235 14 44MB
English Pages 287 [288] Year 2023
Computational Intelligence for Data Analysis (Volume 2) Machine Intelligence for Internet of Medical Things: Applications and Future Trends Edited by Mariya Ouaissa
Moulay Ismail University Meknes Morocco
Mariyam Ouaissa
Moulay Ismail University Meknes Morocco
Zakaria Boulouad Hassan II University Casablanca Morocco
Inam Ullah Khan
Kings College London London, United Kingdom
& Sailesh Iyer
Rai School of Engineering Rai University Ahmedabad India
Computational Intelligence for Data Analysis (Volume 2) Machine Intelligence for Internet of Medical Things: Applications and Future Trends Editors: Mariya Ouaissa, Mariyam Ouaissa, Zakaria Boulouad, Inam Ullah Khan and Sailesh Iyer ISSN (Online): 2810-9465 ISSN (Print): 2810-9457 ISBN (Online): 978-981-5080-44-5 ISBN (Print): 978-981-5080-45-2 ISBN (Paperback): 978-981-5080-46-9 ©2023, Bentham Books imprint. Published by Bentham Science Publishers Pte. Ltd. Singapore. All Rights Reserved. First published in 2023.
BSP-EB-PRO-9789815080445-TP-268-TC-15-PD-20230511
BENTHAM SCIENCE PUBLISHERS LTD.
End User License Agreement (for non-institutional, personal use) This is an agreement between you and Bentham Science Publishers Ltd. Please read this License Agreement carefully before using the ebook/echapter/ejournal (“Work”). Your use of the Work constitutes your agreement to the terms and conditions set forth in this License Agreement. If you do not agree to these terms and conditions then you should not use the Work. Bentham Science Publishers agrees to grant you a non-exclusive, non-transferable limited license to use the Work subject to and in accordance with the following terms and conditions. This License Agreement is for non-library, personal use only. For a library / institutional / multi user license in respect of the Work, please contact: [email protected].
Usage Rules: 1. All rights reserved: The Work is the subject of copyright and Bentham Science Publishers either owns the Work (and the copyright in it) or is licensed to distribute the Work. You shall not copy, reproduce, modify, remove, delete, augment, add to, publish, transmit, sell, resell, create derivative works from, or in any way exploit the Work or make the Work available for others to do any of the same, in any form or by any means, in whole or in part, in each case without the prior written permission of Bentham Science Publishers, unless stated otherwise in this License Agreement. 2. You may download a copy of the Work on one occasion to one personal computer (including tablet, laptop, desktop, or other such devices). You may make one back-up copy of the Work to avoid losing it. 3. The unauthorised use or distribution of copyrighted or other proprietary content is illegal and could subject you to liability for substantial money damages. You will be liable for any damage resulting from your misuse of the Work or any violation of this License Agreement, including any infringement by you of copyrights or proprietary rights.
Disclaimer: Bentham Science Publishers does not guarantee that the information in the Work is error-free, or warrant that it will meet your requirements or that access to the Work will be uninterrupted or error-free. The Work is provided "as is" without warranty of any kind, either express or implied or statutory, including, without limitation, implied warranties of merchantability and fitness for a particular purpose. The entire risk as to the results and performance of the Work is assumed by you. No responsibility is assumed by Bentham Science Publishers, its staff, editors and/or authors for any injury and/or damage to persons or property as a matter of products liability, negligence or otherwise, or from any use or operation of any methods, products instruction, advertisements or ideas contained in the Work.
Limitation of Liability: In no event will Bentham Science Publishers, its staff, editors and/or authors, be liable for any damages, including, without limitation, special, incidental and/or consequential damages and/or damages for lost data and/or profits arising out of (whether directly or indirectly) the use or inability to use the Work. The entire liability of Bentham Science Publishers shall be limited to the amount actually paid by you for the Work.
General: 1. Any dispute or claim arising out of or in connection with this License Agreement or the Work (including non-contractual disputes or claims) will be governed by and construed in accordance with the laws of Singapore. Each party agrees that the courts of the state of Singapore shall have exclusive jurisdiction to settle any dispute or claim arising out of or in connection with this License Agreement or the Work (including non-contractual disputes or claims). 2. Your rights under this License Agreement will automatically terminate without notice and without the
need for a court order if at any point you breach any terms of this License Agreement. In no event will any delay or failure by Bentham Science Publishers in enforcing your compliance with this License Agreement constitute a waiver of any of its rights. 3. You acknowledge that you have read this License Agreement, and agree to be bound by its terms and conditions. To the extent that any other terms and conditions presented on any website of Bentham Science Publishers conflict with, or are inconsistent with, the terms and conditions set out in this License Agreement, you acknowledge that the terms and conditions set out in this License Agreement shall prevail. Bentham Science Publishers Pte. Ltd. 80 Robinson Road #02-00 Singapore 068898 Singapore Email: [email protected]
BSP-EB-PRO-9789815080445-TP-268-TC-15-PD-20230511
CONTENTS FOREWORD ........................................................................................................................................... i PREFACE ................................................................................................................................................ ii OBJECTIVE OF THE BOOK ...................................................................................................... ii ORGANIZATION OF THE BOOK ............................................................................................. ii LIST OF CONTRIBUTORS .................................................................................................................. iv CHAPTER 1 INTERNET OF MEDICAL THINGS & MACHINE INTELLIGENCE ................ Inam Ullah Khan, Mariya Ouaissa, Mariyam Ouaissa and Sarah El Himer INTRODUCTION .......................................................................................................................... LITERATURE STUDY ................................................................................................................. BIG DATA & AI FOR HEALTHCARE ...................................................................................... MACHINE LEARNING CONCEPTS FOR THE INTERNET OF MEDICAL THINGS ..... MACHINE LEARNING-BASED APPLICATIONS FOR IOMT ............................................. Early Prediction of Illnesses ................................................................................................... Healthcare E-Records ............................................................................................................. SAFEGUARDING IOMT FROM CYBER-ATTACKS ............................................................. DoS Attack in IoMT ............................................................................................................... DDoS Attack in IoMT ............................................................................................................ FUTURE ADVANCES & CHALLENGES .................................................................................. CONCLUSION ............................................................................................................................... CONSENT FOR PUBLICATON .................................................................................................. CONFLICT OF INTEREST ......................................................................................................... ACKNOWLEDGEMENT ............................................................................................................. REFERENCES ............................................................................................................................... CHAPTER 2 HEALTH SERVICES AND APPLICATIONS POWERED BY THE INTERNET OF MEDICAL THINGS ......................................................................................................................... Briska Jifrina Premnath and Namasivayam Nalini INTRODUCTION .......................................................................................................................... CONCEPT FOR INTERNET-OF-THINGS-BASED HEALTHCARE .................................... TECHNOLOGIES FOR HEALTHCARE SERVICE ................................................................ Cloud Computing .................................................................................................................... Grid Computing ...................................................................................................................... Big Data .................................................................................................................................. Networks ................................................................................................................................. Ambient Intelligence ............................................................................................................... Augmented Reality ................................................................................................................. Wearable ................................................................................................................................. IOT'S HEALTHCARE BENEFITS ............................................................................................. DIFFICULTIES IN IOMT ............................................................................................................ Confidentiality and Security of Data ...................................................................................... Data Management ................................................................................................................... Scalability, Optimization, Regulation, and Standardization ................................................... Interoperability ........................................................................................................................ Business Viability ................................................................................................................... Power Consumption ................................................................................................................ Environmental Consequences ................................................................................................. SECURITY FOR THE INTERNET OF THINGS IN HEALTHCARE ................................... Security Prerequisites ..............................................................................................................
1 1 3 3 4 4 4 4 5 6 6 7 7 7 7 7 8 11 11 13 14 14 14 14 15 15 15 15 15 16 16 17 17 17 18 18 18 18 19
Security Challenges Memory Limitations .............................................................................. A Threat Model ....................................................................................................................... Attack Types ........................................................................................................................... Security Model Proposal ......................................................................................................... IOMT APPLICATIONS ................................................................................................................ Medical-Smart Technology .................................................................................................... Ingestible Cameras .................................................................................................................. Monitoring of Patients in Real-Time is Number (RTPM) ...................................................... System for Monitoring Cardiovascular Health ....................................................................... Skin Condition Monitoring Systems ....................................................................................... Use of an IoMT Device as a Movement Detector .................................................................. Wearable Sensors for Monitoring your Health from Afar ...................................................... IOMT'S PART IN COVID-19 ....................................................................................................... Technologies Collaborated with IoMT to Develop a Smart Healthcare System at COVID-19 Virtual Reality (VR), Augmented Reality (AR), and Mixed Reality (MR) ........................... CONCLUSION ............................................................................................................................... CONSENT FOR PUBLICATON .................................................................................................. CONFLICT OF INTEREST ......................................................................................................... ACKNOWLEDGEMENT ............................................................................................................. REFERENCES ............................................................................................................................... CHAPTER 3 AN APPROACH TO THE INTERNET OF MEDICAL THINGS (IOMT): IOMTENABLED DEVICES, ISSUES, AND CHALLENGES IN CYBERSECURITY ............................. Usha Nandhini Rajendran and P. Senthamizh Pavai INTRODUCTION .......................................................................................................................... PATIENT-MONITORING SYSTEM IN IOMT ......................................................................... At Home .................................................................................................................................. In-Person ................................................................................................................................. At Community ........................................................................................................................ In-Hospital .............................................................................................................................. POTENTIAL OF THE INTERNET OF MEDICAL THINGS ................................................. Cost-Cutting ............................................................................................................................ Better Care .............................................................................................................................. Patients with a Sense of Empowerment .................................................................................. TAXONOMY OF IOMT SECURITY & PRIVACY (S & P) LAYERS ................................... Thought ................................................................................................................................... Internet .................................................................................................................................... Middleware ............................................................................................................................. Android ................................................................................................................................... Marketing ................................................................................................................................ CYBERSECURITY ATTACKS IN EACH LAYER .................................................................. WAYS TO SOLVE IOMT ISSUES .............................................................................................. Inventory of Assets ................................................................................................................. Policy for Strong Passwords ................................................................................................... Multi-Factor Authentication (MFA) ....................................................................................... Segmentation of the Network ................................................................................................. Updates to Security Patches .................................................................................................... Monitoring of Network Traffic ............................................................................................... Encryption ............................................................................................................................... System for Detecting Intrusions .............................................................................................. SECURITY AND PRIVACY IN IOMT .......................................................................................
19 19 20 20 20 20 20 20 21 21 21 21 21 23 23 24 25 25 25 25 31 31 33 33 34 34 35 35 35 35 35 36 36 38 38 38 38 39 41 41 41 41 41 42 42 42 42 43
CHALLENGES OF IOMT’S ........................................................................................................ STEPS TO IMPROVE DEVICE SECURITY ............................................................................. CONCLUSION ............................................................................................................................... CONSENT FOR PUBLICATON .................................................................................................. CONFLICT OF INTEREST ......................................................................................................... ACKNOWLEDGEMENT ............................................................................................................. REFERENCES ...............................................................................................................................
43 44 45 45 45 45 45
CHAPTER 4 INTERNET OF MEDICAL THINGS IN CLOUD EDGE COMPUTING .............. G. Sumathi, S. Rajesh, R. Ananthakumar and K. Kartheeban INTRODUCTION .......................................................................................................................... MEDICAL INTERNET OF THINGS .......................................................................................... IOMT ARCHITECTURE .............................................................................................................. IOMT TECHNOLOGIES .............................................................................................................. Radio Frequency Identification (RFID) .................................................................................. Wireless Sensor Network (WSN) ........................................................................................... MIDDLEWARE ............................................................................................................................. IOMT APPLICATIONS ................................................................................................................ IOMT IN CLOUD .......................................................................................................................... IOMT CLOUD ARCHITECTURE .............................................................................................. HEALTHCARE SERVICE LAYER ............................................................................................ SERVICE-MANAGEMENT-LAYER .......................................................................................... USER LAYER ................................................................................................................................. IOMT CLOUD TECHNOLOGIES .............................................................................................. Cloud Computing .................................................................................................................... Big Data .................................................................................................................................. Artificial Intelligence (AI) ...................................................................................................... IOMT CLOUD APPLICATIONS ................................................................................................. IOMT EDGE CLOUD ................................................................................................................... IOMT EDGE-CLOUD ARCHITECTURE ................................................................................. IOMT EDGE CLOUD TECHNOLOGIES .................................................................................. Edge Computing ..................................................................................................................... Computational Offload ........................................................................................................... IOMT EDGE CLOUD APPLICATIONS .................................................................................... CONCLUSION & FUTURE WORK ........................................................................................... CONSENT FOR PUBLICATON .................................................................................................. CONFLICT OF INTEREST ......................................................................................................... ACKNOWLEDGEMENT ............................................................................................................. REFERENCES ...............................................................................................................................
47
CHAPTER 5 SURVEY OF IOMT INTERFERENCE MITIGATION TECHNIQUES FOR WIRELESS BODY AREA NETWORKS (WBANS) .......................................................................... Izaz Ahmad, Muhammad Abul Hassan, Inam Ullah Khan and Farhatullah INTRODUCTION .......................................................................................................................... Difference Between WBAN vs. WSN Concerning IoMT ...................................................... Wireless Sensor Network (WSN) ........................................................................................... WBAN ARCHITECTURE ............................................................................................................ WBAN APPLICATIONS ............................................................................................................... Rehabilitation and Therapy ..................................................................................................... Wearable Health Monitoring System ...................................................................................... Disaster Aid Network ............................................................................................................. TECHNOLOGIES ..........................................................................................................................
47 49 49 51 51 52 53 53 54 54 55 55 55 56 56 56 56 57 57 57 58 58 59 59 60 61 61 61 61 64 64 66 67 67 68 68 69 69 69
Bluetooth ................................................................................................................................. Low Energy Bluetooth ............................................................................................................ ZigBee ..................................................................................................................................... IEEE 802.11 ..................................................................................................................................... IEEE 802.15.4 .................................................................................................................................. IEEE 802.15.6 .................................................................................................................................. TECHNIQUES AND COMPARISON ......................................................................................... CONCLUSION ............................................................................................................................... CONSENT FOR PUBLICATON .................................................................................................. CONFLICT OF INTEREST ......................................................................................................... ACKNOWLEDGEMENT ............................................................................................................. REFERENCES ...............................................................................................................................
69 69 70 70 70 71 71 79 79 79 79 79
CHAPTER 6 ARTIFICIAL INTELLIGENCE-BASED IOT APPLICATIONS IN FUTURE PANDEMICS ........................................................................................................................................... Tarun Virmani, Anjali Sharma, Ashwani Sharma, Girish Kumar and Meenu Bhati ....... INTRODUCTION .......................................................................................................................... IOT AND AI IN HEALTH CARE ................................................................................................ IOT AND AI: APPLICATIONS ................................................................................................... AI AND IOT-ENABLED REMOTE SCREENING .................................................................... Patients and IoT ...................................................................................................................... IoT for Doctors ....................................................................................................................... IoT in Hospitals ....................................................................................................................... Diagnosis ................................................................................................................................. MONITORING AND CONTROL OF EPIDEMIC VIA ML-BASED IOT ............................. Drug Discovery and Vaccine Research .................................................................................. Applicability of AI-Enabled System ....................................................................................... FUTURE PANDEMIC PREDICTION ....................................................................................... CONCLUSION ............................................................................................................................... CONSENT FOR PUBLICATON .................................................................................................. CONFLICT OF INTEREST ......................................................................................................... ACKNOWLEDGEMENT ............................................................................................................. REFERENCES ...............................................................................................................................
83 83 84 87 87 89 93 93 93 93 95 96 97 98 99 100 100 100 100
CHAPTER 7 CYBER SECURE AIOT APPLICATIONS IN FUTURE PANDEMICS ................ Maria Nawaz Chohan and Sana Nawaz Chohan INTRODUCTION .......................................................................................................................... LITERATURE STUDY ................................................................................................................. ARTIFICIAL INTERNET OF THINGS APPLICATIONS FOR HEALTHCARE ............... H-AIoT Based Hardware ........................................................................................................ H-AIoT Based Software ......................................................................................................... Communication/Routing Protocols ......................................................................................... UAV’s/Drones in the Healthcare Industry .............................................................................. Wearable AI-IoT Sensors ....................................................................................................... AI-IoT-Based Monitoring System .......................................................................................... Detection of Cyber-Attacks in IoMT ...................................................................................... Machine Learning Techniques for COVID-19 ....................................................................... Industry 5.0 for Smart Healthcare Systems ............................................................................ Industry 5.0 Related Challenges ............................................................................................. Using Flying Vehicles in Health Industry .............................................................................. Future Challenges ................................................................................................................... CONCLUSION ...............................................................................................................................
107 107 107 109 110 110 110 110 110 111 111 111 112 113 114 114 115 115
CONSENT FOR PUBLICATION ................................................................................................ CONFLICT OF INTEREST ......................................................................................................... ACKNOWLEDGEMENT ............................................................................................................. REFERENCES ...............................................................................................................................
115 115 115 116
CHAPTER 8 MACHINE LEARNING SOLUTION FOR ORTHOPEDICS: A COMPREHENSIVE REVIEW .............................................................................................................. Muhammad Imad, Muhammad Abul Hassan, Shah Hussain Bangash and Naimullah INTRODUCTION .......................................................................................................................... LITERATURE REVIEW .............................................................................................................. METHODOLOGY ......................................................................................................................... CONCLUSION ............................................................................................................................... CONSENT FOR PUBLICATION ................................................................................................ CONFLICT OF INTEREST ......................................................................................................... ACKNOWLEDGEMENT ............................................................................................................. REFERENCES ...............................................................................................................................
120 120 120 124 133 133 133 133 134 134
CHAPTER 9 A REVIEW OF MACHINE LEARNING APPROACHES FOR IDENTIFICATION OF HEALTH-RELATED DISEASES ............................................................... Muhammad Yaseen Ayub, Farman Ali Khan, Syeda Zillay Nain Zukhraf and Muhammad Hamza Akhlaq INTRODUCTION .......................................................................................................................... Supervised Learning ............................................................................................................... Unsupervised Learning ........................................................................................................... MOTIVATION ............................................................................................................................... LITERATURE STUDY ................................................................................................................. Heart Diseases Detection ........................................................................................................ Lung Diseases Detection ......................................................................................................... Skin Disease Detection ........................................................................................................... Brain Diseases Detection ........................................................................................................ Liver Diseases Detection ........................................................................................................ ALGORITHMS EXPLOITED FOR VARIOUS DISEASES DETECTION ............................ TOOLS AND LIBRARIES USED FOR DISEASE DETECTION ............................................ CONCLUSION AND FUTURE TRENDS ................................................................................... CONSENT FOR PUBLICATON .................................................................................................. CONFLICT OF INTEREST ......................................................................................................... ACKNOWLEDGEMENT ............................................................................................................. REFERENCES ............................................................................................................................... CHAPTER 10 MACHINE LEARNING IN DETECTION OF DISEASE: SOLUTIONS AND OPEN CHALLENGES ........................................................................................................................... Tayyab Rehman, Noshina Tariq, Ahthasham Sajid and Muhammad Hamza Akhlaq INTRODUCTION .......................................................................................................................... MACHINE LEARNING APPROACHES ................................................................................... Supervised Learning (SL) ....................................................................................................... Unsupervised Learning ........................................................................................................... Reinforcement Learning (RL) ................................................................................................. Data Mining (DM) .................................................................................................................. DETECTION OF DISEASE BY USING DIFFERENT MACHINE-LEARNING CLASSIFICATION ........................................................................................................................ CHRONIC DISEASE: DETECTION OF HEART DISEASE ................................................... Naive Bayes (NB) ...................................................................................................................
137 137 138 139 140 140 140 142 142 143 143 144 145 146 147 147 147 147 149 149 149 150 151 151 152 152 153 154 154
Decision Tree (DT) ................................................................................................................. K-Nearest Neighbor (K-NN) .................................................................................................. Issues and Challenges ............................................................................................................. CHRONIC DISEASE: DETECTION OF DISEASE BREAST CANCER ............................... CAD System ........................................................................................................................... Deep Learning ......................................................................................................................... Machine-Learning Techniques ............................................................................................... Convolutional Neural Network Model (CNN) ....................................................................... Issues and Challenges ............................................................................................................. CHRONIC DISEASE: DETECTION OF DISEASE DIABETES ............................................. Logistic Regression (LR) ........................................................................................................ Random Forest Classifier (RFC) ............................................................................................ Gradient Boosted Trees (GBT) ............................................................................................... Weighted Ensemble Model (WEM) ....................................................................................... Issues and Challenges ............................................................................................................. CHRONIC DISEASE: DETECTION OF LIVER DISEASE .................................................... Data Selection and Pre-Processing ......................................................................................... Feature Selection ..................................................................................................................... Classification Algorithm ......................................................................................................... Supervised Learning and Unsupervised Learning .................................................................. Performance Metrics Analysis ................................................................................................ Predicted Results ..................................................................................................................... Issues and Challenges ............................................................................................................. SEASONAL DISEASE: DETECTION OF DENGUE DISEASE .............................................. Issues and Challenges ............................................................................................................. SEASONAL DISEASE: DETECTION OF COVID-19 DISEASE ............................................ Issues and Challenges ............................................................................................................. CONCLUSION ............................................................................................................................... CONSENT FOR PUBLICATON .................................................................................................. CONFLICT OF INTEREST ......................................................................................................... ACKNOWLEDGEMENT ............................................................................................................. REFERENCES ............................................................................................................................... CHAPTER 11 BREAKTHROUGH IN MANAGEMENT OF CARDIOVASCULAR DISEASES BY ARTIFICIAL INTELLIGENCE IN HEALTHCARE SETTINGS ............................................. Lakshmi Narasimha Gunturu, Girirajasekhar Dornadula and Raghavendra Naveen Nimbagal INTRODUCTION .......................................................................................................................... MATERIALS AND METHODS ................................................................................................... ALGORITHMS USED IN CARDIOVASCULAR DISEASES ................................................. K-Nearest Neighbour (KNN) .................................................................................................. Artificial Neural Network (ANN) ........................................................................................... Decision Tree (DT) ................................................................................................................. Logistic Regression (LR) ........................................................................................................ AdaBoost (AB) ....................................................................................................................... Support Vector Machine (SVM) ............................................................................................. RESULTS AND DISCUSSION ..................................................................................................... Impact of AI on Echocardiography (ECG) ............................................................................. Role of AI on Magnetic Resonance Imaging (MRI) .............................................................. Use of AI on Cardiac Computed Tomography (CT) .............................................................. Impact of AI on Electrocardiography .....................................................................................
155 155 156 156 157 157 157 158 158 160 161 161 161 161 162 162 163 164 164 165 165 165 165 166 167 167 169 169 170 170 170 170 177 177 181 182 183 183 183 184 184 184 185 186 186 186 187
CHALLENGES ............................................................................................................................... CONCLUSION ............................................................................................................................... CONSENT FOR PUBLICATON .................................................................................................. CONFLICT OF INTEREST ......................................................................................................... ACKNOWLEDGEMENT ............................................................................................................. REFERENCES ............................................................................................................................... CHAPTER 12 SMART CANE: OBSTACLE RECOGNITION FOR VISUALLY IMPAIRED PEOPLE BASED ON CONVOLUTIONAL NEURAL NETWORK ................................................ Adnan Hussain, Bilal Ahmad and Muhammad Imad INTRODUCTION .......................................................................................................................... LITERATURE STUDY ................................................................................................................. MATERIALS AND METHODS ................................................................................................... Dataset Description ................................................................................................................. Methods ................................................................................................................................... Ultrasonic Sensors .................................................................................................................. Visual Sensor .......................................................................................................................... Buzzer Sensor ......................................................................................................................... Jumper Wires .......................................................................................................................... Breadboard .............................................................................................................................. Bus Strip .................................................................................................................................. Socket Strip ............................................................................................................................. Power Bank ............................................................................................................................. Earphone/Speaker ................................................................................................................... Traditional Cane ...................................................................................................................... Smart/Modern Cane ................................................................................................................ Proposed Device Architecture ................................................................................................ Deep Convolutional Neural Network ..................................................................................... EXPERIMENTAL RESULTS ANALYSIS ................................................................................. CONCLUSION ............................................................................................................................... CONSENT FOR PUBLICATON .................................................................................................. CONFLICT OF INTEREST ......................................................................................................... ACKNOWLEDGEMENT ............................................................................................................. REFERENCES ............................................................................................................................... CHAPTER 13 A SURVEY ON BRAIN-COMPUTER INTERFACE AND RELATED APPLICATIONS ..................................................................................................................................... Krishna Pai, Rakhee Kallimani, Sridhar Iyer, B. Uma Maheswari, Rajashri Khanai and Dattaprasad Torse INTRODUCTION .......................................................................................................................... RELATED WORKS ....................................................................................................................... APPLICATIONS OF BCI ............................................................................................................. ISSUES AND CHALLENGES, AND FUTURE DIRECTIONS ................................................ Neuro-Psycho-Physiological Issues ........................................................................................ Technical Issues ...................................................................................................................... Ethical Issues .......................................................................................................................... CONCLUSION ............................................................................................................................... CONSENT FOR PUBLICATON .................................................................................................. CONFLICT OF INTEREST ......................................................................................................... ACKNOWLEDGEMENT ............................................................................................................. REFERENCES ...............................................................................................................................
189 190 190 190 190 190 194 194 196 197 197 197 199 199 200 200 200 201 201 201 201 202 202 203 204 204 207 208 208 208 208 210 210 212 218 221 221 221 221 223 223 223 223 223
CHAPTER 14 DATA AUGMENTATION WITH IMAGE FUSION TECHNIQUES FOR BRAIN TUMOR CLASSIFICATION USING DEEP LEARNING ................................................... Tarik Hajji, Ibtissam Elhassani, Tawfik Masrour, Imane Tailouloute and Mouad Dourhmi INTRODUCTION .......................................................................................................................... BACKGROUND ............................................................................................................................. Deep Learning ......................................................................................................................... Data Augmentation ................................................................................................................. Image Fusion ........................................................................................................................... Related Work .......................................................................................................................... METHODOLOGY ......................................................................................................................... Dataset ..................................................................................................................................... Deep Learning Approach with Classical Data Augmentation ................................................ Data Pre-Processing for the Model ......................................................................................... Generation of many Manipulated Images from a Directory ................................................... Design of the Model Architecture ........................................................................................... Convolution Layer ......................................................................................................... Pooling Layer ................................................................................................................ Flatten Layer ................................................................................................................. Dense Layer .................................................................................................................. Learning and Same Parameters ............................................................................................... Data Augmentation: A Comparative Study ............................................................................ Data Augmentation with Image Fusion .................................................................................. Auto-Encoder Architecture ..................................................................................................... RESULTS AND DISCUSSION ..................................................................................................... CNN Result without Data Augmentation ............................................................................... CNN Result with Data Augmentation Automatic Generator ................................................. CNN Result-Based DA using IF with BWT ........................................................................... CNN Result-Based DA using IF with Auto-Encoder Proposed Approach ............................ CONCLUSION ............................................................................................................................... CONSENT FOR PUBLICATON .................................................................................................. CONFLICT OF INTEREST ......................................................................................................... ACKNOWLEDGEMENT ............................................................................................................. REFERENCES ............................................................................................................................... CHAPTER 15 CONVERGENCE TOWARDS BLOCKCHAIN-BASED PATIENT HEALTH RECORD AND SHARING SYSTEM: EMERGING ISSUES AND CHALLENGES ..................... Mahendra Kumar Shrivas, Ashok Bhansali, Hoshang Kolivand and Kamal Kant Hiran INTRODUCTION .......................................................................................................................... METHODOLOGY ......................................................................................................................... THE HEALTHCARE DATA MANAGEMENT SYSTEM (HDMS) OR HEALTHCARE INFORMATION SYSTEM (HIS) ................................................................................................. Evolution of the Health Data Management System (HDMS)/ Health Information System (HIS) ....................................................................................................................................... Current Status, Issues, and Challenges ................................................................................... Huge Data Volume and Velocity and Paper-Based Record Keeping ........................... Interoperability and Data Sharing ................................................................................ Data Governance, Manipulation, Privacy and Security Threats .................................. BLOCKCHAIN FUNDAMENTALS, CONCEPTS, AND FEATURES ................................... Blockchain Categorization ......................................................................................................
229 229 231 231 231 232 233 234 235 235 236 236 237 237 238 238 238 238 240 241 241 242 242 243 243 244 245 245 245 245 245 248 249 249 250 250 251 251 251 252 252 253
Evolution of Blockchain Technology ..................................................................................... Transaction in Blockchain Network ....................................................................................... HEALTHCARE AND BLOCKCHAIN ....................................................................................... Blockchain-Based Systems Models for the HDMS/HIS ........................................................ How Does Blockchain Address Security, Consensus, and Data Manipulation Issues? ......... How Does Blockchain Address Privacy Issues? .................................................................... Preventing PHR, EMR Manipulation, and Sharing Records Securely using Blockchain ...... ISSUES, CHALLENGES, AND RECOMMENDATIONS ........................................................ CONCLUSION ............................................................................................................................... CONSENT FOR PUBLICATON .................................................................................................. CONFLICT OF INTEREST ......................................................................................................... ACKNOWLEDGEMENT ............................................................................................................. REFERENCES ...............................................................................................................................
253 254 255 255 258 259 259 260 261 262 262 262 262
SUBJECT INDEX .................................................................................................................................... 2
i
FOREWORD The COVID-19 pandemic has shed light on the importance of having a more efficient healthcare system. In the era of Industry 4.0, Artificial Intelligence and the Internet of Things have introduced themselves today as must-have technologies in almost every sector, including healthcare. This book introduces an emerging yet interesting concept, the Internet of Medical Things (IoMT). It refers to an infrastructure of highly connected healthcare devices that can communicate and share data to optimize different medical actions and interventions. The book goes further into suggesting solutions that can provide better performance and security for the Internet of Medical Things. Moreover, this book presents different successful case studies of combinations of IoMT with Artificial Intelligence and their application in different medical use cases, such as preventing future pandemics, optimizing brain tumor detection, obstacle detection for visually impaired patients, etc. In its last chapter, this book offers an opening to further development in the IoMT area by exploring the possibilities offered by Blockchain technology in securing medical data. This book aspires to provide a relevant reference for students, researchers, engineers, and professionals working in the IoMT area, particularly those interested in grasping its diverse facets and exploring the latest advances in IoMT.
Yassine Maleh Sultan Moulay Slimane University Beni Mellal Morocco
ii
PREFACE The growing development in the field of computing has encouraged the integration of a variety of sophisticated devices inside houses and facilities. These devices communicate with each other to help users in particular situations and according to their needs, such as safety, comfort, and even health. The devices form an object connection environment known as the Internet of Things (IoT). Healthcare professionals are now embracing the Internet of Medical Things (IoMT), which refers to a connected infrastructure of devices and software applications that can communicate with various healthcare IT systems. One of these technologies — Remote Patient Monitoring — is commonly used for the treatment and care of patients. Often associated with the IoT, Artificial Intelligence (AI) opens the field of possibilities in the medical area, in particular, by allowing the development of new diagnostic and interpretation tools of exceptional reliability and by assessing the large volumes of data that can be generated through the networks by sensors and users.
OBJECTIVE OF THE BOOK The objective of this book is to focus on how to use IoT, AI and Machine Learning (ML), to keep patients safe and healthy and, at the same time, to empower physicians to deliver superlative care. This book discusses the applications, opportunities, and future trends of machine intelligence in the medical domain, including both basic and advanced topics. This book provides core principles, algorithms, protocols, emerging trends, security problems, and the latest e-healthcare services findings. It also includes deep feedforward networks, regularization, optimization algorithms, convolutional networks, sequence modeling, practical methodology, and how they can be used to provide better solutions to healthcare-related issues.
ORGANIZATION OF THE BOOK Chapters 1-3: The authors introduce the concept of the Internet of Medical Things (IoMT), its roles, challenges, and the opportunities it may present to the healthcare system. Chapter 4: The authors present a cloud-edge-based IoMT architecture and discuss the performance optimization it may provide in the context of Medical Big Data. Chapter 5: The authors provide a comprehensive survey on different IoMT interference mitigation techniques for Wireless Body Area Networks (WBANs). Chapters 6 and 7: The authors explore the possibilities that Artificial Intelligence and the Internet of Things can provide to prevent future pandemics. Chapters 8-10: The authors provide a comprehensive review of the newest Machine Learning based solutions in different medical areas. Chapter 11: The authors go through the latest discoveries in curing cardiovascular diseases by implementing Artificial Intelligence in healthcare settings.
iii
Chapter 12: The authors propose a Deep Learning based solution to optimize obstacle recognition for visually impaired patients. Chapter 13: The authors provide a survey on the latest breakthroughs in Brain-Computer Interfaces and their applications. Chapter 14: The authors propose a solution to optimize the performance of Deep Learning for brain tumor detection. Chapter 15: The authors explore the possibilities that Blockchain may offer inpatient data management.
Mariya Ouaissa Moulay Ismail University Meknes Morocco Mariyam Ouaissa Moulay Ismail University Meknes Morocco Zakaria Boulouad Hassan II University Casablanca Morocco Inam Ullah Khan Kings College London London, United Kingdom & Sailesh Iyer Rai School of Engineering Rai University Ahmedabad India
iv
List of Contributors Adnan Hussain
Islamia College University Peshawar, Peshawar, Pakistan
Anjali Sharma
Pharmacovigilance Expert, Uttar Pradesh, India
Ashok Bhansali
Department of Computer Science Engineering, O. P. Jindal University, India
Ashwani Sharma
School of Pharmaceutical Sciences, MVN University, Palwal, Haryana, India
B. Uma Maheswari
Department of Computer Science and Engineering, Amrita School of Engineering, Bengaluru, Amrita Vishwa Vidyapeetham, KA, India
Bilal Ahmad
Islamia College University Peshawar, Peshawar, Pakistan
Briska Jifrina Premnath
Department of Biochemistry and Biotechnology, Faculty of Science, Annamalai University, Tamilnadu, India
Dattaprasad Torse
Department of Computer Science and Engineering, KLE Dr. M.S. Sheshgiri College of Engineering and Technology, Udyambag, Belagavi, KA, India
Farhatullah
School of Automation, China University of Geosciences, Wuhan, China
Farman Ali Khan
COMSATS University Islamabad, Attock Campus, Pakistan
G. Sumathi
Department of IT, Kalasalingam Institute of Technology Krishnankoil, Tamilnadu, India
Girirajasekhar Dornadula
Department of Pharmacy Practice, Annamacharya College of Pharmacy, Rajampeta, India
Girish Kumar
School of Pharmaceutical Sciences, MVN University, Palwal, Haryana, India
Hoshang Kolivand
Faculty of Engineering and Technology, Liverpool John Moores University, Liverpool, England
Ibtissam Elhassani
Artificial Intelligence for Engineering Sciences Team (IASI), Laboratory of Mathematical Modeling, Simulation and Smart Systems (L2M3S), ENSAM, Moulay Ismail University, Meknes, Morocco
Imane Tailouloute
Artificial Intelligence for Engineering Sciences Team (IASI), Laboratory of Mathematical Modeling, Simulation and Smart Systems (L2M3S), ENSAM, Moulay Ismail University, Meknes, Morocco
Inam Ullah Khan
Kings College London, London, United Kingdom
Izaz Ahmad
Department Computing and Technology, Abasyn University, Peshawar, Pakistan
K. Kartheeban
Department of CSE, Kalasalingam Academy of Research and Education, Krishnankoil, Tamilnadu, India
Kamal Kant Hiran
Faculty of IT and Design, Aalborg University, Copenhagen, Denmark
Krishna Pai
Department of Electronics and Communication Engineering, KLE Dr. M.S. Sheshgiri College of Engineering and Technology, Udyambag, Belagavi, KA, India
v Lakshmi Narasimha Gunturu
Scientimed Solutions Private Limited, Mumbai, Maharashtra, India
Dourhmi Mouad
Artificial Intelligence for Engineering Sciences Team (IASI), Laboratory of Mathematical Modeling, Simulation and Smart Systems (L2M3S), ENSAM, Moulay Ismail University, Meknes, Morocco
Mahendra Kumar Shrivas
Department of Computer Science Engineering, O. P. Jindal University, Raigarh, India
Maria Nawaz Chohan
National Defence University, Islamabad, Pakistan
Mariya Ouaissa
Moulay Ismail University, Meknes, Morocco
Meenu Bhati
School of Pharmaceutical Sciences, MVN University, Palwal, Haryana, India
Muhammad Abul Hassan
Abasyn University Peshawar, Peshawar, Pakistan
Muhammad Hamza Akhlaq
Allama Iqbal Open University, Islamabad, Pakistan
Muhammad Imad
Abasyn University, Peshawar, Pakistan
Muhammad Yaseen Ayub
COMSATS University Islamabad, Attock Campus, Pakistan
Naimullah
Abasyn University, Peshawar, Pakistan
Namasivayam Nalini
Department of Biochemistry and Biotechnology, Faculty of Science, Annamalai University, Tamilnadu, India
P. Senthamizh Pavai
Faculty of Education, Dr. M.G.R. Educational and Research Institute, Chennai, India
R. Ananthakumar
Department of CSE, Kalasalingam Institute of Technology, Krishnankoil, Tamilnadu, India
Raghavendra Naveen Nimbagal
Department of Pharmaceutics, Sri Adichunchanagiri College of Pharmacy, Adichunchanagiri University, Karnataka 571418, India
Rajashri Khanai
Department of Electronics and Communication Engineering, KLE Dr. M.S. Sheshgiri College of Engineering and Technology, Udyambag, Belagavi, KA, India
Rakhee Kallimani
Department of Electrical and Electronics Engineering, KLE Dr. M.S. Sheshgiri College of Engineering and Technology, Udyambag, Belagavi, KA, India
S. Rajesh
Department of CSE, Mepco Schlenk Engineering College, Sivakasi, Tamil Nadu, India
Sajid Ahthasham
Department of Computer Science, Faculty of ICT, BUITEMS, Quetta, Baluchistan, Pakistan
Sana Nawaz Chohan
Foundation University Institute of Rehabilitation Sciences, Islamabad, Pakistan
Sarah El Himer
Sidi Mohammed Ben Abdellah University, Fez, Morocco
Shah Hussain Bangash
Abasyn University, Peshawar, Pakistan
vi Sridhar Iyer
Department of Electronics and Communication Engineering, KLE Dr. M.S. Sheshgiri College of Engineering and Technology, Udyambag, Belagavi, KA, India
Syeda Zillay Nain Zukhraf
KIOS Research and Innovation Center of Excellence, Department of Electrical and Computer Engineering, University of Cyprus, Nicosia, Cyprus
Tarik Hajji
Artificial Intelligence for Engineering Sciences Team (IASI), Laboratory of Mathematical Modeling, Simulation and Smart Systems (L2M3S), ENSAM, Moulay Ismail University, Meknes, Morocco
Tawfik Masrour
Artificial Intelligence for Engineering Sciences Team (IASI), Laboratory of Mathematical Modeling, Simulation and Smart Systems (L2M3S), ENSAM, Moulay Ismail University, Meknes, Morocco
Tarun Virmani
School of Pharmaceutical Sciences, MVN University, Palwal, Haryana, India
Tayyab Rehman
Faculty of Computing, SZABIST, Islamabad, Pakistan,
Usha Nandhini Rajendran
Faculty of Education, Dr. M.G.R. Educational and Research Institute, Chennai, India
Computational Intelligence for Data Analysis, 2023, Vol. 2, 1-10
1
CHAPTER 1
Internet of Medical Things & Machine Intelligence Inam Ullah Khan1, Mariya Ouaissa2,*, Mariyam Ouaissa2 and Sarah El Himer3 Kings College London, London, United Kingdom Moulay Ismail University, Meknes, Morocco 3 Sidi Mohammed Ben Abdellah University, Fez, Morocco 1 2
Abstract: Recently, the internet of medical things has been the widely utilized approach to interconnect various machines. While, IoT in combination with machine intelligence, has given new directions to the healthcare industry. Machine intelligence techniques can be used to promote healthcare solutions. The merger of IoT in medical things is a completely advanced approach. The intelligent behavior of machines provides accurate decisions, which greatly helps medical practitioners. For real-time analysis, artificial intelligence improves accuracy in different medicinal techniques. The use of telemedicine has increased so much due to COVID-19. Gathering unstructured data where the concept of electronic databases should be used in the health care industry for advancement. Big data and cyber security play an important role in IoMT. An intrusion detection system is used to identify cyber-attacks which helps to safeguard the entire network. This article provides a detailed overview of the internet of medical things using machine intelligence applications, future opportunities, and challenges. Also, some of the open research problems are highlighted, which gives insight into information about the internet of medical things. Different applications are discussed related to IoMT to improve communication standards. Apart from that, the use of unmanned aerial vehicles is increased, which are mostly utilized in rescuing and sending medical equipment from one place to another.
Keywords: Big Data, IoMT, IoT, Machine Intelligence, UAVs. INTRODUCTION With the development of IoT, the healthcare industry is revolutionized, where a massive amount of data can be transferred from one place to another. Therefore, IoMT is introduced to connect medical devices, which can improve decisionmaking process. Data resource management is the central point of discussion in IoMT. However, machine learning techniques enhance the accuracy level, which has shifted researcher’s attention to secure communication betweenQRGH Corresponding author Mariya Ouaissa: Moulay Ismail University, Meknes, Morocco; Tel: +212 604483006; E-mail: [email protected] *
Mariya Ouaissa, Mariyam Ouaissa, Zakaria Boulouad, Inam Ullah Khan, Sailesh Iyer (Eds.) All rights reserved-© 2023 Bentham Science Publishers
2 Computational Intelligence for Data Analysis, Vol. 2
Khan et al.
COVID-19 is considered the most dangerous virus which affects the respiratory system. Machine intelligence-based techniques can be used for the effective treatment of viruses. AI and machine learning refers to solving big problems related to healthcare [1, 2]. An advance in the healthcare industry has enhanced standards for different stakeholders like patients, doctors and researchers. Therefore, AI, machine learning, cyber security, big data and 5G can be integrated with IoMT to give optimal solutions [3]. Sensors and a high level of hardware equipment are needed to modify healthcare industry processes. Due to that, IoT with medical is quite helpful [4]. Integrated applications are designed using AI models for disease treatments [5, 6]. This research work provides novel ideas related to the internet of medical things using artificial intelligence, machine learning, and meta-heuristic search optimization to give directions to researchers. However, the major contribution of this article is as below: ● ● ● ● ●
Big data and AI-designed techniques for the health care industry. Machine learning concepts for IoMT. Applications for IoMT. Safeguarding IoMT from cyber-security attacks. Future advances and challenges.
The contribution points are fully incorporated in the rest of the paper, which gives a detailed overview of IoMT applications, challenges, AI, big data and machine learning techniques. Fig. (1) illustrates the concept of tele-medicine, which was mostly used during COVID-19 for online consultation with medical doctors. Also, the whole architectural view of tele-medicine is presented.
Fig. (1). Internet of medical things (tele-medicine).
Internet of Medical Things
Computational Intelligence for Data Analysis, Vol. 2 3
LITERATURE STUDY IoT has interconnected patients, doctors and related equipment’s in the healthcare industry. However, different sensors are used to collect, send and manage the data. Various applications of IoT are utilized which use to tackle COVID-19. Therefore, IoT connects each and everything while machine learning techniques diagnose diseases [7]. The Internet of health things has changed the dynamics in health management. Federated learning is a new concept that is sub part of machine learning. This novel technique takes data in central servers and local devices, which makes the data safer in contrast with other traditional methods [8]. However, local models must be properly updated using 5G communication networks [9]. Protocols are designed while integrating 5G networks with federated learning [10]. Lightweight protocols are proposed to bring trust between two nodes in IoMT [11, 12]. COVID-19 has disturbed our daily life routine, where we have to maintain social distancing and make people aware of vaccination [13, 14]. While, the health status of patients and much more information can be easily made available due to various advancements in IoT, cyber security, big data, AI and machine learning [15 - 19]. In addition, UAVs are widely used during COVID-19 to send medical equipment and rescue operations. Also, tele-medicine is nowadays commonly utilized by doctors to properly advise patients. Therefore, secure routing is needed between nodes. BIG DATA & AI FOR HEALTHCARE Artificial intelligence is making life easier for humans. Due to advanced communication technologies, life has become more comfortable. AI merger with big data has solved major problems related to healthcare. Electronic healthcare records are quite helpful in improving tumors to optimize treatment methods [20]. The healthcare industry is based on data that should be authentic. Due to this, decision-making process will be quite efficient. The data usually flow from patients to doctors where to share possible information to give possible treatment. However, in traditional methods, the data or record cannot be preserved for a long period of time. While, digitalization utilizing big data analytics and artificial intelligence has improved the standards of technological equipment’s [21].
4 Computational Intelligence for Data Analysis, Vol. 2
MACHINE LEARNING MEDICAL THINGS
CONCEPTS
Khan et al.
FOR
THE
INTERNET
OF
Machine learning techniques have played an important role in developing data and records more efficiently. Moreover, machine learning presented novel ideas to digitalize and improve computer-aided technologies [22]. The growth of mobile devices and the merger of the Internet of things has changed the dynamics of the medical field. The quality of service has solved unprecedented problems and given mobile medical services, including tele-medicines. Using mobile technology in the form of web services, patient consultation with the doctor is quite easy nowadays [23, 24]. Health applications like “Good Doctor Online” is utilized more for tele-medicine during COVID-19. Mobile data will be quite useful in the future to observe the needs of patients. While, for remote treatment, health expert systems provide many services like audio, video and short messages. Therefore, initial data will be taken from patients with the help of technology, and based on that, the doctor will give possible guidance [25]. In healthcare engineering, following machine learning techniques can be utilized to solve problems. ● ●
Classification & Clustering Prediction & Anomaly identification
MACHINE LEARNING-BASED APPLICATIONS FOR IOMT In addition, machine learning has many related applications that have improved healthcare standards. Some of the applications are as below: Early Prediction of Illnesses For better treatment, three diseases, coronavirus, heart disease and diabetes model, are formulated in the form of an android mobile application. A supervised learning model is utilized for training the database on real-time data, which shows results in android applications. Therefore, logistic regression is used for the early prediction of illness [26]. Healthcare E-Records In the fourth industrial revolution, an electronic health record is an optimal way to save data. Medical data is so importantthat intruders can try to hijack the entire system, which is very dangerous for the patient. Medical healthcare systems are changed with the passage of time, but had vulnerabilities as well. Due to cyber-
Internet of Medical Things
Computational Intelligence for Data Analysis, Vol. 2 5
attacks, falsification, data loss, end-to-end delay, jitter and modification of packets are possible, which endanger the life of patient. Therefore, an intelligent & secure electronic health record system is designed to reduce cost and improve trust using blockchain [27]. Apart from that, many more applications are available, or either scientists or engineers are working to improve the healthcare industry, which is as below: ● ● ●
Humanoid robot surgery Novel disease breakthrough Drug discovery and clinical trials
Table 1 describes the applications related to IoMT. Table 1. Internet of medical things applications. Author
Year
Application
Chao et al. [33]
2014
Smart Medical Nursing System
Lei et al. [34]
2012
Smart Hospital
Harshal et al. [35]
2016
IoT-based Smart Medical System
Rashmi et al. [36]
2016
Medical Healthcare System
Micheal et al. [37]
2016
Medical Bot
SAFEGUARDING IOMT FROM CYBER-ATTACKS Wearable devices will be utilized in the near future to collect data from humans and send information to the external device. The data or information can either be viewed by using a laptop or computer or might be mobile. Emergency response sensors can be deployed at home or the workplace to monitor emergencies and send locations to the base station. The entire information can be viewed through web applications or mobile devices. Mobile applications are developed to facilitate patients for proper and timely medication. For this purpose, sometimes an alert message is sent to the mobile device of the patient. Internet of medical things architecture is divided into three phases which include: ● ● ● ● ●
Wireless Body Area Sensor Networks Wireless Personal Area Network Wireless Wide Area Network Medical Server (Laptop or Mobile) Emergency Service Provider
6 Computational Intelligence for Data Analysis, Vol. 2
Khan et al.
Due to the extensive use of wireless communication technologies, the ratio of cyber-attacks is increasing. While, the medical industry is the main focus of intruders to take information or change the data [28]. Some of the commonly used attacks are discussed as under: DoS Attack in IoMT Denial of service is also called third-party attack. Due to this attack, the intruder tries to take full control of the network or either modify data packets. Broadcasting illegal data packets in a continuous pattern affects the process of the Internet of things [29]. DDoS Attack in IoMT Distributed denial of service attack is considered the most dangerous threat to every network. In DDoS, the entire cluster or group attacks another to send false information to create congestion and take control. DDoS is the extended version of a DoS attack [30]. Some other attacks also disrupt the entire communication in IoMT, which are as under: ● ● ● ● ● ●
Routing Attack False Alarm Attack Unbalancing High Accuracy Attack Overhead Attack Data Traffic Attack Unwanted Nodes Attack
Fig. (2). Different types of attacks.
Internet of Medical Things
Computational Intelligence for Data Analysis, Vol. 2 7
Fig. (2) describes different types of security attacks which can affect overall communication in IoMT. FUTURE ADVANCES & CHALLENGES Machine learning techniques like supervised, unsupervised, and reinforcement have greatly changed the healthcare industry. Quality of service is improved due to technological inventions to modernize the healthcare field. ML in health technologies has given deep information to improve treatment methods. While, due to cognitive computing identification of different diseases can be easily treated in a better way. Drug discovery, medical imaging, behavioral medicines, a database for healthcare records and data collection will be improved with the help of machine learning algorithms [31]. However, AI-based telemedicine will give new directions to the entire world [32]. Moreover, unmanned aerial vehicles are widely used during COVID-19 to send medical equipment and maintain physical distancing. With the usage of new technology, some of the problems to a normal human being exist. People should update their knowledge about every subject, especially healthcare, as information technology has improved. CONCLUSION The role of machine learning algorithms has a direct impact on the internet of medical things. Therefore, machines are trained to give an optimal prediction for illness or disease. AI-based tools have advanced the way of treatment. Also, IoMT is a combination of the internet of things and the medical field. This research paper gives knowledge related to big data, AI, machine learning, cyber-attacks, and various applications related to IoMT. In addition, future directions and challenges are incorporated, which is very much helpful for engineers, scientists, researchers and practitioners. AI, ML and meta-heuristic search algorithms will be deployed in the future to enhance communication within IoMT. CONSENT FOR PUBLICATON Declared none. CONFLICT OF INTEREST The author declares no conflict of interest, financial or otherwise. ACKNOWLEDGEMENT Declared none.
8 Computational Intelligence for Data Analysis, Vol. 2
Khan et al.
REFERENCES [1]
S.H. Sharma, and H. Ahteshamul, "Artificial Intelligence, Machine Learning & Internet of Medical Things (IoMT) for COVID-19 & Future Pandemics: An Exploratory Study", IEEE Smart Cities Newsletter, IEEE Smart Cities, 2021.
[2]
H.K. Bharadwaj, A. Agarwal, V. Chamola, N.R. Lakkaniga, V. Hassija, M. Guizani, and B. Sikdar, "A Review on the Role of Machine Learning in Enabling IoT Based Healthcare Applications", IEEE Access, vol. 9, pp. 38859-38890, 2021. [http://dx.doi.org/10.1109/ACCESS.2021.3059858]
[3]
R.F. Mansour, A.E. Amraoui, I. Nouaouri, V.G. Díaz, D. Gupta, and S. Kumar, "Artificial Intelligence and Internet of Things Enabled Disease Diagnosis Model for Smart Healthcare Systems", IEEE Access, vol. 9, pp. 45137-45146, 2021. [http://dx.doi.org/10.1109/ACCESS.2021.3066365]
[4]
F.F. Gong, X.Z. Sun, J. Lin, and X.D. Gu, "Primary exploration in establishment of China’s intelligent medical treatment", Mod. Hos Manag., vol. 11, no. 2, pp. 28-29, 2013.
[5]
K.H. Abdulkareem, M.A. Mohammed, A. Salim, M. Arif, O. Geman, D. Gupta, and A. Khanna, "Realizing an effective COVID-19 diagnosis system based on machine learning and IOT in smart hospital environment", IEEE Internet Things J., vol. 8, no. 21, pp. 15919-15928, 2021. [http://dx.doi.org/10.1109/JIOT.2021.3050775] [PMID: 35782183]
[6]
A.A. Mutlag, M. Khanapi Abd Ghani, M.A. Mohammed, M.S. Maashi, O. Mohd, S.A. Mostafa, K.H. Abdulkareem, G. Marques, and I. de la Torre Díez, "MAFC: Multi-agent fog computing model for healthcare critical tasks management", Sensors (Basel), vol. 20, no. 7, p. 1853, 2020. [http://dx.doi.org/10.3390/s20071853] [PMID: 32230843]
[7]
B. Shahab, S. Ardabili, A. Yarahmadi, B. Pahlevanzadeh, A.K. Kiani, A. Beheshti, H.A. Rokny, I. Dehzangi, and A. Mosavi. "Machine Learning and Internet of Medical Things for Handling COVID19: Meta-Analysis," 2022. [http://dx.doi.org/10.20944/preprints202302.0083.v1]
[8]
M.A. Rahman, M.S. Hossain, M.S. Islam, N.A. Alrajeh, and G. Muhammad, "Secure and Provenance Enhanced Internet of Health Things Framework: A Blockchain Managed Federated Learning Approach", IEEE Access, vol. 8, pp. 205071-205087, 2020. [http://dx.doi.org/10.1109/ACCESS.2020.3037474] [PMID: 34192116]
[9]
S. Zhou, H. Huang, W. Chen, Z. Zheng, and S. Guo, "PIRATE?: A blockchain-based secure framework of distributed machine learning in 5G networks", arXiv:1912.07860, vol. 1, no. 1, pp. 1-8, 2019.
[10]
G. Hua, L. Zhu, J. Wu, C. Shen, L. Zhou, and Q. Lin, "Blockchain-based federated learning for intelligent control in heavy haul railway", IEEE Access, vol. 8, pp. 176830-176839, 2020. [http://dx.doi.org/10.1109/ACCESS.2020.3021253]
[11]
M. Kamal, and M. Tariq, "Light-weight security and blockchain based provenance for advanced metering infrastructure", IEEE Access, vol. 7, pp. 87345-87356, 2019. [http://dx.doi.org/10.1109/ACCESS.2019.2925787]
[12]
M. Kamal, G. Srivastava, and M. Tariq, "Blockchain-Based Lightweight and Secured V2V Communication in the Internet of Vehicles," IEEE Transactions on Intelligent Transportation Systems, vol. 22, no. 7, pp. 3997-4004, 2021. [http://dx.doi.org/10.1109/TITS.2020.3002462]
[13]
M.A. Rahman, M.S. Hossain, N.A. Alrajeh, and N. Guizani, "B5G and explainable deep learning assisted healthcare vertical at the edge—COVID-19 perspective", IEEE Netw., vol. 34, no. 4, pp. 98105, 2020. [http://dx.doi.org/10.1109/MNET.011.2000353]
[14]
Y. Abdulsalam, and M. S. Hossain, "COVID-19 Networking Demand: An Auction-Based Mechanism
Internet of Medical Things
Computational Intelligence for Data Analysis, Vol. 2 9
for Automated Selection of Edge Computing Services," IEEE Transactions on Network Science and Engineering, vol. 9, no. 1, pp. 308-318, 2022. doi: 10.1109/TNSE.2020.3026637. [http://dx.doi.org/10.1109/TITS.2020.3002462] [15]
X. Yang, T. Zhang, C. Xu, and M.S. Hossain, "Automatic visual concept learning for social event understanding", IEEE Trans. Multimed., vol. 17, no. 3, pp. 346-358, 2015. [http://dx.doi.org/10.1109/TMM.2015.2393635]
[16]
Y. Xiao, G. Shi and M. Krunz, "Towards ubiquitous AI in 6G with federated learning", arXiv:2004.13563, vol. 1, no. 1, pp. 1-7, 2020, [online] Available: https://arxiv.org/abs/2004.13563
[17]
M.F. Alhamid, M. Rawashdeh, H. Al Osman, M.S. Hossain, and A. El Saddik, "Towards contextsensitive collaborative media recommender system", Multimedia Tools Appl., vol. 74, no. 24, pp. 11399-11428, 2015. [http://dx.doi.org/10.1007/s11042-014-2236-3]
[18]
S. Qian, T. Zhang, C. Xu, M.S. Hossain, "Social event classification via boosted multimodal supervised latent Dirichlet allocation", ACM Trans. Multimedia Comput. Commun. Appl., vol. 11, no. 2, pp. 27.1-27.22, 2015.
[19]
I.U. Khan, I.M. Qureshi, M.A. Aziz, T.A. Cheema, and S.B.H. Shah, "Smart IoT Control-Based Nature Inspired Energy Efficient Routing Protocol for Flying Ad Hoc Network (FANET)", IEEE Access, vol. 8, pp. 56371-56378, 2020. [http://dx.doi.org/10.1109/ACCESS.2020.2981531]
[20]
Y.C. Yang, S.U. Islam, A. Noor, S. Khan, W. Afsar, and S. Nazir, "Influential usage of big data and artificial intelligence in healthcare", Comput. Math. Methods Med., vol. 2021, pp. 1-13, 2021. [http://dx.doi.org/10.1155/2021/5812499] [PMID: 34527076]
[21]
N. Mehta, A. Pandit, and S. Shukla, "Transforming healthcare with big data analytics and artificial intelligence: A systematic mapping study", J. Biomed. Inform., vol. 100, p. 103311, 2019. [http://dx.doi.org/10.1016/j.jbi.2019.103311] [PMID: 31629922]
[22]
A. Khare, M. Jeon, I.K. Sethi, and B. Xu, "Machine learning theory and applications for healthcare", J. Healthc. Eng., vol. 2017, pp. 1-2, 2017. [http://dx.doi.org/10.1155/2017/5263570] [PMID: 29090076]
[23]
Y. Zhang, L. Sun, H. Song, and X. Cao, "Ubiquitous WSN for healthcare: recent advances and future prospects", IEEE Internet Things J., vol. 1, no. 4, pp. 311-318, 2014. [http://dx.doi.org/10.1109/JIOT.2014.2329462]
[24]
F. Xiao, and W. Ding, "Divergence measure of Pythagorean fuzzy sets and its application in medical diagnosis", Appl. Soft Comput., vol. 79, pp. 254-267, 2019. [http://dx.doi.org/10.1016/j.asoc.2019.03.043]
[25]
H.N. Bischel, L. Caduff, S. Schindelholz, T. Kohn, and T.R. Julian, "Health risks for sanitation service workers along a container-based urine collection system and resource recovery value chain", Environ. Sci. Technol., vol. 53, no. 12, pp. 7055-7067, 2019. [http://dx.doi.org/10.1021/acs.est.9b01092] [PMID: 31082211]
[26]
N. Kumar, N. Narayan Das, D. Gupta, K. Gupta, and J. Bindra, "Efficient automated disease diagnosis using machine learning models", J. Healthc. Eng., vol. 2021, pp. 1-13, 2021. [http://dx.doi.org/10.1155/2021/9983652] [PMID: 34035886]
[27]
A.M. Hasib, K. Tamzid, I. Chowdhury, S. Sakib, M.M. Khan, N. Alsufyani, A. Alsufyani, and S. Bourouis, "Electronic Health Record Monitoring System and Data Security Using Blockchain Technology", Secur. Commun. Netw., no. 2022, 2022. [http://dx.doi.org/10.1155/2022/2366632]
[28]
M. Anandarajan, and S. Malik, "Protecting the Internet of medical things: A situational crimeprevention approach", Cogent Med., vol. 5, no. 1, 2018. [http://dx.doi.org/10.1080/2331205X.2018.1513349]
10 Computational Intelligence for Data Analysis, Vol. 2
Khan et al.
[29]
A. Abdollahi, and M. Fathi, "An intrusion detection system on ping of death attacks in IoT networks", Wirel. Pers. Commun., vol. 112, no. 4, pp. 2057-2070, 2020. [http://dx.doi.org/10.1007/s11277-020-07139-y]
[30]
I.U. Khan, A. Abdollahi, R. Alturki, M.D. Alshehri, M.A. Ikram, H.J. Alyamani, and S. Khan, "Intelligent Detection System Enabled Attack Probability Using Markov Chain in Aerial Networks", Wirel. Commun. Mob. Comput., vol. 2021, pp. 1-9, 2021. [http://dx.doi.org/10.1155/2021/1542657]
[31]
P. Singh, K. S. Singh and H. Vikram Singh, "Machine Learning for Healthcare: A Survey & its Algorithm for the Security of Medical Images," 2021 10th International Conference on Internet of Everything, Microwave Engineering, Communication and Networks (IEMECON), pp. 01-05, 2021. [http://dx.doi.org/10.1109/IEMECON53809.2021.9689076]
[32]
S. Ellahham, N. Ellahham, and M.C.E. Simsekler, "Application of artificial intelligence in the health care safety context: opportunities and challenges", Am. J. Med. Qual., vol. 35, no. 4, pp. 341-348, 2020. [http://dx.doi.org/10.1177/1062860619878515] [PMID: 31581790]
[33]
C.H. Huang, and K.W. Cheng, "RFID technology combined with IoT application in medical nursing system", Bulletin of Networking, Computing, Systems, and Software, vol. 3, no. 1, pp. 20-24, 2014.
[34]
H. Arbat, S. Choudhary, and K. Bala, “IOT smart health band,” Imperial Journal of Interdisciplinary Research, vol. 2, no. 5, 2016.
[35]
L. Yu, Y. Lu, and X. Zhu, “Smart hospital based on internet of things,” Journal of Netwroks, vol. 7, no. 10, 2012. [http://dx.doi.org/10.4304/jnw.7.10.1654-1661]
[36]
R. Singh, “A proposal for mobile E-Care health service system using IOT for Indian scenario,” Journal of Network Communications and Emerging Technologies, vol. 6, no. 1, 2016.
[37]
M. Fischer and M. Lam, “From books to bots: Using medical literature to create a chat bot,” In Proc. First Workshop on IoT-enabled Healthcare and Wellness Technologies and Systems, June 30, 2016.
Computational Intelligence for Data Analysis, 2023, Vol. 2, 11-30
11
CHAPTER 2
Health Services and Applications Powered by the Internet of Medical Things Briska Jifrina Premnath1 and Namasivayam Nalini1,* Department of Biochemistry and Biotechnology, Faculty of Science, Annamalai University, Tamil Nadu, India 1
Abstract: The traditional healthcare system model is now out of date. As the digital era progresses, new advanced technologies and service platforms are highly demanded. The Internet of Medical Things (IoMT), a subset of the Internet of Things, is one such technology. The Internet of Things (IoT) is a network of wireless, interconnected, and linked digital devices that can collect, send and store data without requiring human-tohuman or human-to-computer interaction. Understanding how established and emerging IoT technologies help health systems provide safe and effective care is more important than ever. For example, the rapid spread of Coronavirus disease (COVID-19) has alerted the entire healthcare system. The Internet of Medical Things (IoMT) has dramatically improved the situation, and COVID-19 has inspired scientists to create a new 'Smart' healthcare system focused on early diagnosis, prevention of spread, education, and treatment to facilitate living in the new normal. This paper provides an overview of the IoMT design and how cloud storage technology can help healthcare applications. This chapter should assist researchers in considering previous applications, benefits, problems, challenges, and threats of IoMT in the healthcare field and the role of IoMT in the COVID-19 pandemic. This review will be helpful to researchers and professionals in the field, allowing them to recognize the enormous potential of IoT in the medical world.
Keywords: Applications, Benefits, Challenges, COVID-19, Healthcare, IoMT, IoT, Medical, Threats. INTRODUCTION Significant changes have taken place in the healthcare industry over the last few years. One crucial factor in this change is the use of new information technology across the business right now. Hospitals and nursing homes need help from many different IT service platforms and cutting-edge technology to meet the growing healthcare demand. The Internet of Medical Things, or IoMT, is one of the most Corresponding author Namasivayam Nalini: Department of Biochemistry and Biotechnology, Faculty of Science, Annamalai University, Tamil Nadu, India; E-mail: [email protected] *
Mariya Ouaissa, Mariyam Ouaissa, Zakaria Boulouad, Inam Ullah Khan, Sailesh Iyer (Eds.) All rights reserved-© 2023 Bentham Science Publishers
12 Computational Intelligence for Data Analysis, Vol. 2
Premnath and Nalini
commonly used technologies in the healthcare field today. A subset of the Internet of Things is the Internet of Medical Things [1]. The term “Internet of Things” refers to a network of physical things, or “Things,” meant to communicate with each other through the Internet. Ashton first talked about the Internet of Things in 1999. Since then, it has overgrown, with about 10 billion connected devices today and an estimated 25 billion by 2025 [2]. Taking care of a person's physical, mental, or emotional well-being is called “health care,” usually done by trained and licensed professionals like doctors and other healthcare workers. There are not enough doctors, nurses, or hospital beds because there has been much growth in the population, and a lot more people are getting sick. Scientists who use the latest techniques and methodologies develop new medicine and healthcare trends every day. Researchers have recently focused their attention on the Internet of Things (IoT) because of its popularity as a perfect solution for healthcare systems that do not put much pressure on them [3]. Today, healthcare and modern technology businesses, especially healthcare systems, play a big part in our lives. The main goal of integrating technology into healthcare systems is to make it easier for patients and caregivers to communicate with each other. This will make medical devices and services more efficient and easier to get. The Internet of Medical Things (IoMT) has been essential in monitoring healthcare from afar (RHM). Wearable sensors and devices are often used to get data on patients remotely and store it in cloud databases. The Internet of Things (IoT) is used primarily for this. These data can be used by caregivers right away for analysis and planning [4]. IoMT consists of three main parts: the device layer [Body Sensor Network (BSN)], the fog layer, and the cloud service. The main goal of the device layer (sensing layer) is to build an effective and accurate sensing technology that can collect different types of health-based data. Communication technologies like Bluetooth, RFID (NFC), Wi-Fi, IrDA, UWB, and ZIGBEE help the IoMT system build network solutions and infrastructures. In the cloud layer (data layer), the data is processed and kept safe and sound. Furthermore, the cloud gets the patient's data to analyze, process, and store it. Healthcare workers can then use such data [5 - 12]. The IoMT is a group of medical strategies connected to networks. People can connect their smart glasses, head-mounted devices, belt-worn clothes, smartwatches, woven clothes, and smart wristbands to Wi-Fi, Bluetooth, or the Internet to get information about their health. Diagnostic machines such as ultrasonography, MRI machines, infusion pumps, ventilators, and X-ray machines in healthcare facilities use IOMT technologies. These IOMT wearable devices can
Internet of Medical Things
Computational Intelligence for Data Analysis, Vol. 2 13
be used to keep track of people's health of all ages. They are usually easy to wear and use. IOMT devices are used in applications and software, such as remote data analytics, medical assistance, operations augmentation, medicine monitoring, and accounting systems [13]. Remote Health Monitoring (RHM) is a way to track a person's health data regularly. People's heart rate, temperature, blood pressure, physical activities, and dietary habits are all monitored. The cloud sends health data wirelessly to both patients and caregivers. So, IoMT can make real-time, quick, remote, and trustworthy decisions for various disorders. This process generates many data, which is then analyzed and monitored. Due to the hectic pace of today's lives, most people do not go to the doctor regularly. In addition, healthcare costs are rising, and governments spend a lot of money on healthcare each year. People in Europe and the United States also prefer to get their health care at home rather than in a hospital. These problems can be solved if real-time healthcare monitoring can be done from afar and in real-time. The use of wearable gadgets and sensors to provide continuous monitoring for patients and the elderly has received a lot of interest [14 - 25]. Imagine a world where billions of things are connected through IP (Internet Protocol) networks and have built-in intelligence, communication, sensing, and actuation abilities. This is called the Internet of Things (IoT). Our current Internet has moved a lot away from hardware-based options (computers, fibers, and Ethernet connections) and toward market-based ones (such as apps) (Facebook, Amazon) [26]. This chapter will look at the technologies that make up IoMT and the benefits, problems, security concerns, and ways that IoMT can be used in healthcare. IoMT's role in COVID-19 is also addressed briefly. CONCEPT FOR INTERNET-OF-THINGS-BASED HEALTHCARE It is meant to allow for a wide range of types and services, each of which has a different set of Medicare solutions. There is not yet a complete list of IoT services in healthcare. Health care services can be hard to tell apart from other solutions or applications in some cases. It also looks at how potentially building blocks can be used in general service. In Medicare settings, IoT frameworks and protocols have been updated a little to make them work better. Simple, safe, low-power and quick discovery of new devices and services can be made and done quickly. There are many subtopics under the term “health service” that deal with future and emerging health services [3]. Fig. (1) illustrates the concept of IoMT in health care.
14 Computational Intelligence for Data Analysis, Vol. 2
Premnath and Nalini
Fig. (1). Concept of IoMT in healthcare.
TECHNOLOGIES FOR HEALTHCARE SERVICE The Internet of Things (IoT) healthcare services use many different technologies. However, the proposed system explains a few technologies at the heart of medical assistance. Cloud Computing If cloud computing is implemented, many cloud computing benefits come with IoT-based healthcare services. These features include always-on access to shared resources, services offered in response to network requests, and operations that meet the company's needs. Grid Computing Grid computing, also known as cluster computing, is the foundation of cloud computing. Grid computing makes it possible for all healthcare networks to deal with the computational limitations of sensor nodes spread out all over a patient's body. Big Data Big data collect essential health data from many different sensors all over our bodies. It also gives us tools that help us make the health diagnostic and monitoring stages and procedures more consistent.
Internet of Medical Things
Computational Intelligence for Data Analysis, Vol. 2 15
Networks The physical architecture of IoT-based healthcare networks includes a variety of networks ranging from short-range communications, such as 6LoWPANs, WPANs, WBANs, WLANs, and WSNs, to long-range communications, such as any type of cellular network. In developing low-power communication protocols and medical sensor devices, technologies such as Bluetooth Low Energy, radio frequency identification (RFID), and near-field communication (NFC) play a role. Ambient Intelligence Ambient Intelligence continuously monitors human activities and executes any action required for the recognized event. It is critical because it is constantly monitoring what individuals do. Augmented Reality Augmented reality emphasizes surgery and other standard patient examinations more than other technologies. Wearable Sensors like pulse, body temperature, pulse oximetry, and respiration rate are wearable because they are planned with soft, smooth, and easily wearable features. It is also easy to use and apply to the body of anyone. Wearables have many benefits, like gamification, connected data, and healthcare communities [3]. IOT'S HEALTHCARE BENEFITS As with any new technology, there are some downsides to the Internet of Things, but it is primarily advantageous. The Internet of Things is causing a significant shift in medical health care. The Internet of Things applications and technologies have changed the world in ways that people did not think possible in the 1990s. Internet of Things created a significant change in people's communication with each other on the Internet. It led to the creation of many new and demanding fields, most notably in the area of medical things. Most people follow and utilize it due to its accurate and straightforward features. So, it bridges the gap between doctors, patients, and healthcare services. Thanks to the Internet of Things, health care workers can now do their jobs better and more quickly, with less effort and intelligence. This use of IoT in the medical field has been perfect for patients. IoT is also straightforward to use. People who use IoT can benefit from the following points:
16 Computational Intelligence for Data Analysis, Vol. 2 ● ● ● ● ● ● ● ● ● ● ● ●
● ● ● ● ● ● ● ●
Premnath and Nalini
Makes life easier. Cuts down on the cost of healthcare. Improves the health of patients. Diseases are taken care of in real-time. Improves the quality of life. Improves the end experience of users. The quality of care for patients is better. Expenses will be cut. The goal is for people to live longer with better health. Care and prevention of diseases should be done in the best way possible. The progress of children and parents who are old is kept on track. When there is a significant change in a patient's health, an automatic alarm is sent to many people, saving lives and time. Internet of Things (IoT) resources, as well as other IoT gadgets. There are rules about how to take medicine. Family members will be kept up to date on the patient's condition. The ability to make money. The ease of use. Making the best use of energy, including time. Think about money as an example, and you will get the idea. Doctors who use the Internet of Things to provide on-demand medical care [27].
DIFFICULTIES IN IOMT Before IOMT was widely used, many problems and consequences needed to be solved. These include data privacy and security, data management, scalability and upgradeability, legislation, interoperability and cost-effectiveness. Confidentiality and Security of Data One of the most challenging things to do in some applications is ensuring that cyber security works well in healthcare monitoring systems. The security of the massive amount of sensitive health data sent between systems is still an open question. Akhbarifar et al. (2020) came up with a way to make it easier to monitor people's health from afar in a cloud-based IoMT setting. Lin et al. (2021) suggested using a smartcard system for a single sign-on (SC-UCSSO) for telemedicine that protects patients' privacy while increasing security and performance. Covi-Chain uses “block-chain technology” to solve security and privacy problems while not exposing data and increasing storage capacity [28 31].
Internet of Medical Things
Computational Intelligence for Data Analysis, Vol. 2 17
Data Management Data management is the ability to get at, combine, control, and manage data information flow to make sense. There are numerous methods for providing computer programs with only the information they require while concealing other information. These methods include data anonymization, data integration, and data synchronization. Scalability, Optimization, Regulation, and Standardization Scalability is a term that refers to how well a healthcare device can change with changes in the environment. As a result, a highly scalable system keeps connected devices in sync and can run efficiently while taking advantage of available resources. Having a system that can be easily changed is better now and in the future. Continuous changes and improvements in IoMT technology have kept the devices up to date. The Healthcare Industry needs to ensure that the wide range of IoMT-based devices that use communication protocols, data aggregation, and gateway interfaces made by many different manufacturers or suppliers who follow standard design principles and protocols are safe. People must do it like the Information Technology and Innovation Foundation (IETF), the European Telecommunications Standards Institute (ETSI), and the Internet Protocol for Smart Things (IPSO). There must also be a check on IoMT devices that record EMR data. Scholars, organizations, and standard-setting bodies must work together to do this. People who want to use IoMT devices quickly cannot because of laws like the Health Information Technology for Economic and Clinical Health Act, HIPAA compliance, and the general data protection regulation (GDPR) [2]. Interoperability The standards that allow different industries to make apps are other. In addition, the wide range of equipment and data makes it hard to use, primarily because of differences between operators. Interoperability is a problem because it is hard to move data between different IoMT systems with different abilities. As a result, the development of standard interfaces becomes essential, especially for applications that allow for cross-organizational and cross-system communication. People in the IoMT world exchange information, which creates many data. People have to figure out how to manage the data and control the interconnecting devices while keeping energy costs in check [32].
18 Computational Intelligence for Data Analysis, Vol. 2
Premnath and Nalini
Business Viability Financial stress has spread to many people, businesses, and even whole organizations during the time of COVID-19, which has made it hard for people to use IoMT. So, cost-effectiveness is a big issue that needs to take into concern. The costs of producing, installing, and using an IoMT system must be fair. In IoMT-based systems, many medical devices and sensors are connected and used. Unfortunately, these are expensive to keep and upgrade. So, both the manufacturer and the end-user must pay for it. As a result, low-maintenance sensors with low start-up costs could help make more IoMT devices and more people use them. Power Consumption Another factor impeding the widespread adoption of IoMT devices is power consumption. Most IoMT devices are battery-powered, and if a sensor is mounted, the battery needs to be changed often or must use a high-power battery. Making self-sufficient healthcare devices or connecting the IoMT system to renewable energy systems should be our main focus which can also help mitigate the global issue of energy [2]. Environmental Consequences IoMT systems use a variety of biological sensors to do different things, as discussed above. These are formed by mixing semiconductors with earth metals and other potentially dangerous compounds. As a result, regulatory bodies monitor and control the manufacturing of sensors. More research should be directed toward designing and manufacturing biodegradable sensors [33]. SECURITY FOR THE INTERNET OF THINGS IN HEALTHCARE When it comes to technology, the Internet of Things is becoming a technology that can be used in almost all areas and at virtually any point. However, the Internet of Things is becoming more and more popular in the medical field, both in technology and protection. The Internet of Things will eventually lead to many technology prescriptions for doctors and patients to follow and monitor their health. As a result, security could be a big problem. The device has the patient's medical information, and hackers can quickly get into the sensors today. As a result, the main challenge for IoT-based healthcare services is to stop hackers from protecting and storing the patient's critical unique data on a database through encryption requirements.
Internet of Medical Things
Computational Intelligence for Data Analysis, Vol. 2 19
Security Prerequisites Security requirements in medicine focus on the security of patient data and the communication system between IoT gateways and care advisors, where the possibility of data trapping exists. Confidentiality: Confidentiality means an unapproved person cannot access a patient's medical data. Patients' data can only be accessed by people who have the proper permissions. Approved people can only handle or use the data. Even eavesdroppers will not be able to get the information. For the benefit of both patients and healthcare providers, this is all about keeping things private for them. Authentication: Many things go into authentication, like giving each sensor device a unique ID to identify another person communicating with them and unlock the device when they want to open it. Availability: When an IoT healthcare service needs to be available for as long as possible, either by local and global servers or by cloud services that can identify and authorize people who need the sensor, even if they try to shut down the service. Data Freshness: Data freshness makes sure that it has all of the new data added to it every day. Moreover, it ensures that no data or messages are repeated or stored in its database and that no advisory repeats old statements. Security Challenges Memory Limitations Most IoT-enabled healthcare devices have memory used by system software, an application binary, or an embedded operating system to store data. As a result, memory may not be enough to run complex security procedures. A Threat Model There are many ways for hackers to get into IoT health services, networks, and devices because there are many places to attack. The three plotlines in IoT healthcare make the services safe for everyone, no matter how they are used. The scenarios make it possible for cloud networks, native networks, and cloud services to grow. When this is done, it makes sure that there is a way to report between IoT networks and devices and between IoT networks and cloud services. The last scenario is very clear about the hardware and software limitations of the device.
20 Computational Intelligence for Data Analysis, Vol. 2
Premnath and Nalini
Attack Types Attacks based on information disruptions and attacks based on host and network properties are in this taxonomy. Security Model Proposal It is not very safe to use IoT in healthcare as it can be affected by any security problems since it has a patient's personal information. It is more challenging to figure out and diagnose all attacks, and threats, making it more difficult. The dynamic properties are made to achieve security goals. Security plans should be able to identify and respond to unexpected or unanticipated issues that may arise [3]. IOMT APPLICATIONS Medical-Smart Technology Devices and kits that are smart are included in this. These are now being used by paramedics to help people who need medical care and help right away. One way to do this would be to use medical drones. Because medical drones are the fastest to get to an emergency scene, they were first used to help people with heart attacks. Paramedics might not respond quickly enough because they might get stuck in traffic and not be able to respond quickly; this encourages intelligent medical robots to do surgery in a hospital. VR/AR and AI-based medical technologies were also used for other medical reasons. AI-based medical technologies are also used to ensure that they are more accurate [34 - 40]. Ingestible Cameras A patient can swallow these cutting-edge, low-cost capsules (in-vivo/in-vitro) to get real-time, live monitoring of their internal organs to detect chronic diseases and cancer early. Data recorder capsules, ingestible endoscopic optical scanning devices for endoscopy, and ingestible hydrogel devices are just three examples among many ingestible devices. X-ray or camera capsules, tracking/recording systems, and diagnostics toolkits are used to look at ingestible devices when they are being looked at [41 - 44]. Monitoring of Patients in Real-Time is Number (RTPM) As they use smart devices so much in their daily lives, this is a new trend growing among the younger generation, including millennials and Generation Y. As it turns out, RTPM monitors a patient's health from afar. Sensors attached to the patient's body are used to watch their health, either through a home health
Internet of Medical Things
Computational Intelligence for Data Analysis, Vol. 2 21
telehealth system or a telecare monitoring system. If a person wants to keep an eye on your fitness level, blood sugar level, respiration rate, and heart rate, he may want to keep track of these things. Many new RTPM trends are now available, like the Apple Watch app for monitoring depression, Apple's Research Kit for Parkinson's disease, and ADAMM Intelligence Asthma Monitoring, which allows monitoring one's body [45 - 51]. System for Monitoring Cardiovascular Health When a person has these systems, they look at their heart rate and pulse wave velocity. Using the Body Cardio Scale, People can measure pulse wave velocity (PWV) to see if they have early peripheral artery disease (PAD) [52]. Skin Condition Monitoring Systems These systems determine how big the wound is and how quickly it will heal. The authors wanted to start a group to talk about mobile telemedicine with smartphones. Ten experts used an experimental mobile app to look at a diabetic person's foot from afar. They used this method to test the platform's viability and usability. Use of an IoMT Device as a Movement Detector These devices are for people who are unable to move. These devices have to track where these groups of people are going. As a result, smart watches and sensors are attached to the patient's clothes, bed, or body to keep an eye on them. It will also help keep track of involuntary actions and give a better idea of controlling medical care. Data from these devices is encrypted from start to finish, giving patients more peace of mind [53]. Wearable Sensors for Monitoring your Health from Afar Since both the user and the patients use this app, it gets much attention. The services are only for making the user's app. In a nutshell, services are for developers, and applications are for users. On the Internet of Things-based healthcare application, applications have been made for patients to watch their health, like apps for oxygen saturation monitoring and rehabilitation systems, blood pressure monitoring, body temperature monitoring, etc. [3]. IOMT'S PART IN COVID-19 By using technology, IoMT can help doctors make more accurate diagnoses, avoid mistakes, and save money on health care by letting patients send their health data to doctors. This is very important in light of recent events because
22 Computational Intelligence for Data Analysis, Vol. 2
Premnath and Nalini
wireless connectivity means that there is no need to go to the doctor in person, which stops the spread of COVID-19. There are a few drawbacks to IoMT, but one of them may be that patients may not trust tech systems with their personal information. Global COVID-19 made IoMT even more essential and helped technology improve, but this is not the only thought. IoMT plays a big part in solving these problems, as medical industry researchers are looking for the best way to screen patients and keep an eye on their symptoms quickly. IoT in 5G, cloud computing, and block-chain powered by artificial intelligence will build a complex and efficient health-tech environment that will be very hard to break down [54]. COVID-19-related research mostly tries to get around the virus's problems. 1. Technical ways to ensure the distance between people. 2. IoMT helps doctors monitor their patients from afar. 3. IoMT-based methods for detecting infections. 4. IoMT provides thermal screening to stop the spread of disease [55]. Cognitive IoT is a new technology that makes it easier to use a limited spectrum. This method is suitable for COVID-19 patient tracking. A part of the Internet of Things called Cognitive IoMT (CIoMT) is the Internet of Things that is smart (IoT). IoMT was used a lot in high-risk places to fight the COVID-19 pandemic. CIoMT has been used in a lot of different ways. 1. Keeping an eye on people in real-time. 2. Monitoring patients who have COVID-19 from afar. 3. The pandemic can be diagnosed quickly. 4. Getting to know patients. 5. Preventing and controlling pandemic sickness. IoMT can also help orthopaedic patients who have COVID-19 infections. COVID-19 patients can get the following services from IoMT: 1. Monitoring of patients.
Internet of Medical Things
Computational Intelligence for Data Analysis, Vol. 2 23
2. Wearable technology. 3. High-quality health care [56, 57]. IoMT was used to stop the spread of the COVID-19 pandemic. Developing more effective solutions using IoMT with other technologies, like artificial intelligence, big data, and block-chain is easier. Many people are working on making IoMT applications that are safe. It is possible to monitor COVID-19 patients in their own homes with IoMT. Many patient parameters, such as blood pressure and heart rate, will be stored and sent over the cloud to healthcare workers, who will be able to see them. It stops the spread of disease and will also be good for people's health. The source of the COVID-19 epidemic was found and tracked with the help of IoMT. A Taiwanese company has made a wearable Internet of Things device that can tell when the temperature is too high and alert the right people. The cloud-based temperature monitoring technology reduces medical staff mistakes on time and the risk of infection for healthcare administrators. The IoMT has been used in a Taiwanese hospital to scan visitors, check their temperature, and see if they are wearing face masks. The AI-enabled IoMT is used in the hospital [58 - 61]. Technologies Collaborated with IoMT to Develop a Smart Healthcare System at COVID-19 During the COVID-19 pandemic, it is crucial to observe the health of many people at all times, both before and after they get sick. The Internet of Medical Things (IoMT) has made it easy for caregivers and patients to use telemedicine to monitor, screen, and treat patients from afar. When there is a global pandemic, intelligent devices powered by IoMT are changing the world, especially in the middle [2]. Virtual Reality (VR), Augmented Reality (AR), and Mixed Reality (MR) Virtual, mixed, and augmented reality can be used for various things, but there are four main types: clinical/therapeutic, entertainment, business/industry, and education/training. By changing reality experiences, VR technology makes a three-dimensional, appealing multimodal world so that people can feel like they are there. It looks like people are “transported to a three-dimensional, life-like world” when they wear a head-mounted display (HMD) that has a very close screen to their face. Virtual reality is good because it uses distraction, extinction learning, cognitive-behavioural principles, gate-control theory, and the spotlight theory of attention to make people pay attention. Virtual reality treats mental health and anxiety disorders and manages strokes, pain, and obesity. It also controls and stops weight gain. Virtual reality can help cancer patients keep track
24 Computational Intelligence for Data Analysis, Vol. 2
Premnath and Nalini
of their treatment by changing how they think and feel. They lessen the psychological symptoms of cancer, which makes the patient's emotional wellbeing better. Immersive VR-induced distant remembrance could help people with mild cognitive impairment deal with their anxiety, according to a study by Yahara et al. in 2021. Palliative care and the unpleasant effects of the pandemic can be lessened without people having to travel. VR can be used for this because it allows people to have video chats and simulate how people would interact in the real world. Computer images are superimposed on top of real-world images. In addition to being a great teaching tool, augmented reality can help people visualize and annotate ideas that are not visible to the naked eye. They can do this by moving around in a virtual world. XR-Health came up with a virtual reality (VR) system that helps people who cannot leave their homes deal with stress and anxiety by doing physical and mental exercises. Immersive VR Education Company made the “Engage” platform, which is used for virtual reality instruction and group work. EON Reality came up with a virtual reality/augmented reality platform that businesses, schools, and governments can use in quarantine situations, and it was made by EON Reality. Also, a virtual reality environment was made to look at the COVID-19 virus's structure, molecular dynamics, enzymes, and proteins. Surgeons can practice and rehearse surgical methods in a controlled environment with haptic features that give them a sense of touch. This kind of training is called “fundamental VR.” When surgeons implanted spinal screws in cadavers, XVision Augmedics used a 3D representation to make it look like X-ray vision. This helped them see the patient's anatomy more clearly. To help anxious people who have mental illnesses, Oxford VR makes people feel less stressed [62 - 69]. CONCLUSION The Internet of Medical Things (IoMT) is a group of medical equipment and applications that communicate with healthcare information technology systems. Medical equipment with Wi-Fi allows IoMT to work. This will enable machines to communicate with each other. When medical devices and software applications connect wirelessly to healthcare information technology systems, they form the Internet of Medical Things (IoMT). This network includes medical devices and software applications. By clicking people, data, and processes through connected medical devices and mobile apps, the goal is to improve patients' health. IoT deals with many data, some of which can be very important in the case of IoMT. Some of this could be about the patient's health, where they live, and what kind of treatment they are getting. This data needs strong security and protection from cyber-attacks, which have become more common recently. This manuscript will expose people to how the Internet of Things (IoT) can help people get better care
Internet of Medical Things
Computational Intelligence for Data Analysis, Vol. 2 25
and how IoMT played a role in the COVID-19 pandemic. This work will help scholars and practitioners understand the enormous benefits of IoT in the medical field and point out important issues with IOMT. Besides, it will also help researchers understand IOT applications in the healthcare industry. CONSENT FOR PUBLICATON Declared none. CONFLICT OF INTEREST The author declares no conflict of interest, financial or otherwise. ACKNOWLEDGEMENT Declared none. REFERENCES [1]
A. Giri, S. Chatterjee, P. Paul, S. Chakraborty, S. Biswas, "Impact of Smart Applications of IoMT (Internet of Medical Things) on Health- Care Domain in India", International Journal of Recent Technology and Engineering (IJRTE), vol. 1, no. 6, pp. 18-25, 2019. [http://dx.doi.org/10.35940/ijrte.D7474.118419]
[2]
R. Dwivedi, D. Mehrotra, and S. Chandra, "Potential of Internet of Medical Things (IoMT) applications in building a smart healthcare system: A systematic review", J. Oral Biol. Craniofac. Res., vol. 12, no. 2, pp. 302-318, 2022. [http://dx.doi.org/10.1016/j.jobcr.2021.11.010] [PMID: 34926140]
[3]
S. Devadharshini, R.J. Anitha, "IoT Based Health Care Services", International Research Journal of Multidisciplinary Technovation (Irjmt), vol. 1, no. 6, pp. 18-25, 2019.
[4]
O. AlShorman, B. AlShorman, M. Al-khassaweneh, F. Alkahtani, “A review of Internet of medical things (IoMT) - based remote health monitoring through wearable sensors: a case study for diabetic patients”, Indonesian Journal of Electrical Engineering and Computer Science, Vol. 20, No. 1, 2020.
[5]
A. Mihovska, and M. Sarkar, Smart connectivity for Internet of things (iot) applications. New Advances in the Internet of Things, 2018, pp. 105-118. [http://dx.doi.org/10.1007/978-3-319-58190-3_7]
[6]
A. Yassine, S. Singh, M.S. Hossain, and G. Muhammad, "IoT big data analytics for smart homes with fog and cloud computing", Future Gener. Comput. Syst., vol. 91, pp. 563-573, 2019. [http://dx.doi.org/10.1016/j.future.2018.08.040]
[7]
J. Mahapatro, "Nagdeo, Sumit Kumar, Mittal, Sumitt Bhutia, Tenzing Thundup, “Selection of Priority in Wireless Body Area Network Sensors using IoT Technology", 5th International Conference for Convergence in Technology, 2019 .
[8]
Y. Miao, G. Wu, C. Liu, M.S. Hossain, and G. Muhammad, "Ghulam Muhammad, “Green cognitive body sensor network: architecture, energy harvesting and smart-clothing based applications”", IEEE Sens. J., vol. 19, no. 19, pp. 8371-8378, 2019. [http://dx.doi.org/10.1109/JSEN.2018.2870251]
[9]
N. Abbas, M. Asim, N. Tariq, T. Baker, and S. Abbas, "A Mechanism for Securing IoT-enabled Applications at the Fog Layer", Journal of Sensor and Actuator Networks, vol. 8, no. 1, p. 16, 2019. [http://dx.doi.org/10.3390/jsan8010016]
26 Computational Intelligence for Data Analysis, Vol. 2
Premnath and Nalini
[10]
H. Tao, M.Z.A. Bhuiyan, A.N. Abdalla, M.M. Hassan, J.M. Zain, and T. Hayajneh, "Secured Data Collection With Hardware-Based Ciphers for IoT-Based Healthcare", IEEE Internet Things J., vol. 6, no. 1, pp. 410-420, 2019. [http://dx.doi.org/10.1109/JIOT.2018.2854714]
[11]
S. Sharma, V. Chang, U.S. Tim, J. Wong, and S. Gadia, "Cloud and IoT-based emerging services systems", Cluster Comput., vol. 22, no. 1, pp. 71-91, 2019. [http://dx.doi.org/10.1007/s10586-018-2821-8]
[12]
B. Mallikarjuna, and D. A. K. Reddy, Healthcare Application Development in Mobile and Cloud Environments, in Internet of Things and Personalized Healthcare Systems”, Appl. Sci. Technol., pp. 93-103, 2019.
[13]
W Azeem, A.A. Malik, K. Amjad, M. Asad and H. U. Rashid, “Applications of Internet of Medical Things (IoMT) In Telemedicine for Medical and Health Care System”, Pakistan Journal of Science, Vol. 72, No. 4, 2020.
[14]
"M. Al khafajiy, Thar Baker, Carl Chalmers, Muhammad Asim, Hoshang Kolivand, Muhammad Fahim and Atif Waraich,“ Remote health monitoring of elderly through wearable sensors", Multimedia Tools Appl., vol. 78, pp. 24681-24706, 2019.
[15]
L.P. Malasinghe, N. Ramzan, and K. Dahal, "Remote patient monitoring: a comprehensive study", J. Ambient Intell. Humaniz. Comput., vol. 10, no. 1, pp. 57-76, 2019. [http://dx.doi.org/10.1007/s12652-017-0598-x]
[16]
N. Bashi, Mohanraj Karunanithi, Farhad Fatehi, Hang Ding and Darren Walters, “Remote monitoring of patients with heart failure: an overview of systematic reviews”, J. Med. Internet Res., vol. 19, no. 1, p. e18, 2017. [http://dx.doi.org/10.2196/jmir.6571] [PMID: 28108430]
[17]
I. Azimi, A.M. Rahmani, P. Liljeberg, and H. Tenhunen, "Internet of things for remote elderly monitoring: a study from user-centered perspective", J. Ambient Intell. Humaniz. Comput., vol. 8, no. 2, pp. 273-289, 2017. [http://dx.doi.org/10.1007/s12652-016-0387-y]
[18]
J.T. Giger, N.D. Pope, H.B. Vogt, C. Gutierrez, L.A. Newland, J. Lemke, and M.J. Lawler, "Remote patient monitoring acceptance trends among older adults residing in a frontier state", Comput. Human Behav., vol. 44, pp. 174-182, 2015. [http://dx.doi.org/10.1016/j.chb.2014.11.044]
[19]
B. Noah, M.S. Keller, S. Mosadeghi, L. Stein, S. Johl, S. Delshad, V.C. Tashjian, D. Lew, J.T. Kwan, A. Jusufagic, and B.M.R. Spiegel, "Impact of remote patient monitoring on clinical outcomes: an updated meta-analysis of randomized controlled trials", NPJ Digit. Med., vol. 1, no. 1, p. 20172, 2018. [http://dx.doi.org/10.1038/s41746-017-0002-4] [PMID: 31304346]
[20]
A. Benjemmaa, "Hela Ltifi, Mounir Ben Ayed, “Design of Remote Heart Monitoring System for Cardiac Patients", International Conference on Advanced Information Networking and Applications,. pp. 963-976, 2019.
[21]
J. Pagan, R. Fallahzadeh, M. Pedram, J.L. Risco-Martin, J.M. Moya, J.L. Ayala, and H. Ghasemzadeh, "Ramin Fallahzadeh, Mahdi Pedram, Jose L. Risco-Martin, Jose M. Moya, Jose L. Ayala, Hassan Ghasemzadeh,“Toward ultra-low-power remote health monitoring: An optimal and adaptive compressed sensing framework for activity recognition”", IEEE Trans. Mobile Comput., vol. 18, no. 3, pp. 658-673, 2019. [http://dx.doi.org/10.1109/TMC.2018.2843373]
[22]
A. Depari, D. Fernandes Carvalho, P. Bellagente, P. Ferrari, E. Sisinni, A. Flammini, and A. Padovani, "An IoT Based Architecture for Enhancing the Effectiveness of Prototype Medical Instruments Applied to Neurodegenerative Disease Diagnosis", Sensors (Basel), vol. 19, no. 7, pp. 1564-1582, 2019. [http://dx.doi.org/10.3390/s19071564] [PMID: 30935157]
Internet of Medical Things
Computational Intelligence for Data Analysis, Vol. 2 27
[23]
B. Farahani, and F. Firouzi, Krishnendu Chakrabarty, Sani Nassif, “Healthcare IoT”, Intelligent Internet of Things. Springer, 2020, pp. 515-545. [http://dx.doi.org/10.1007/978-3-030-30367-9_11]
[24]
M. Pham, "Yehenew Mengistu; Ha Manh Do; Weihua Sheng, “Cloud-based smart home environment (CoSHE) for home healthcare", IEEE International Conference on Automation Science and Engineering (CASE),. pp. 483-488, 2016. [http://dx.doi.org/10.1109/COASE.2016.7743444]
[25]
A.E. Hassanien, Nilanjan Dey, Surekha Borr, “Medical Big Data and Internet of Medical Things: Advances”, Challenges and Applications. CRC Press, 2018. [http://dx.doi.org/10.1201/9781351030380]
[26]
S. Cirani, G. Ferrari, M. Picone, L. Veltri, “Internet of things: architectures, protocols and standards”, Wiley 2019.
[27]
G.J. Joyia, R.M. Liaqat, A. Farooq, S. Rehman, “Internet of Medical Things (IOMT): Applications, Benefits and Future Challenges in Healthcare Domain”, J. Commun., vol. 12, no. 4, pp. 240-247, 2017. [http://dx.doi.org/10.12720/jcm.12.4.240-247]
[28]
S. Akhbarifar, H.H.S. Javadi , A.M. Rahmani, “A secure remote health monitoring model for early disease diagnosis in cloud-based IoMT environment”, Pers Ubiquit Comput, 2020. [http://dx.doi.org/10.1007/s00779-020-01475-3]
[29]
TW Lin, CL Hsu, TV Le, CF Lu, BY Huang, “A Smartcard-Based User-Controlled Single Sign-On for Privacy Preservation in 5G-IoMT Telemedicine Systems”, Sensors, Vol. 21, No. 8, pp. 2880, 2021. [http://dx.doi.org/10.3390/s21082880]
[30]
D.A. Chaudhari, E. Umamaheswari, “A new adaptive XOR, hashing and encryption-based authentication protocol for secure transmission of the medical data in Internet of Things (IoMT)”, Biomed Tech (Berl), vol. 66, no. 1, pp. 91-105, 2021. [http://dx.doi.org/10.1515/bmt-2019-0123]
[31]
S. Alsamhi, B. Lee, “Block-chain for multi-robot collaboration to combat COVID-19 and future pandemics”, IEEE Access, p. 1, 2020.
[32]
S. Kumar, P. Tiwari, and M. Zymbler, "Internet of Things is a revolutionary approach for future technology enhancement: a review", J. Big Data, vol. 6, no. 1, p. 111, 2019. [http://dx.doi.org/10.1186/s40537-019-0268-2]
[33]
B. Pradhan, S. Bhattacharyya, and K. Pal, "IoT-Based Applications in Healthcare Devices", J. Healthc. Eng., vol. 2021, pp. 1-18, 2021. [http://dx.doi.org/10.1155/2021/6632599]
[34]
L.A. Haidari, S.T. Brown, M. Ferguson, E. Bancroft, M. Spiker, A. Wilcox, R. Ambikapathi, V. Sampath, D.L. Connor, and B.Y. Lee, "The economic and operational value of using drones to transport vaccines", Vaccine, vol. 34, no. 34, pp. 4062-4067, 2016. [http://dx.doi.org/10.1016/j.vaccine.2016.06.022] [PMID: 27340098]
[35]
A. Pulver, R. Wei, and C. Mann, "Locating AED enabled medical drones to enhance cardiac arrest response times", Prehosp. Emerg. Care, vol. 20, no. 3, pp. 378-389, 2016. [http://dx.doi.org/10.3109/10903127.2015.1115932] [PMID: 26852822]
[36]
M. Li, R.H. Taylor, “Spatial motion constraints in medical robot using virtual fixtures generated by anatomy”, In IEEE International Conference on Robotics and Automation. Proceedings. ICRA'04, vol. 2, pp. 1270–1275, 2004.
[37]
F. Jiang, Y. Jiang, H. Zhi, Y. Dong, H. Li, S. Ma, Y. Wang, Q. Dong, H. Shen, Y. Wang, “Artificial intelligence in healthcare: past, present and future”, Stroke Vasc. Urol., Vol. 2 No. 4, pp. 230–243, 2017.
28 Computational Intelligence for Data Analysis, Vol. 2
Premnath and Nalini
[38]
J. Hunter, “Adopting AI is essential for a sustainable pharma industry”, Drug Discov. World. pp. 69–71, 2016.
[39]
P. Shah, F. Kendall, and S. Khozin, "Ryan Goosen, Jianying Hu, Jason Laramie, Michael Ringel & Nicholas Schork, “Artificial intelligence and machine learning in clinical development: a translational perspective”", NPJ Digit. Med., vol. 2, no. 69, pp. 1-5, 2019.
[40]
P. Agrawal, "Artificial intelligence in drug discovery and development", J. Pharmacovigil., vol. 6, no. 2, p. 6, 2018. [http://dx.doi.org/10.4172/2329-6887.1000e173]
[41]
A. Kiourti, K.A. Psathas, and K.S. Nikita, "Implantable and ingestible medical devices with wireless telemetry functionalities: A review of current status and challenges", Bioelectromagnetics, vol. 35, no. 1, pp. 1-15, 2014. [http://dx.doi.org/10.1002/bem.21813] [PMID: 24115132]
[42]
D.R. Marshall, “Swallow able data recorder capsule medical device”, US Patent 6, 632,175, 2003.
[43]
W.R. Bandy, B.G. Jamieson, K.J. Powell, K.E. Salsman, R.C. Schober, J. Weitzner, M.R. Arneson, “Ingestible endoscopic optical scanning device”, US Patent 8,529,441, 2013.
[44]
X. Liu, C. Steiger, S. Lin, G.A. Parada, J. Liu, H.F. Chan, H. Yuk, N.V. Phan, J. Collins, S. Tamang, G. Traverso, and X. Zhao, "Ingestible hydrogel device", Nat. Commun., vol. 10, no. 1, p. 493, 2019. [http://dx.doi.org/10.1038/s41467-019-08355-2] [PMID: 30700712]
[45]
S.L. Toohey, A. Wray, and W. Wiechmann, Michelle Lin, Megan Boysen-Osborn, “Ten tips for engaging them illennial learner and moving an emergency medicine residency curriculum into the 21st century”., West. J. Emerg. Med., vol. 17, no. 3, pp. 337-343, 2016. [http://dx.doi.org/10.5811/westjem.2016.3.29863] [PMID: 27330668]
[46]
F. Santoso, S.J. Redmond, “Indoor location-aware medical systems for smart homecare and telehealth monitoring: state-of-the-art”, Physiol. Measur., Vol. 36, No. 10, R53-87, 2015. [http://dx.doi.org/10.1088/0967-3334/36/10/R53]
[47]
S Mc Farland, A. Coufopolous, D. Lycett, “The effect of tele health versus usual care for home- care patients with long-term conditions: a systematic review, meta-analysis and qualitative synthesis”, J. Telemed. Telecare, 1–19, 2019. [PMID: 31394973]
[48]
P. Cullin, T. Bergdahl, “A telecare system”, US Patent App. 16/310,127, 2019.
[49]
N. Saeed, M. Manzoor, P. Khosravi,“An exploration of usability issues in telecare moni- toring systems and possible solutions: a systematic literature review”, Disabil. Rehabil. Assistive Technol., Vol. 15, No. 3, pp. 271-281, 2020. [http://dx.doi.org/10.1080/17483107.2019.1578998]
[50]
K. Anderson, O. Burford, L. Emmerton, “Mobile health apps to facilitate self-care: a qualitative study of user experiences”, PLoS ONE, Vol. 11, No. 5, pp. e0156164, 2016. [http://dx.doi.org/10.1371/journal.pone.0156164] [PMID: 27214203]
[51]
M. Kang, E. Park, B.H. Cho, and K.S. Lee, "Recent patient health monitoring platforms incorporating Internet of things-enabled smart devices", Int. Neurourol. J., vol. 22, no. 2, suppl. Suppl. 2, pp. S76S82, 2018. [http://dx.doi.org/10.5213/inj.1836144.072]
[52]
G.S. Chrysant, "Peripheral vascular disease is associated with increased pulse wave velocity and augmentation index: clinical implications", J. Clin. Hypertens. (Greenwich), vol. 16, no. 11, pp. 788789, 2014. [http://dx.doi.org/10.1111/jch.12407] [PMID: 25243631]
[53]
S. Kumar, A.K. Arora, P. Gupta, B.S. Saini, “A Review of Applications, Security and Challenges of Internet of Medical Things”, Springer Nature Switzerland AG, Cognitive Internet of Medical Things
Internet of Medical Things
Computational Intelligence for Data Analysis, Vol. 2 29
for Smart Healthcare, Studies in Systems, Decision and Control Vol. 311, 2021. [http://dx.doi.org/10.1007/978-030-55833-8_1] [54]
C. Emily, “Importance of IoMT in Healthcare 2020”, Article: Digital Healthcare, July 20, 2020.
[55]
M.A. Jabbar., “Internet of medical things (IoMT): a systematic review of applications, trends, challenges, and future directions”, The Internet of Medical Things Enabling technologies and emerging applications. In: S.K. Pani, P. Patra, G. Ferrari, R. Kraleva, X. Wang, Eds. The Institution of Engineering and Technology, London, United Kingdom, 2021. [http://dx.doi.org/10.1049/PBHE034E_ch1]
[56]
W. Ejaz and M. Ibnkahla, “Multiband spectrum sensing and resource allocation for IoT in cognitive 5G networks”, IEEE Internet Things J., Vol. 5, No. 1, pp. 150-163, 2018.
[57]
R. Pratap Singh, M. Javaid, A. Haleem, R. Vaishya, and S. Ali, "Internet of Medical Things (IoMT) for orthopaedic in COVID-19 pandemic: Roles, challenges, and applications", J. Clin. Orthop. Trauma, vol. 11, no. 4, pp. 713-717, 2020. [http://dx.doi.org/10.1016/j.jcot.2020.05.011] [PMID: 32425428]
[58]
D.S.W. Ting, L. Carin, V. Dzau, and T.Y. Wong, "Digital technology and COVID-19", Nat. Med., vol. 26, no. 4, pp. 459-461, 2020. [http://dx.doi.org/10.1038/s41591-020-0824-5] [PMID: 32284618]
[59]
Y. Song, J. Jiang, X. Wang, D. Yang and C. Bai,”Prospect and application of Internet of Things technology for prevention of SARIs”, Clin. eHealth, Vol. 3, pp. 1-4, 2020. [http://dx.doi.org/10.1016/j.ceh.2020.02.001]
[60]
D. Koh, “Temp Pal Smart Thermometer Helps Reduce COVID- 19 Spread in Hospitals”, 2020.
[61]
Microsoft, Trying to Shield Hospital Staff and Patients from COVID-19 with Help from AI, Cloud, and Intelligent Edge - Asia News Center, 2020.
[62]
A. Chirico, F. Lucidi, M. De Laurentiis, C. Milanese, A. Napoli, and A. Giordano, "Virtual reality in health system: beyond entertainment. A mini-review on the efficacy of VR during cancer treatment", J. Cell. Physiol., vol. 231, no. 2, pp. 275-287, 2016. [http://dx.doi.org/10.1002/jcp.25117] [PMID: 26238976]
[63]
M. Yahara, K. Niki, K. Ueno, M. Okamoto, T. Okuda, H. Tanaka, Y. Naito, R. Ishii, M. Ueda, and T. Ito, "Remote Reminiscence Using Immersive Virtual Reality May Be Efficacious for Reducing Anxiety in Patients with Mild Cognitive Impairment Even in COVID-19 Pandemic: A Case Report", Biol. Pharm. Bull., vol. 44, no. 7, pp. 1019-1023, 2021. [http://dx.doi.org/10.1248/bpb.b21-00052]
[64]
A. Li, Z. Montaño, V.J. Chen, and J.I. Gold, "Virtual reality and pain management: current trends and future directions", Pain Manag. (Lond.), vol. 1, no. 2, pp. 147-157, 2011. [http://dx.doi.org/10.2217/pmt.10.15] [PMID: 21779307]
[65]
B. Birckhead, C. Khalil, X. Liu, S. Conovitz, A. Rizzo, I. Danovitch, K. Bullock, and B. Spiegel, "Recommendations for methodology of virtual reality clinical trials in health care by an international working group: iterative study", JMIR Ment. Health,. Vol. 6, No. 1, p. e11973, 2019. [http://dx.doi.org/10.2196/11973] [PMID: 30702436]
[66]
S.S.Y. Wang, W.Z.W. Teo, W.Z.Y. Teo, and Y.W. Chai, "Virtual reality as a bridge in palliative care during COVID-19", J. Palliat. Med., vol. 23, no. 6, p. 756, 2020. [http://dx.doi.org/10.1089/jpm.2020.0212] [PMID: 32324080]
[67]
A. Asadzadeh, T. Samad-Soltani, P. Rezaei-Hachesu, “Applications of virtual and augmented reality in infectious disease epidemics with a focus on the COVID-19 outbreak”, Inform Med Unlocked, Vol. 24, p. 100579, 2021. [http://dx.doi.org/10.1016/j.imu.2021.100579]
[68]
S. Wang, M. Parsons, J. Stone-McLean, P. Rogers, S. Boyd, K. Hoover, O. Meruvia-Pastor, M. Gong, and A. Smith, "Augmented reality as a telemedicine platform for remote procedural training", Sensors
30 Computational Intelligence for Data Analysis, Vol. 2
Premnath and Nalini
(Basel), vol. 17, no. 10, p. 2294, 2017. [http://dx.doi.org/10.3390/s17102294] [PMID: 28994720] [69]
J. Gerup, C.B. Soerensen, and P. Dieckmann, "Augmented reality and mixed reality for healthcare education beyond surgery: an integrative review", Int. J. Med. Educ., vol. 11, pp. 1-18, 2020. [http://dx.doi.org/10.5116/ijme.5e01.eb1a] [PMID: 31955150]
Computational Intelligence for Data Analysis, 2023, Vol. 2, 31-46
31
CHAPTER 3
An Approach to the Internet of Medical Things (IoMT): IoMT-Enabled Devices, Issues, and Challenges in Cybersecurity Usha Nandhini Rajendran1 and P. Senthamizh Pavai1,* 1
Faculty of Education, Dr. M.G.R. Educational and Research Institute, Chennai, India Abstract: As the number of devices connected to the Internet (Internet of Things: IoT) grows, ensuring reliable security and privacy becomes more difficult. With the widespread usage of online medical facilities, security and privacy in the medical arena have become a severe problem that is only becoming worse. The criticality and sensitivity of data in the healthcare industry make guaranteeing the security and privacy of the Internet of medical things (IoMT) even more difficult. The privacy of the patients will be threatened, and their lives may be threatened if effective measures are not implemented in IoMT. Also, it provides novel services, such as remote sensing, elder care assistance, and e-visit, improving people’s health and convenience while lowering medical institution costs per-patient. However, with the rise of mobile, wearable, and telemedicine options, security can no longer be assessed just inside the confines of clean physical walls. Nonetheless, by implementing recognized and applicable safeguards, the risk of exploiting vulnerabilities can be greatly decreased. This article provides an outline of the key security and privacy measures that must be implemented in current IoMT environments to protect the users and stakeholders involved. The overall approach can be seen as a best-practice guide for safely implementing IoMT systems.
Keywords: Cybersecurity, Internet of Medical Things (IoMT), Online Medical Facilities, Patient Monitoring, Privacy, Protection. INTRODUCTION The Internet of Medical Effects (IoMT) is a network of Internet-connected medical outfits, tackle structure and software operations that connect healthcare IT. IoMT, as mentioned in Fig. (1), shows the Internet of Effects in healthcare, allows wireless and remote bias to securely connect via the Internet to allow for rapid-fire and flexible medical data processing. The impact of the Internet of Effects on the healthcare business is apparent and unrecoverable [1]. Likewise, Corresponding author P. Senthamizh Pavai: Faculty of Education, Dr. M.G.R. Educational and Research Institute, Chennai, India; E-mail: [email protected]
*
Mariya Ouaissa, Mariyam Ouaissa, Zakaria Boulouad, Inam Ullah Khan, Sailesh Iyer (Eds.) All rights reserved-© 2023 Bentham Science Publishers
32 Computational Intelligence for Data Analysis, Vol. 2
Rajendran and Pavai
IoMT operations are explosively associated with sensitive healthcare services, particularly because they manage sensitive patient information similar as names, addresses, and health conditions. The crucial problem in the IoMT sector is maintaining patient sequestration while maintaining a high position of security. Likewise, applicable security and sequestration results should only take minimum calculations and coffers [2].
Fig. (1). Internet of medical things.
IoMT refers to internet-connected bias and operations used by healthcare providers to cut costs and ameliorate patient care. The bias can be employed in hospitals or at the homes of cases. These bias prisoners, store, transmit, and process sensitive data similar to names, addresses, and the state of your health. The following are some instances of it: ●
●
●
●
Smart fitness devices and smart blood pressure devices are examples of personal wearable devices. Infusion pumps and dialysis machines are examples of medical devices that can be used at home. Defibrillators, anaesthetic machines, patient monitoring, and Personal Emergency Response Systems (PERS) are examples of in-hospital and clinical gadgets. Cameras that can be consumed.
This application allows patients to provide medical data in real-time via their mobile phone and internet connection. Patients with diabetes, heart disease, and high blood pressure, in particular, have found it handy to communicate medical records to doctors using remote health-monitoring apps on their smartphones.
Challenges in Cybersecurity
Computational Intelligence for Data Analysis, Vol. 2 33
These gadgets are commonly used in hospitals and nursing homes to comfort patients and assist doctors in reducing the number of journeys they make to see patients. The IoMT market is massive, and it’s only going to get bigger shortly. Doctors were unable to continuously monitor a patient’s health before the Internet of Medical Things, making even disease identification difficult. The influence of linked gadgets on patients’ life is becoming more apparent as their use grows. IoMT lowers healthcare expenditures and delivers more precise data, as well as reduces frequent hospital visits, while enhancing treatment quality. PATIENT-MONITORING SYSTEM IN IOMT Patient monitoring in real-time (RTPM), such as glucose and heart rate monitoring, IoMT is primarily concerned with healthcare and medical applications. IoMT requires a more robust security structure than other IoT systems due to the perceptivity and rigorous laws around healthcare data. Fig. (2) represents the Case-Monitoring functions of the system, which explains that patient health issues are observed through a medical device and, in turn, shared with the physicians through an assistant using the internet of effects. The Internet of Effects has a variety of goods on healthcare assiduity. These changes are particularly conspicuous when it's used in the home, on the body, in the community, and in the hospital.
Fig. (2). Patient-monitoring system in IoMT.
At Home People can transfer medical data from their homes to other locales, similar to their primary care provider or a hospital, using in-home IoMT. Remote case monitoring (RPM), for illustration, is the use of a medical outfit to shoot criteria like blood pressure and oxygen saturation from lately discharged cases to their hospital for
34 Computational Intelligence for Data Analysis, Vol. 2
Rajendran and Pavai
assessment by their doctors. This can help to help readmissions to the medical centre by feting problems before they become serious. Telehealth, or the use of communication technologies for remote healthcare, adds to the inflexibility by allowing lately discharged cases to communicate with their doctors ever to handle small difficulties. Outside of the case environment, the use of IoMT bias in confluence with telemedicine can be salutary for ongoing treatment. Particular exigency response systems (PERS), for illustration, can determine circumstances like a fall or a heart attack and automatically hail deliverance. PERS can give security to persons who are in trouble, similar to seniors who wish to live at home but are concerned about their safety. In-Person Wearable medical bias connected to remote shadowing or monitoring systems is appertained to as on-body IoMT. On-body IoMT, unlike in-home IoMT, can be used outside of the home while people go about their diurnal conditioning. Onbody consumer IoMT bias is wearable health trackers that anybody can buy for a particular use or share with healthcare providers. These bias can give early warning pointers for more serious health enterprises, in addition to recording a typical metric like heart rate. The Apple Watch, for illustration, can warn druggies of abnormal cardiac measures. Clinical on-body IoMT biases are analogous to consumer IoMT bias, except they have a larger detector selection. Diabetic individuals, for illustration, can wear glucose observers to descry changes in their blood glucose situations. Numerous of these widgets can directly communicate data with a case’s physicians, icing that they admit prompt and accurate treatment. At Community The use of IoMT bias throughout a larger city or geographic area is appertained to as IoMT. Mobility services, for illustration, are widgets that track cases while they are in an auto. Paramedics and first askers employ exigency response intelligence systems to track patient criteria outside of the medical centre setting. Community IoMT includes technologies that offer remote services in addition to mobile and exigency care. Point-of-care bias, for illustration, can be used by healthcare providers in non-conventional medical settings like a field medical centre, and alcoves can be used to apportion drugs to cases in locales where a traditional structure is limited or missing. IoMT bias may also be used in logistics by suppliers to help in the transfer of healthcare goods or medical outfits. Detectors, for illustration, can keep an eye on the temperature-or pressure-sensitive shipping holders to ensure that quality is maintained throughout the transport process.
Challenges in Cybersecurity
Computational Intelligence for Data Analysis, Vol. 2 35
In-Hospital Hospitals must keep track of the quality and force of their medical means throughout time, as well as how workers and cases use the installation. IoMT detectors and other shadowing bias are used by healthcare interpreters to track all of these relations so that directors can gain a complete picture of what's going on. IoMT bias may also be used in logistics by suppliers to help in the payload of healthcare inventories. POTENTIAL OF THE INTERNET OF MEDICAL THINGS The Internet of Effects/ Internet of Medical Effects has the implicit to deliver further cost-effective, advanced-quality, more individualized care while also perfecting patient commission in the following ways: Cost-Cutting Solutions like the IoMT are becoming even more relevant as the United States spends 18 percent of its annual GDP on healthcare (a figure that is anticipated to climb). The “digital revolution” is expected to save more than $300 billion in healthcare spending, according to estimates. According to the Dell blog (through Goldman Sachs), the largest actual savings opportunity is in chronic disease management, which is worth more than $200 billion. It is also worth highlighting the “infinitely vast” savings prospects in the domain of behaviour modification in the treatment of diseases like obesity and smoking cessation. Better Care Devices that are part of the IoT/IoMT can provide remote data to caregivers and clinicians, such as temperature, heart rates, and glucose levels, that was previously unavailable. The capacity to communicate health data automatically can aid in determining risk for a variety of acute and chronic conditions. Patients with a Sense of Empowerment Patient empowerment is critical not only in an era of patient-centred care but also in a time when we are learning just how much of a difference a patient’s attitude toward treatment can make in clinical outcomes. Telehealth will play a key role in this field, particularly in terms of using remote doctor access to modify behaviour and promote healthier living [1].
36 Computational Intelligence for Data Analysis, Vol. 2
Rajendran and Pavai
TAXONOMY OF IOMT SECURITY & PRIVACY (S & P) LAYERS As medical systems are so sensitive, managing the hundreds of medical effects dissimilarly connected across the Internet is critical to ensure the loftiest position of trustworthiness. As a result, a flexible layered armature is needed. The five levels of IoMT model described depicts a layered armature in which each level performs a distinct function. Each capability has its own set of S & P problems [3]. As a result, we classify S & P enterprises into the following orders grounded on their frequency in each level, as depicted in Fig. (3).
Thought
IoMT
Internet
Marketing
Android
Intermediate Fig. (3). Taxonomy of IoMT security & privacy layers.
Thought The thought layer is in charge of carrying and collecting data (e.g., body temperature, heart rate, etc.) through the use of physical outfits (e.g., detectors), and also transmitting that data to the network (i.e., Internet). In health monitoring systems, numerous detectors communicate to guarantee that the case is always covered and receives help as instantly as possible. The four types of medical effects engage with stakeholders (e.g., physicians and cases) in the following ways: Wearable Gadgets: Wearable widgets, like smartwatches, give for accurate, endless, and real-time case monitoring. They are: ● ● ● ●
Position detectors: Cases’ localities are tracked using this. Temperature detectors: It's used to measure the body's temperature. Pressure detectors: It's used to keep track of a case’s blood pressure. Respiratory detectors: To keep track of how well cases are breathing.
Challenges in Cybersecurity ● ●
Computational Intelligence for Data Analysis, Vol. 2 37
Palpitation oximeters: It measure oxygen saturation and palpitation rate. Biometric detectors: To keep the identity protected (e.g., fingerprints, face detection).
Implantable Gadgets: This biasfits inside the case’s body. Current gadgets include: ●
●
Swallowable camera capsule visualises the case’s gastrointestinal tract from the inside. Embedded cardiac gathers data and sends it over a radio link to a near-universal network [3].
Ambient Devices: These gadgets ascertain the case’s surroundings and examiner exertion patterns and transmit caregiver’s announcements if aberrant patterns are detected. Detectors in the terrain include: ●
● ● ●
● ●
Detectors that discover the movement of individualities in a room are known as stir detectors. Room temperature detectors are used to record the temperature of a room. Pressure detectors measure the volume of gas, liquid, and air. Door detectors determine the status of the door (open or unrestricted) for Alzheimer’s cases or infection control. Cases’ body movements on beds are detected using vibration detectors. Daylight detectors determine natural light and modify the lighting zones in the space automatically [4].
Stationary Devices: These widgets determine the case’s surroundings and examiner exertion patterns and shoot caregiver’s announcements if aberrant patterns are detected. Detectors in the terrain include: ● ● ●
●
Glamorous resonance imaging (MRI), reckoned tomography (CT). Scanners and X-rays are exemplifications of imaging systems that give visual. Representations of the innards of a body for clinical study and medical intervention. Instruments used in surgeries or operations are appertained to as surgical bias. Attacks on these bias compromise the integrity and sequestration of data, maybe performing in grave consequences.
38 Computational Intelligence for Data Analysis, Vol. 2
Rajendran and Pavai
Internet Content delivery, content discovery, content routing to the destination, and network addressing are all handled by the internet layer. The following are the media that medical effects use: Wired/Wi-Fi: IoT devices link to the gateway and end- user via wired or wireless networks. Due to the need for a dependable power source, a bias that employs this mode of communication is constantly immobile. This is applicable for IoMT systems that demand a high position of speed. Radio Communication: To connect with user end devices, certain low-power IoT mobile devices use radio communication similar to 3G, 4G, LTE, Bluetooth, and RFID. Bluetooth low energy (BLE) is a vital short-range communication technology used in various wearable medical devices. In a hospital room, a stationary Bluetooth core connects numerous devices across the room, allowing data transmission to and down. Long-distance communication is possible with cellular-grounded IoT devices (3G/ 4G/ LTE). Wireless Networks: Medical Things can also connect using Wireless Networks (WSNs), which use regular Wi-Fi or a low-power wireless personal area network (6LoWPAN) to communicate. Middleware Data delivery, discovery and routing to the destination, and network addressing are all handled by the internet layer. Android This layer connects the user interface with people through IoMT devices in an intermediate layer. Due to the superior scalability and flexibility, operation manufacturers have been learning further about hosting operations in cloud computing. Marketing This layer is in charge of handling the business intellection of the healthcare provider as well as supporting the business process lifecycle, which includes monitoring, managing, and optimizing business processes. It is also in charge of gleaning perceptivity from IoMT data. Attacks on this level have been mooted preliminarily, but they have an advanced impact then because the data contains sensitive medical information.
Challenges in Cybersecurity
Computational Intelligence for Data Analysis, Vol. 2 39
CYBERSECURITY ATTACKS IN EACH LAYER Understanding the challenges enclosing diverse cyber-attacks and designing defence approaches (i.e., countermeasures) that save the confidentiality, integrity, and accessibility of any digital and information technologies are all part of cybersecurity. The following illustration (Fig. 4), listed out the cyber-attacks in each layer (Table 1). Cybersecurity Attacks in Each Layers
Perception Layer • Extracting restricted data • Data duplication • Data hacking and Monitoring
Network Layer • Data interference • Data duplication • Fraudulent data • Declination
Application Layer • Data driven attack • Account hacking • Cryptoviral extortion • Unbreakable codes
Middleware • Session riding • Spoofing • Software attack
Business Layer • Information leakage • Data misleading • Interruption and data possessing without right
Fig. (4). Cybersecurity attacks in each layers of IoMT. Table 1. Aspects of cyber security and protection mechanisms. Aspect
Privacy
Reliability
Protection Mechanism
Description
Privacy
Ensures that a reused asset remains unknown to anyone outside of the interacting realities.
Verification
Authorizations are challenged predicated on their identification and authorization.
Flexibility
In the event of a failure, the protection is preserved.
Reliability
Ensures that all interacting realities are apprehensive when an asset has been modified.
Suppression
Guarantees that transactions take place according to a set of rules, eliminating the freedom of choice and liability in the event of a disclosure.
Validation
Guarantees that deals take place according to a set of rules, barring the freedom of choice and liability in the event of exposure.
40 Computational Intelligence for Data Analysis, Vol. 2
Rajendran and Pavai
(Table 1) cont.....
Aspect
Protection Mechanism
Description
Connection
In the event of a failure, commerce is saved.
Alarm
Informs the stoner that commerce is being or has passed.
Secure
A contract between the asset holder and the interacting reality is also included. It may also include public legal protection and warnings as a primary to legal action.
Accessibility
PII: Personal Identifiable Information
Data Collection
Data Access
Data Usage
Note: Source [5].
Consent
Requires the PII holder’s voluntary, specific, and informed concurrence to the PII’s processing. Without the holder’s authorization, PII cannot be participated in or given to a third party.
Opt-in
Before applicable concurrence, includes a policy or process in which the PII proprietor explicitly agrees to the PII’s processing.
Objectivity
Ensures that PII is only collected, used, or discovered for the purposes for which it was gathered, using fundamentals similar to data minimization and delicacy.
Accountability
Based on a set of PII, the results in identifying the PII owner, either directly or indirectly. Accountability, de-identification, or anonymity should all be included.
Notification
Notifies the PII owner that his or her information is being collected.
Auditability
Provides adequate mechanisms for identifying and controlling PII data access.
Liability
Supports the GDPR principles of legality, fairness, and openness by assuring that the PII holder can hold the PII processors responsible for clinging to all isolation measures.
Retention
As a precautionary approach to reduce the risk of illegal collection, disclosure, or usage, it ensures that PII that is no longer needed is not kept.
Disposal
Included are systems for erasing or disposing of PII on-demand, as well as the GDPR’s right to be forgotten.
Report
Notifies you that you are having or have had a PII encounter.
Break or Incident Response
Manages a data breach involving personally identifiable information (PII).
Challenges in Cybersecurity
Computational Intelligence for Data Analysis, Vol. 2 41
WAYS TO SOLVE IOMT ISSUES Inventory of Assets The first step toward improved security is to create and maintain an inventory of all IoMT. It is frequently unmanaged, as it is not tied to a responsible person who functions as the “owner,” many healthcare organizations fail to identify the devices. As no one is accountable for handling security precautions such as passwords, the inability to link a device and a user creates a security blind hole. Policy for Strong Passwords Default settings and passwords are included with every IoMT device. Threat actors, on the other hand, frequently locate default passwords online. The first step in connecting a new IoMT device to the network should be to set a new, strong password that is specific to the device. Some examples of best practices are as follows: ● ● ● ●
Using upper- and lower-case letters in a creative way. Including at least one number in the equation. Including at least one unique character. We should also ensure that the password is not stored in any online password databases. This reduces the chances of malevolent actors being able to “guess” the password.
Multi-Factor Authentication (MFA) It is a second step in preventing credential theft. MFA requires threat actors to supply additional information verifying that they are who they claim they are, even if they successfully log into the device. To authenticate to a device, network, or application, MFA involves using two or more of the following: ● ● ●
Something you are aware of a password. You have something like a smartphone. Something you are doing like a biometric like fingerprint or face recognition.
Segmentation of the Network The technique of physically or intellectually isolating networks containing sensitive information from those not known as network segmentation. This can be accomplished by storing sensitive data in a separate data centre from public internet-facing apps or limiting access to the network containing sensitive data
42 Computational Intelligence for Data Analysis, Vol. 2
Rajendran and Pavai
with firewalls. Because hostile actors are unable to travel from one network to another, this method reduces the potential consequences of a data leak. Updates to Security Patches Updates to software, operating systems and hardware addresses are identified as security flaws. Threat actors frequently use these flaws to obtain access to devices, networks, and apps. Risk can be mitigated by establishing a regular schedule that prioritizes updating essential IoMT devices and apps. Any network devices or components associated with any IoMT-connected network should also be prioritized in this schedule. Monitoring of Network Traffic Network traffic monitoring allows you to see if your devices are sending or receiving more data than they should. An IoMT device, for example, can be weaponized and utilized as a part of a botnet. In a botnet attack, the BotMaster is in charge of the compromised devices (known as “bots”), and issues command to them. The volume of requests and responses overwhelms the servers, resulting in a Denial of Service (DoS) attack. The healthcare company can discover possibly compromised IoMT devices and mitigate the attack's impact by monitoring for aberrant traffic. Encryption The Internet of Effects (IoMT) sends Electronic Protected Health Information (ePHI) to a linked operation. A connected insulin pump, for illustration, sends data to the app, allowing the case and clinician to keep track of glucose situations. The operation, still, is connected to the public internet. At the network position, data-in-conveyance encryption decreases the impact of wiretapping and man-i-the-middle attacks. Without the necessary decryption technology, encryption scrambles data, rendering it ungraspable. Indeed if bad actors gained access to the network, they would not be suitable to pierce the data [3]. System for Detecting Intrusions Signature-based, specification-based, or anomaly-based IDS are all options. The greatest defence for IoMT is anomaly-based. An anomaly-based IDS keeps an eye on the network for unusual activity. It frequently integrates machine learning to alert you to new threats. The capacity to identify zero-day attacks, which arise from previously undiscovered vulnerabilities, is the fundamental value of anomaly-based IDS. Many devices are not tied to known vulnerabilities because it is a newer technology. Healthcare businesses that wish to improve the security of
Challenges in Cybersecurity
Computational Intelligence for Data Analysis, Vol. 2 43
their IoMT devices should look into a solution that allows them to see new and emerging hazards [6]. SECURITY AND PRIVACY IN IOMT Traditional and zero-day attacks can lose IoMT bias. This is substantially due to the lack of established security norms and measures in device fabrication, as well as the devices and IoT network’s nature. Since the devices are so small, their computational resources and batteries cannot handle cryptography and strong security mechanisms [7]. The security and sequestration norms for IoMT are distinct from those for traditional networks, which are typically pertained to as the CIA- trio (confidentiality, integrity, and vacuity). Other measures for the IoMT system include legal possession and assurance. The descriptions of varied criteria utilised in the IoMT system are listed below [8]. 1. Confidentiality: It prevents illegal access to particular information and ensures that confidential information is protected. Data leakage and, in extreme cases, life-threatening scenarios might result from unauthorised access [9]. 2. Integrity: It ensures that unauthorised parties do not modify, cancel, or fit data or readings from bias. False data injection on an implantable trendsetter, for illustration, can affect death [10]. 3. Availability: It assures that data, computational rudiments, and connections are always available and running when a service needs them. When considering a surgical room equipped with wireless medical bias, system service dislocation poses a threat to cases’ health [11]. 4. Privacy: It ensures that the IoMT system follows privacy standards and allows users to view their personal information [12]. 5. Authentication: This statistic refers to the system’s capability to authenticate the actuality or absence of exertion. It ensures that the transferring knot receives delivery damage and that the entering knot receives verification of the sender’s identity, icing that none of them is refused during the process [13]. CHALLENGES OF IOMT’S Implanted systems are employed in a variety of settings to realize a wide range of operations, including telemedicine, business control, supported living, and smart metropolises. Digital systems (Cyber-systems) control physical effects in these operations, performing in a constant interplay between the digital and physical worlds. Still, the pivotal factors of these operations, particularly when implanted systems are taken into account, give three types of challenges:
44 Computational Intelligence for Data Analysis, Vol. 2
Rajendran and Pavai
1. These systems must meet conditions for responsibility, robustness, and security. Because the physical and physiological world is innately unstable, IoMT- grounded systems must not only be suitable to maintain respectable performance in the face of similar changes but also be suitable to reply properly if necessary. These IoMT systems also produce security enterprises since they constantly regulate conditions in which a system failure might be fatal. As a result, these systems must be suitable to repel a variety of offending attacks. 2. The IoMT must calculate on precise mongrel system models. They live at the crossroads of the digital and physical worlds, challenging accurate physical models as well as exact computational abstractions. Likewise, using a modelgrounded design allows for simulation-grounded optimization of testing procedures. 3. Specific verification and confirmation procedures are needed; the bulk of it's likely to be encyclopaedically distributed and to pass instruments, verification and confirmation protocols at numerous degrees of granularity are needed. 4. These difficulties are multidisciplinary, bringing together experimenters from several disciplines, similar as advanced control theory, computer knowledge, electronics engineering, power electronics, and signal processing. This necessitates the creation of inter-cross system models that include both digital and physical factors, as well as their relations. Eventually, to make the instrument process easier and increase system dependability, verification and confirmation procedures must be described at several scales, ranging from the lowest IoMT device scale to the largest IoMT scale [4, 14]. STEPS TO IMPROVE DEVICE SECURITY ● ● ● ●
● ● ● ●
● ● ●
Make a list of all the devices connected to the network. Passwords on devices should be strengthened. Increase network cleanliness and enforce segmentation regulations. Keep up with known and available patches, especially for devices that are very vulnerable. Monitor network traffic for malicious packets on a regular basis. Sharing of the same mobile number for ‘n’ number of systems. Sharing of the same E-mail ID for ‘n’ number of systems. Allowing ‘n’ no. of people to open and access the system with the same common shared password. Writing the password in a book or table, on walls, etc. Computer illiterates share the password/ID. Same staff used to sell information like mobile no., disease of the patient, etc. for commercial people, banks, etc. this should be identified and stopped.
Challenges in Cybersecurity
Computational Intelligence for Data Analysis, Vol. 2 45
CONCLUSION Despite its benefits, it is susceptible to a good range of pitfalls, issues, and challenges, the utmost of which aim to compromise patient sequestration and thus, the confidentiality, integrity, and vacuity of medical services. We outlined and anatomized the main issues, challenges, and downsides that IoMT faces, also because of the numerous security results which will be espoused to guard and secure IoMT disciplines and their affiliated means, like medical bias, systems, and medical (Cyber Physical System) CPSs, during this study. Likewise, several fabrics, taxonomies, and ways were handed to ensure a more fortified and robust IoMT and ameliorate the health and knowledge of cases. It is also pivotal to guard the numerous wireless communication protocols that the IoMT relies on. Eventually, maintaining a high position of security, sequestration, trust, and delicacy is critical. As a result, it's critical and advised that medical and IT help be trained so that they do not become victims of physical or cyber-attacks. In summary, this work aims to strengthen the connections between numerous technological and non-technical results to make a more comprehensive, safe, and effective system across all IoMT disciplines. CONSENT FOR PUBLICATON Declared none. CONFLICT OF INTEREST The author declares no conflict of interest, financial or otherwise. ACKNOWLEDGEMENT Declared none. REFERENCES [1]
https://www.mspinsights.com/doc/what-is-the-potential-of-the-internet-of-medical-things-0001 [Accessed: 29-Mar-2022]
[2]
J.P.A. Yaacoub, O. Salman, H.N. Noura, N. Kaaniche, A. Chehab, and M. Malli, "Cyber-physical systems security: Limitations, issues and future trends", Microprocess. Microsyst.. vol. 77, p. 103201, 2020. [http://dx.doi.org/10.1016/j.micpro.2020.103201] [PMID: 32834204]
[3]
F. Alsubaei, A. Abuhussein, and S. Shiva, "Security and Privacy in the Internet of Medical Things: Taxonomy and Risk Assessment", Proc. - 2017 IEEE 42nd Conf. Local Comput. Networks Work. LCN Work. 2017, no. September 2018, pp. 112–120, 2017 [http://dx.doi.org/10.1109/LCN.Workshops.2017.72]
[4]
C. Chakraborty, A. Banerjee, L. Garg, and J. J. P. C. Rodrigues, Internet of Medical Things for Smart Healthcare, 2020. [http://dx.doi.org/10.1007/978-981-15-8097-0]
46 Computational Intelligence for Data Analysis, Vol. 2
Rajendran and Pavai
[5]
G. Hatzivasilis, O. Soultatos, S. Ioannidis, C. Verikoukis, G. Demetriou, and C. Tsatsoulis, “Review of security and privacy for the internet of medical things (IoMT): Resolving the protection concerns for the novel circular economy bioinformatics,” Proceedings - 15th Annual International Conference on Distributed Computing in Sensor Systems, DCOSS 2019, 2019.
[6]
https://securityscorecard.com/blog/ways-to-secure-the-internet-of-medical-things [Accessed: 29-Ma-2022]
[7]
S.S. Hameed, W.H. Hassan, L. Abdul Latiff, and F. Ghabban, "A systematic review of security and privacy issues in the internet of medical things; the role of machine learning approaches", PeerJ Comput. Sci., vol. 7, pp. 1-44, 2021. [http://dx.doi.org/10.7717/peerj-cs.414] [PMID: 33834100]
[8]
P.I. Radoglou Grammatikis, P.G. Sarigiannidis, and I.D. Moscholios, "Securing the Internet of Things: Challenges, threats and solutions", Internet of Things, vol. 5, pp. 41-70, 2019. [http://dx.doi.org/10.1016/j.iot.2018.11.003]
[9]
N. Alassaf, and A. Gutub, "Simulating light-weight-cryptography implementation for IoT healthcare data security applications", Int. J. E-Health Med. Commun., vol. 10, no. 4, pp. 1-15, 2019. [http://dx.doi.org/10.4018/IJEHMC.2019100101]
[10]
J.P.A. Yaacoub, M. Noura, H.N. Noura, O. Salman, E. Yaacoub, R. Couturier, and A. Chehab, "Securing internet of medical things systems: Limitations, issues and recommendations", Future Gener. Comput. Syst., vol. 105, pp. 581-606, 2020. [http://dx.doi.org/10.1016/j.future.2019.12.028]
[11]
Y. Sun, F.P.W. Lo, and B. Lo, "Security and Privacy for the Internet of Medical Things Enabled Healthcare Systems: A Survey", IEEE Access, vol. 7, pp. 183339-183355, 2019. [http://dx.doi.org/10.1109/ACCESS.2019.2960617]
[12]
A. Mosenia, and N.K. Jha, "A comprehensive study of security of internet-of-things", IEEE Trans. Emerg. Top. Comput., vol. 5, no. 4, pp. 586-602, 2017. [http://dx.doi.org/10.1109/TETC.2016.2606384]
[13]
K.H. Yeh, "BSNCare+: A robust IoT-oriented healthcare system with non-repudiation transactions", Appl. Sci. (Basel), vol. 6, no. 12, p. 418, 2016. [http://dx.doi.org/10.3390/app6120418]
[14]
A. Gatouillat, Y. Badr, B. Massot, and E. Sejdic, "Internet of Medical Things: A Review of Recent Contributions Dealing With Cyber-Physical Systems in Medicine", IEEE Internet Things J., vol. 5, no. 5, pp. 3810-3822, 2018. [http://dx.doi.org/10.1109/JIOT.2018.2849014]
Computational Intelligence for Data Analysis, 2023, Vol. 2, 47-63
47
CHAPTER 4
Internet of Medical Things in Cloud Edge Computing G. Sumathi1,*, S. Rajesh2, R. Ananthakumar2 and K. Kartheeban2 1 2
Department of IT, Kalasalingam Institute of Technology Krishnankoil, Tamil Nadu, India Department of CSE, Mepco Schlenk Engineering College, Sivakasi, Tamil Nadu, India Abstract: Booming growth of ubiquitous connections and clinical computerization in the 5th generation mobile communication era, the explosive increase and heterogeneity of clinical information have delivered enormous demanding situations to information processing, privacy and security, as well as data access in (IoMT) Internet of Medical Things. Our paper gives a complete evaluation of the way to understand the analysis and timely processing of big data in medical applications and the dropping of healthcare resources in high quality under the limitations of previous medical equipment and the medical environment. We mostly concentrate on the benefits carried via the artificial intelligence, edge computing and cloud computing concepts to IoMT. We also explain how to use clinical resources while keeping the privacy and security of clinical information, so that extremely good clinical services can be given to patients.
Keywords: Cloud Computing, Edge Computing, Fog Computing, Healthcare, Internet of Medical Things. INTRODUCTION With the speedy growth of technology, economy and science, clinical treatment has emerged as one of the motivations of social, personal or even national interest, as described by Sun et al. [1]. The conventional medical design has troubles which include trouble with expensive treatment, occlusion of clinical facts and seeing a physician. But, with the proper establishment of knowledge of the Internet of Things (IoT), as suggested by Al-Fuqaha et al. [2], the application subject of IoT has been concerned in every aspect during today’s generation of Internet-of-Everything (IoE), as mentioned by W. Wang et al., X. Wang et al., as well as Ning et al. [3 - 5]. Regarding the clinical area involved, Internet-ofMedical-Things (IoMT) can be a focused incarnation of the IoT discipline, as well Corresponding author G. Sumathi: Department of IT, Kalasalingam Institute of Technology Krishnankoil, Tamil Nadu, India; E-mail:[email protected] *
Mariya Ouaissa, Mariyam Ouaissa, Zakaria Boulouad, Inam Ullah Khan, Sailesh Iyer (Eds.) All rights reserved-© 2023 Bentham Science Publishers
48 Computational Intelligence for Data Analysis, Vol. 2
Sumathi et al.
as a medium of clinical electronic renovation. IoMT, as described by Shit et al. [6], can be difficult IoT skills inclusive of positioning, sensor and radio frequency identity technology, in addition, carried out to clinical subjects in collaboration with network communication, cell terminals and extra gadgets to understand communication medical gadgets and medical institutions, patients, medical staff among patients, to attain the automation, digitalization as well as the intelligence of clinic. The IoMT helps every phase of clinical areas, such as medical drugs, waste and equipment management, remote monitoring, identification recognition and essential signs monitoring. The medical wireless sensor, as described by Liu et al. [7], can be the base as well as a core of every connection; also, wi-fi sensor communication integration of most forms of wi-fi clinical sensors, inclusive of implantable sensor, biosensor, pressure sensor, performs a vital function in gaining essential symptoms information of patients. Now, sensors have been broadly applied inside the (ICU) Intensive Care Unit, emergency room and operation room to screen and show the important disorders of patients. Moreover, Yang et al. [8] suggest that wearable medical device, including sensor as core, allows patients to acquire modified clinical services everywhere and anytime, importantly decreasing treatment price for both patients and doctors, giving realtime fitness management for patients and interruptions over restriction of space and time. How to rapidly analyze, process, and choose gathered clinical facts in real-time may be essential to immediately connect patients' lifestyles and fitness. Furthermore, as clinical information includes patient’s personal privacy, this can be essential to support the safety of the privacy data of patients and sensitive facts. However, the utility of IoT generation within the healthcare field may assist clinics to understand the intelligence clinical treatment of human beings as well as smart control of things; various healthcare institutions are exceptionally autonomous, therefore it's problematic to obtain useful resource sharing. The IoMT-related cloud computing (CC) presents influential basic resources of IT as well as substantially decreases medicinal costs. It can’t fulfill mass storing of clinical information, however additionally recognize the distribution of medical data via cloud environment, to enhance the effectiveness and excellence of clinical services. But, totally trusting CC will devour enormous network broadcast resources with a big delay, which will probably create a danger to a patient’s life. The capacity of CC to practice information sinks, creating information processing towards the source, as opposed to cloud or external fact that may reduce time delay, as well as accomplish real-time with quicker processing than evaluation of medical statistics. Edge Computing (EC) decreases dependent on centralized remote/local servers, then answers issues present in CC via sensible software of
Cloud Edge Computing
Computational Intelligence for Data Analysis, Vol. 2 49
sources over edge gadgets; this means clinics and hospitals acquire an extra responsive and agile IT community, therefore patients may experience higher medical services. Then EC doesn’t occur in separation. Cloud and Edge Computing accompany each other and perform exceptionally vital roles.CC focuses on the general understanding, while EC concentrates on the local. The normal usage of cloud edge integration will higher support the improvement of medical application cases. Based on the present situation in the healthcare field, our manuscript integrates 3 favorable skills for 1st time to systematically overview the existing work, as well as examine the experiments that IoMT can face in the future. The fundamental offerings of our paper are as follows: 1st, we present traditional IoMT architecture, examine IoT key technologies carried out in the clinical area, then present the directions of research to increase its efficiency in the healthcare area. With the fast boom of healthcare statistics and complication of data structure, we observe 3-tier cloud computing architecture for IoMT, then announce tools inclusion of IOMT cloud computing application. By evaluating conventional IoMT as well as medical IOT cloud, we recall the benefits of EC and present its IoMT architecture. MEDICAL INTERNET OF THINGS Ning et al. [9] explained that the utility of IoT knowledge in the healthcare area subject understands the digital administration of healthcare data, so medical workers are now not busy recording as well as arranging huge quantities of bulky healthcare data, however greater targeted on patients to give higher healthcare services. In this phase, we are able to present the architecture of conventional IoMT, after which we define the key technology applied. IOMT ARCHITECTURE IOMT architecture envelops a few associated contents of Wireless Sensor Network (WSN) and EPCglobal, as described by Liu et al. [10], as well as is the maximum broadly applied architecture in IoT field. The IoMT can be the focused incarnation of IoT technology within the healthcare sector. Here common 3-tier IOT architecture will be followed. It has 3 layers, namely, transmission, network and perception layer. The architecture of IoMT is shown in Fig. (1).
50 Computational Intelligence for Data Analysis, Vol. 2
Sumathi et al.
Fig. (1). IoMT architecture.
The IoMT's priority and challenge are determined by the perceptual layer. Data acquisition sub-layer, as well as data access sub-layer, are the two main sublayers. To finish the identification as well as the perception of IoMT instances, and to gather information about individuals as well as things, the data acquisition sub-layer uses various kinds of signal acquisition and medical perception equipment. To distribute all things and people associated within the network into easy-to-identify (CPS) Cyber-Physical Systems nodes, it utilizes signal acquisition methodologies, including general packet radio service technology, image recognition technology, graphic code, Radio Frequency Identification (RFID) technology, and numerous sensing devices like DNA sensor, chemical
Cloud Edge Computing
Computational Intelligence for Data Analysis, Vol. 2 51
sensor, physiological signal sensor and physical signal sensor. In the Internet of Things, there are three nodes: Internet CPS, active CPS and passive CPS. According to various needs and objects, appropriate identifier is needed in IoMT. A data access sub-layer attaches the collected data by data acquisition sub-layer to a network layer via short-distance data transfer innovation, including Bluetooth, WiFi, ZigBee and so on, and the major access methodologies must be chosen based on IoMT's overall environmental characteristics and the requirements of various objects. The network layer has two sublayers: the data transfer layer with the service layer. Data transfer layer can be IoMT's backbone system, which can be analogous to a person's nerve centre and brain. It transmits the relevant data obtained by the perception layer in a barrier-free manner, reliable, accurate and real-time using the Internet, mobile communications systems, and some other special networks. The IoMT aims to explore the framework of heterogeneous networks suited for hospitals and original networks, rather than entirely replacing them. The service layer is responsible for integrating heterogeneous networks, as well as multiple data types, descriptions, data warehouses, and other data. Simultaneously, it develops a support service strategy focused on this that offers interfaces to different application layer services, allowing 3rd parties to create appropriate applications for medical professionals and another majority to utilize. Health information applications as well as health information decision-making applications, are the two layers of the application layer. Healthcare data management applications include inpatient treatment data processing, outpatient data processing, patient data processing, medical devices and material management, and so on. Therapy information analysis and diagnosis, pharmaceutical information analysis, illness information analysis, Patient information analytics, and so on are examples of medical information decisionmaking applications. IOMT TECHNOLOGIES Radio Frequency Identification (RFID) RFID can be a novel contactless computerized identity framework, the main technology of IoMT. This applies radio frequency technique to become aware of the certain goal as well as read/write the applicable records; also, particular goal and identification gadget may verify identity without physical touch, i.e., contactless 2-way communication. This has the benefits of robust antiinterference, lengthy distance, and without human intervention, it performed. RFID framework can be commonly integrated using 3 elements: reader and
52 Computational Intelligence for Data Analysis, Vol. 2
Sumathi et al.
information control gadget, and electronic radio frequency tag. Digital Radio frequency tags can be the main component of the RFID framework. Its fundamental feature is to shop appropriate statistics records of the desired target as well as connect with the reader/writer. RFID may accomplish moving object identification in high-speed mode and recognition of multi-target, which can be broadly utilized in personnel identification, waste tracking and medical equipment clinical asset control. It also can be applied to accumulate important symptom information of patients, consisting of (ECG) Electrocardiogram, blood pressure, breath, and so forth, that is useful for applicable evaluation of patient’s illness. Wireless Sensor Network (WSN) In the healthcare industry, WSN is the main research area due to its real-time, security, rapidity, reliability and additional benefits. Its application comprises hospital ICU/general ward, emergency monitoring, patient’s physiological features, real-time monitoring and many others. To enhance the attainment of physiological parameters of patient’s, linearity and high sensitivity sort of wearable graphene healthcare sensor, as presented by da Cruz et al. [11]. To perform the three roles of data collecting, processing, and transmission, WSN incorporates a detector, as mentioned by Xin et al. [12], distributed data processing techniques, communications technologies, as well as other technologies. It is a self-organizing and multi-hop wireless system made up of a variety of tiny sensor instances with wireless computation and communication capacities that are periodically distributed inside or around the monitoring region. It can sense, monitor and gather all categories of data from various objects or environments in real-time, and afterward wirelessly transfer the extracted features to the client. Sun et al. [13] described how different forms of sensors can minimize the chances of people's well-being by monitoring person's pressure levels in real-time via their perspiration, movement speed, and body temperature. Many scholars have been focusing on WSN optimization in recent years. Glazkova et al. [14] suggested that a low-energy security network method related to a smart grid WSN monitoring system can be provided to address the WSN security challenge, while Sun et al. [15] proposes the Balanced-Weighted Shortest Path (B-WSP) routing method to minimize energy through efficiently limiting effects of high-load transmitting on endpoint operations. Zhao et al. [16] propose an optimal clustering optimization technique on compression sensors as well as principal component analysis to reduce the usage of energy. In a nutshell, WSN innovation offers a lot of opportunities for advancement in the healthcare profession.
Cloud Edge Computing
Computational Intelligence for Data Analysis, Vol. 2 53
MIDDLEWARE Middleware, as described by Li et al. [17], performs a key position within IoT, in addition to IoMT. It can be placed among reader/writer and backend application software device, which performs an intermediate function. This may meet the wishes of a large quantity of applications, help distributed computing, as well as face sensor gadgets. It can standardize heterogeneous application environments, including data transmission amongst application systems, by establishing a single interface and platform. It will record the events or data obtained via the sensor system, perform collection, filtering, proofreading and other processing, but then transfer those to back-end application dbms or a RFID reader to recognize the information exchange between the RFID reader and the back-end implementation database system. IoMT middleware accepts interface technology and preferred protocol, which could enhance extraordinary middleware for various healthcare needs and software services, including medical electronic record data broadcast middleware, healthcare worker managing middleware, healthcare kit managing middleware, and so on, and every middleware need to be enhanced dependent on standards and requirements of IoMT utility services to attain broadcast standardization of information. The processing methods, transmission and structure of data acquired through the sensing element are quite diverse due to the IoMT's complicated business operations, greater density of IoT devices and huge scale. We describe the semantic middleware approach, as explained by Yu et al. [18], which integrates semantic technology and technology for perfect interoperability. IOMT APPLICATIONS The IoMT understands intelligence healthcare remedy as well as management of things and people; this decreases the treatment cost, then confirms humans’ fitness. Healthcare statistics have a great sensitivity. Whether the recognition system of identity has, sufficient protection performs a critical function in the IoMT environment. Biometric single-mode identity has medium safety as well as progressively can’t manage the volatile increase of statistics in the medical area. Zhang et al. [19] mentioned that Biometric multi-mode identity, which combines 3 biometrics of finger vein, fingerprint and face, has an excessive recognition rate as well as greater security traits; this can be an unavoidable fashion in the future growth of healthcare area. Whether it’s real-time tracking of important symptoms, as described by Qi et al. [20] as well as the telemedicine realization, this is attached from the aid of wearable gadgets. The look of wearable gadgets, including watches, rings, and wristbands, makes management easier among patients and healthcare workers. But, subversive adjustments in the healthcare field, the interconnectedness of wearable gadgets, flexibility and design speed
54 Computational Intelligence for Data Analysis, Vol. 2
Sumathi et al.
needs to be enhanced. The hybrid lean-agile technique presented by Wang et al. [21] can be favorable to the enhancement of wearable gadgets. IOMT IN CLOUD The conventional IoMT can be hard to obtain the interconnectedness among the huge quantity of various healthcare data systems and healthcare institutions, which ends in a separated healthcare service data island; this can create replicated information in the healthcare process, resulting in a big quantity of resource wastage, as well as boom interoperability trouble among clinical records systems. Recently, the application of CC era in IoMT, including its scalability, sharing and excessive reliability, has found the leap ahead improvement in healthcare. Here, we introduce key technology and architecture of the IoMT Cloud, as well as evaluate the security of healthcare information in the cloud. IOMT CLOUD ARCHITECTURE Fig. (2) describes IOMT cloud architecture is specially separated into 3 layers: healthcare service, service administration and customer layer. The service Administration layer specifically consists of a complete Monitoring Scheme (MS) for System Applications (SAM), Virtual Resources (VRM), Physical Resources Monitoring (PRM), and Management Functions (MF) with Disaster Recovery (DRM), Resource (RM), Service (SM), and User Management (UM).
Fig. (2). IOMT cloud architecture.
Cloud Edge Computing
Computational Intelligence for Data Analysis, Vol. 2 55
HEALTHCARE SERVICE LAYER This can be specially separated into 3 sublayers: Healthcare infrastructure, healthcare platform and healthcare software layer. The healthcare infrastructure layer gives network resources, storage, powerful computing and services related to the resource pool comprised of virtualization. The creation of medical institutions doesn’t want to construct individual data, then rent desired resources in step with the need, to save a charge for a largescale medical institution. Datacenter may be applied to medical evaluation and decision-making data, statistical data of medicinal systems with substances, saving private statistics records of patients, and so on, arranging basis for the combination of medical systems and institutions, analysis and data sharing. The healthcare service layer affords a simple platform as well as equivalent technical assistance to creators of healthcare data machines. Depending on the platform, builders can rapidly form, expand and increase the efficient modules of the healthcare records system, which carries super accessibility to improvement, enhancing improvement effectiveness, and protecting the growth value. The healthcare software layer understands the interface and discharge of purposeful framework of clinical data machine, especially for customers who need to apply the machine. SERVICE-MANAGEMENT-LAYER This part can be the main portion of cloud IOMT as well as the base of the healthcare service layer. It guarantees that the whole healthcare data system may perform constantly and effectively and gives technologies and features in renovation, multi-management, and all sides. It can perform a very good part in collaborating with customers, absolutely examining and understanding customer wishes, which is favorable to the healthcare-service layer to remove or avoid timely faults timely, so enhancing quality and performance of software improvement as well as decreasing enhancement costs. USER LAYER User’s layer confers with customers of healthcare data sharing frameworks, consisting of healthcare institutions at unique ranges, community clinics, disease manipulation centers, and so on. Customers may access the consumer layer via an interface of computers, mobile phones, and additional terminal gadgets, which also turns into a fellow of the cloud after admission. During adoring services provided through cloud computing, customers may transmit their very own sources to the cloud for different customers to apply, to understand resource sharing and real information sharing. For medical doctors, getting the right of
56 Computational Intelligence for Data Analysis, Vol. 2
Sumathi et al.
entry to the cloud can obtain and track affected person data, treatment information, treatment plans and many others. In real-time, patients may select any healthcare institution for remedy in keeping with their situations via a cloud system. IOMT CLOUD TECHNOLOGIES Cloud Computing Cloud computing offers a good possibility for digital hospitals, additionally carrying tremendous modifications to the computation, sharing and processing of medical statistics and facts. Inappropriately, CC carries comfort, but it’s hard for customers to realize whether cloud service is secured or not, so they can't successfully manage the whole. Wang et al. [22] presented a measurable security assessment version to resolve this issue. The safety of records kept in the cloud server can be crucial. Iliashenko et al. [23] mentioned that an active audit system of big data applying fuzzy identity guarantees the integrity of customer statistics. Big Data Big data knowledge normally consists of preprocessing and data collection, data mining, analysis and storage. Conventional data collection foundations are solitary, as well as quantity of records saved, analyzed and managed can be tremendously unimportant. Lin et al. [24] mentioned how many can be managed via parallel database and relational database, at the same time as medical big data series is normally separated into. Unstructured, semi-structured and structured, as well as the size, can be regularly in TB/PB. The accrued information is cleaned, extracted, and analyzed, as well as remaining operations to achieve data highquality. Applying big data generation to correctly mine as well as studies, as mentioned by Ning et al. [25], big medical data will make a high impact on disease treatment and diagnosis as well as medical studies turn out to be a source of encouraging digital hospital construction. Artificial Intelligence (AI) As described by Liu et al. [26], AI can be a novel skill that may expand and simulate the technology, method, theory and application framework of human’s intellectual method and wise manner. The important technologies of AI, as mentioned by Guo et al. [27] and Yang et al. [28], include expert systems, computer vision, robots, Natural Language Processing (NLP), and Machine Learning (ML). Among all, ML is the main core skill in AI. In the healthcare industry, ML expertise can diagnose and predict diseases which in large part avoids low performance, excessive errors, as well as emergent of most important
Cloud Edge Computing
Computational Intelligence for Data Analysis, Vol. 2 57
sicknesses of diagnosis in artificial nature. The software of clinical robots substantially enhances the impact of diagnosis, nursing as well as treatment. Computer vision may rapidly study as well as diagnose medical image information which includes Ultrasound (US), microscope and X-ray images. The utility of AI expertise in the clinical treatment area, as described by Hui et al. [29], carries technical innovation, however, it carries the exchange of healthcare service mode. IOMT CLOUD APPLICATIONS The IOMT cloud can be used to create a healthcare data sharing stage dependent on electronic fitness statistics, which will remedy the issue of records island in special healthcare institutions. This is extremely good for hospitals and patients to add electrical health records shaped through private data, critical symptoms and signs, and other applicable facts of sufferers to the cloud for storing and united control. IOMT EDGE CLOUD The IoMT cloud era uploads healthcare information from terminal gadgets to the remote cloud and proceeds outcomes to the terminal system after computation. But, if the massive quantity of records created through rising clinical gadgets can be updated to CC, it'll motivate enormous pressure on the cloud, resulting in great power consumption and a big delay because of the great cloud load. CC alone can’t assist with a huge dataset as well as give a real-time reaction. Thus, the important thing to the improvement of IOMT cloud, as described by Pathinarupothi et al. [30] and Queralta et al. [31], is to efficiently extend the capability of CC and utilize the resources of distributed computing, i.e., to develop computing jobs in the edge of network, as well as the utility edge computing may meet the particular demand of computing. IOMT EDGE-CLOUD ARCHITECTURE We separate IOMT Edge-Cloud into cloud computing, edge computing and terminal layer. Fig. (3) displays the IOMT cloud edge architecture. The CC layer is within the middle of the whole network. This can be an effective data processing, with network sources, storage and massive computing, that could analyze, summarize, and completely keep the records updated through a layer of edge computing. The layer of edge computing includes a massive quantity of community edge nodes; this can be smart terminal gadgets, which include tablets, smart telephones and so forth, or community gadgets, which include routers, gateways and so on. These edge devices are broadly implemented among cloud
58 Computational Intelligence for Data Analysis, Vol. 2
Sumathi et al.
and terminal systems, including clinics, hospitals, and others. They could offer storage edge computing and network services for obtained facts due to less number of nodes among terminal systems and edge nodes, clinics, hospitals, and so on. They can achieve a greater responsive and agile network. It may lessen delay request, and efficiently prevent information leakage because of lengthydistance broadcasts and other safety problems, which can be critical for extraordinarily sensitive healthcare information. The terminal layer comprises of a selection of IOMT gadgets, like RFID tags, wearable gadgets, medical sensors and many others. This can be the nearest layer to the customer, particularly answerable for gathering information from local equipment and updating data to edge gadgets with input-method as a carrier.
Fig. (3). IOMT cloud edge architecture.
IOMT EDGE CLOUD TECHNOLOGIES Edge Computing As defined by Emam et al. [32], Edge computing is an open-platform that gives the closest service through core abilities of application, storage, computing, and network on network edge near the source of gadgets or information, people. The storage capability and computing of the cloud will sink to network’s edge, as well as unique facts may not be updated to the cloud for processing in centralized manner as a whole lot is viable, then processing and evaluation of data resources may be understood regionally via storage as well as distributed computing. Then variation among cloud and edge-computing lies in 2-way communication between CC center and terminal system, i.e., the terminal gadget may forward requests to
Cloud Edge Computing
Computational Intelligence for Data Analysis, Vol. 2 59
CC center, may be given processing, data storage and united control of CC center as well as complete adjustment techniques and computing tasks issued through CC center. Edge computing expertise has traits of schedulable, definable, high security, low delay and so forth. When a network can be interrupted and abnormal, edge nodes could understand self-recovery and nearby autonomy as well as have excessive robustness. Computational Offload Computational offload expertise specifically has 2 issues: resource allotment and decision-making. Wei et al. [33] mentioned that the decision of computational offloading primarily depends on suboptimal algorithmic program and game theory. It is used in multiaccess edge networks to store overall computing price and gain Nash equilibrium via restricted step length. After the decision-making of unloading, succeeding attention can be rationality of allotment of resources. Resource allotment specifically implements whether or not to unload the computing mission to at least one or greater MEC servers; this relies upon whether or not the computing mission may be separated as well as whether there’s an association among the separated components. Rachakonda et al. [34] mentioned that allotment of resource approach dependent on MEC server(s) can be presented, which may resolve the issue of broadcast delay as a result of resource allotment between more than one MEC server. The resource allotment method, primarily dependent on deterministic-differential equations, as described by Dhunna et al. [35], may correctly allot resources and make the stability and security of the MEC application. IOMT EDGE CLOUD APPLICATIONS Edge computing enhances the working effectiveness of the healthcare experience of sufferers and healthcare workers. As mentioned by Bouadem et al. [36], healthcare workers don’t need to forward affected person fitness facts to remote data-center for processing and waiting for outcomes. They may install a datacenter of edge computing to calculate, process as well as save patient health facts accrued with edge devices domestically that carry high-quality modifications to the clinical area. Yadav et al. [37] described how EC allows wearable health tracking systems to carry out real-time processing and evaluation domestically, even if it is offline, which can be useful for affected person's remote monitoring. Zgheib et al. [38] have presented a fitness monitoring architecture dependent on AI, EC and different technologies. The architecture may be applied to gather EEG, blood stress, ECG, and other kinds of critical sign statistics and enhance disorder analysis accuracy.
60 Computational Intelligence for Data Analysis, Vol. 2
Sumathi et al.
Table 1. The comparison of edge and cloud computing. Features
Edge Computing
Cloud Computing
User experience
Strong
Weak
Computing cost
Low
High
Computing resources
Limited
Unlimited
Reliability
Low
High
Security
More
Less
Real-time
Strong
Weak
Latency
Less
More
Data processing
Fast
Slow
Energy consumption
Low
High
Bandwidth load
Less
More
Server node location
Data-center
Edge network
Architecture
Distributed
Centralized
The comparison of edge with cloud computing is shown in Table 1. Edge computing, we believe, is a natural outgrowth of cloud computing, but it cannot totally replace cloud computing. CC and EC have a complimentary and collaborative relationship. The edge endpoints can quickly process and assess a significant amount of real-time information, but most aren't used only once. It must still be gathered from the edge end to the cloud after it has been processed at the edge. The cloud is required for the storage of critical information, analysis and mining of large amounts of data, and the connection of numerous edge devices, as well as edge management and virtualization resources. Cloud computing and Edge computing can only achieve varied demand situations when they operate closely together, increasing the application value of cloud computing and edge computing. CONCLUSION & FUTURE WORK In our manuscript, we examined conventional IoMT, IoMT cloud, IOMT edge cloud, focusing on healthcare information processing of medical edge cloud as well as the growth of telemedicine. Especially, we initially added conventional IoMT architecture and important technologies involved, mentioning the utility of IoMT technology. Next, we examined the benefits of EC compared with CC and discussed the optimization of EC, as well as presented that cloud edge association may attain maximal application in the healthcare field. After 5G arrival and the constant growth of EC, the forthcoming telemedicine may develop limitlessly. In addition, the growth of conversation protection protocol, distributed intrusion
Cloud Edge Computing
Computational Intelligence for Data Analysis, Vol. 2 61
detection technology in edge, and trust control of edge nodes to save you from malicious and avoid security attacks will be considered as future work. CONSENT FOR PUBLICATON Declared none. CONFLICT OF INTEREST The author declares no conflict of interest, financial or otherwise. ACKNOWLEDGEMENT Declared none. REFERENCES [1]
L. Sun, X. Jiang, H. Ren, and Y. Guo, "Edge-Cloud Computing and Artificial Intelligence in Internet of Medical Things: Architecture, Technology and Application", IEEE Access, vol. 8, pp. 101079101092, 2020. [http://dx.doi.org/10.1109/ACCESS.2020.2997831]
[2]
A. Al-Fuqaha, M. Guizani, M. Mohammadi, M. Aledhari, and M. Ayyash,"The origin and prevention of pandemics", Clin. Infect. Dis., vol. 50, no. 12, pp. 1636–40, 2010. [http://dx.doi.org/10.1109/COMST.2015.2444095]
[3]
W. Wang, S. Yu, T. M. Bekele, X. Kong, and F. Xia, ``Scientic collaboration patterns vary with scholars' academic ages," Scientometrics, vol. 112, no. 1, pp. 329-343, 2017.
[4]
X. Wang, Z. Ning, X. Hu, L. Wang, L. Guo, B. Hu, and X. Wu, "Future communications and energy management in the Internet of Vehicles: Toward intelligent energy-harvesting," IEEE Wireless Commun., vol. 26, no. 6, pp. 87-93, 2019. [http://dx.doi.org/10.1109/MWC.001.1900009]
[5]
Z. Ning, X. Hu, Z. Chen, M. Zhou, B. Hu, J. Cheng, and M. S. Obaidat, "A cooperative quality-aware service access system for social Internet of vehicles," IEEE Internet Things J., vol. 5, no. 4, pp. 25062517, 2018. [http://dx.doi.org/10.1109/JIOT.2017.2764259]
[6]
R.C. Shit, S. Sharma, D. Puthal, A.Y. Zomaya, "Location of things (LoT): A review and taxonomy of sensors localization in IoT infrastructure," IEEE Commun. Surveys Tuts., vol. 20, no. 3, pp. 20282061, 2018. [http://dx.doi.org/10.1109/COMST.2018.2798591]
[7]
X. Liu, A. Liu, T. Qiu, B. Dai, T. Wang, L. Yang, "Restoring connectivity of damaged sensor networks for long-term survival in hostile environments," IEEE Internet Things J., vol. 7, no. 2, pp. 1205-1215, 2020. [http://dx.doi.org/10.1109/JIOT.2019.2953476]
[8]
G. Yang, G. Pang, Z. Pang, Y. Gu, M. Mäntysalo, H. Yang, "Non-invasive _exible and stretchable wearable sensors with nano-based enhancement for chronic disease care," IEEE Rev. Biomed. Eng., vol. 12, pp. 34-71, 2019.
[9]
Z. Ning, P. Dong, X. Wang, X. Hu, L. Guo, B. Hu, Y. Guo, and R. Y. K. Kwok, "Mobile edge computing enabled 5g health monitoring for Internet of medical things: A decentralized game theoretic approach," IEEE J. Sel. Areas Commun., pp. 1-16, 2020.
62 Computational Intelligence for Data Analysis, Vol. 2
Sumathi et al.
[10]
X. Liu, T. Qiu, and T. Wang, "Load-balanced data dissemination for wireless sensor networks: A nature-inspired approach," IEEE Internet Things J., vol. 6, no. 6, pp. 9256-9265, 2019. [http://dx.doi.org/10.1109/JIOT.2019.2900763]
[11]
M.A.A. da Cruz, J.J.P.C. Rodrigues, J. Al-Muhtadi, V.V. Korotaev, and V.H. C. de Albuquerque, "A reference model for Internet of Things middleware," IEEE Internet Things J., vol. 5, no. 2, pp. 871883, 2018. [http://dx.doi.org/10.1109/JIOT.2018.2796561]
[12]
Y. Xin, L. Kong, Z. Liu, C. Wang, H. Zhu, M. Gao, C. Zhao, X. Xu, "Multimodal feature-level fusion for biometrics identi_cation system on iomt platform," IEEE Access, vol. 6, pp. 21418-21426, 2018.
[13]
Y. Sun, H. Qiang, J. Xu, G. Lin, "Internet of Things-based online condition monitor and improved adaptive fuzzy control for a medium-lowspeed maglev train system," IEEE Trans. Ind. Informat., vol. 16, no. 4, pp. 2629-2639, 2020
[14]
N. Glazkova, C. Fortin, and T. Podladchikova, "Application of lean-agile approach for medical wearable device development", Proc. 14th Annu. Conf. Syst. Syst. Eng. (SoSE),. pp. 75-80, 2019. [http://dx.doi.org/10.1109/SYSOSE.2019.8753825]
[15]
A. Sun, G. Gao, T. Ji, and X. Tu, "One quanti_able security evaluation model for cloud computing platform", Proc. 6th Int. Conf. Adv. Cloud Big Data (CBD). pp. 197-201, 2018.
[16]
C. Zhao, L. Xu, J. Li, F. Wang, and H. Fang, "Fuzzy identity-based dynamic auditing of big data on cloud storage," IEEE Access, vol. 7, pp. 160459_160471, 2019. [http://dx.doi.org/10.1109/ACCESS.2019.2950938]
[17]
M. Li, C. Wang, L. Yan, and S. Wei, "Research on the application of medical big data", Proc. 14th Int. Conf. Comput. Sci. Edu. (ICCSE),. pp. 478-482, 2019. [http://dx.doi.org/10.1109/ICCSE.2019.8845376]
[18]
Y. Yu, M. Li, L. Liu, Y. Li, and J. Wang, "Clinical big data and deep learning: Applications, challenges, and future outlooks," Big Data Mining Analytics, vol. 2, no. 4, pp. 288-305, 2019. [http://dx.doi.org/10.26599/BDMA.2019.9020007]
[19]
K. Zhang, S. Leng, X. Peng, L. Pan, S. Maharjan, and Y. Zhang, "Arti_- cial intelligence inspired transmission scheduling in cognitive vehicular communications and networks," IEEE Internet Things J., vol. 6, no. 2, pp. 1987-1997, 2019.
[20]
W. Qi, B. Landfeldt, Q. Song, L. Guo, and A. Jamalipour, "Traf_c differentiated clustering routing in DSRC and C-V2X hybrid vehicular networks," IEEE Trans. Veh. Technol., early access, 2020, doi:10.1109/TVT.2020.2990174. [http://dx.doi.org/10.1109/TVT.2020.2990174]
[21]
W. Wang, J. Chen, J. Wang, J. Chen, and Z. Gong, "Geography-aware inductive matrix completion for personalized point-of-interest recommendation in smart cities," IEEE Internet Things J., vol. 7, no. 5, pp. 4361-4370, 2020. [http://dx.doi.org/10.1109/JIOT.2019.2950418]
[22]
W. Wang, J. Chen, J. Wang, J. Chen, J. Liu and Z. Gong, "Trust-Enhanced Collaborative Filtering for Personalized Point of Interests Recommendation," in IEEE Transactions on Industrial Informatics, vol. 16, no. 9, pp. 6124-6132, 2020. [http://dx.doi.org/10.1109/TII.2019.2958696]
[23]
O. Iliashenko, Z. Bikkulova, and A. Dubgorn, Opportunities and challenges of arti_cial intelligence in healthcare.Proc. E3S Web Conf. vol. Vol. 110. EDP Sciences: Les Ulis, France, 2019, p. 02028.
[24]
P. Lin, Q. Song, and A. Jamalipour, "Multidimensional cooperative caching in CoMP-integrated ultradense cellular networks," IEEE Trans. Wireless Commun., vol. 19, no. 3, pp. 1977-1989, 2020. [http://dx.doi.org/10.1109/TWC.2019.2960329]
[25]
Z. Ning, X. Wang, and J. Huang, "Mobile edge computing-enabled 5G vehicular networks: Toward
Cloud Edge Computing
Computational Intelligence for Data Analysis, Vol. 2 63
the integration of communication and computing," IEEE Veh. Technol. Mag., vol. 14, no. 1, pp. 54-61, 2019. [http://dx.doi.org/10.1109/MVT.2018.2882873] [26]
X. Liu, T. Qiu, X. Zhou, T.Wang, L. Yang, and V. Chang, "Latency-aware path planning for disconnected sensor networks with mobile sinks," IEEE Trans. Ind. Informat., vol. 16, no. 1, pp. 350361, 2020. [http://dx.doi.org/10.1109/TII.2019.2916300]
[27]
H. Guo, J. Liu, and J. Zhang, "Effcient computation offoading for multiaccess edge computing in 5G HetNets," in Proc. IEEE Int. Conf. Commun. (ICC), May 2018, pp. 1-6.
[28]
Y. Yang, Y. Wang, R. Wang, and S. Chu, "A resource allocation method based on the core server in the collaborative space for mobile edge computing," in Proc. IEEE/CIC Int. Conf. Commun. China (ICCC), pp. 568-572, 2018. [http://dx.doi.org/10.1109/ICCChina.2018.8641218]
[29]
H. Hui, C. Zhou, X. An, F. Lin, "A new resource allocation mechanism for security of mobile edge computing system," IEEE Access, vol. 7, pp. 116886-116899, 2019. [http://dx.doi.org/10.1109/ACCESS.2019.2936374]
[30]
R.K. Pathinarupothi, P. Durga, E.S. Rangan, "IoT-based smart edge for global health: Remote monitoring with severity detection and alerts transmission," IEEE Internet Things J., vol. 6, no. 2, pp. 2449-2462, 2019. [http://dx.doi.org/10.1109/JIOT.2018.2870068]
[31]
J.P. Queralta, T.N. Gia, H. Tenhunen, T. Westerlund, "Edge-AI in LoRa-based health monitoring: Fall detection system with fog computing and LSTM recurrent neural networks," In Proc. 42nd Int. Conf. Telecom-mun. Signal Process. (TSP), pp. 601-604, 2019. [http://dx.doi.org/10.1109/TSP.2019.8768883]
[32]
A. Emam, A.A. Abdellatif, A. Mohamed, and K.A. Harras, "Edge-Health: An energy-ef_cient edgebased remote mHealth monitoring system," InProc. IEEE Wireless Commun. Netw. Conf. (WCNC), pp. 1-7, 2019.
[33]
W. Wei, and Y. Qi, "Information potential fields navigation in wireless ad-hoc sensor networks", Sensors (Basel), vol. 11, no. 5, pp. 4794-4807, 2011. [http://dx.doi.org/10.3390/s110504794] [PMID: 22163876]
[34]
L. Rachakonda, P. Sundaravadivel, S.P. Mohanty, E. Kougianos, M. Ganapathiraju, “A smart sensor in the iomt for stress level detection,”in 2018 IEEE International Symposium on Smart Electronic Systems(iSES)(Formerly iNiS). IEEE, pp. 141–145, 2018.
[35]
G.S. Dhunna, and I. Al-Anbagi, "A low power wsns attack detection andisolation mechanism for critical smart grid applications", IEEE Sens. J., vol. 19, no. 13, pp. 5315-5324, 2019. [http://dx.doi.org/10.1109/JSEN.2019.2902357]
[36]
N. Bouadem, R. Kacimi, and A. Tari, “B-d’wsp selection algorithm: A loadbalancing convergecast for wsns,” in 2018 Wireless Days (WD). IEEE, 2018, pp. 101-103. [http://dx.doi.org/10.1109/WD.2018.8361702]
[37]
S. Yadav, V. Kumar, S.B. Dhok, D.N. Jayakody, “Energy-efficient design of mi communication-based 3-d non-conventional wsns,” IEEE Systems J., vol. 14, no. 2, pp. 2585-2588, 2020. [http://dx.doi.org/10.1109/JSYST.2019.2918184]
[38]
R. Zgheib, E. Conchon, and R. Bastide, “Semantic middleware architecturesfor iot healthcare applications,” in Enhanced Living Environments. Springer, 2019, pp. 263-294. [http://dx.doi.org/10.1007/978-3-030-10752-9_11]
64
Computational Intelligence for Data Analysis, 2023, Vol. 2, 64-82
CHAPTER 5
Survey of IoMT Interference Mitigation Techniques for Wireless Body Area Networks (WBANs) Izaz Ahmad1, Muhammad Abul Hassan1,*, Inam Ullah Khan2 and Farhatullah3 Department Computing and Technology, Abasyn University, Peshawar, Pakistan Kings College London, London, United Kingdom 3 School of Automation, China University of Geosciences, Wuhan, China 1 2
Abstract: Medical data can be stored and analyzed using the Internet of Medical Things (IoMT), which is a collection of smart devices that link to a wireless body area network (WBAN) using mobile edge computing (MEC). The Wireless Body Area Network (WBAN) is the most practical, cost-effective, easily adaptable, and noninvasive electronic health monitoring technology. WBAN consists of a coordinator and several sensors for monitoring the biological indications and jobs of the human body. The exciting field has led to a new research and standardization process, especially in WBAN performance and consistency. In duplicated mobility or WBASN scenarios, signal integrity is unstable, and system performance is greatly reduced. Therefore, the reduction of disturbances in the project must be considered. WBAN performance may compromise if co-existing other wireless networks are available. A complete detailed analysis of coexistence and mitigation solutions in WBAN technology is discussed in this paper. In particular, the low power consumption of IEEE 802.15.6 and IEEE 802.15.4, 3 of one of WBAN's leading Wi-Fi wireless technologies, have been investigated. It will elaborate on a comparison of WBAN interference mitigation schemes.
Keywords: IEEE 802.11, IEEE 802.15.4, IEEE 802.15.6, Interference Avoidance Schemes, IoMT, WBAN, Wi-Fi. INTRODUCTION The Internet of Things (IoT) has transformed modern healthcare, gradually displacing traditional therapies in favor of ubiquitous healthcare technologies. As a result, medical practices must design strategies for providing the best possible Corresponding author Muhammad Abul Hassan: Department Computing and Technology, Abasyn University, Peshawar, Pakistan; E-mail: [email protected]
*
Mariya Ouaissa, Mariyam Ouaissa, Zakaria Boulouad, Inam Ullah Khan, Sailesh Iyer (Eds.) All rights reserved-© 2023 Bentham Science Publishers
Wireless Body Area Networks
Computational Intelligence for Data Analysis, Vol. 2 65
patient care while yet being profitable. WBAN has evolved into a critical tool for obtaining healthcare services at any time and from any location. In real time, these networks can gather physiological, behavioral, and other health-related data. As a result, the WBAN may give those in need of emergency medical treatment with cost-effective and dependable real-time e-healthcare services. The World Wide Web of Things (WBAN) is also known as the Internet of Medical Things (IoMT) because it affects people by monitoring their health using wearable or implantable medical sensors. Wireless Body Networks (WBAN) have many small and inactive devices (sensors) with a wireless connection. Sensors can be placed inside or on the body for the purpose of monitoring health. WBAN is made up of many sensors and coordinators. The sensor is sent to the coordinator of the health control center, the environmental and human functions perceive and transmitted to the coordinators. WBAN applications, as described by Tobon et al. [1], can be used in two areas: (i) medical applications for patient monitoring, (ii) non-medical applications such as biometrics, which is used for training. Communication uses two standards: IEEE 802.15.4 and IEEE 802.15.6. IEEE 802.15.4 identifies the levels of MAC control for short-term communications. Although IEEE 802.15.6 was proposed for localization in the body. WBAN is widely regarded as technical support for many applications, including health and safety monitoring and emergency situations. The most modern invention, the electronics, the sensor, as described by González-Valenzuela et al. [2], is a wireless transceiver, which eliminates the need for the receiving node and the communication infrastructure to transmit the observed data with one or more sensors; the Microcontroller Unit (MCU) allows devices and low-power concepts that can be linked together to create. The human body is mobile, and mobility is supported by this. Suppose the other WBAN service area penetrates the arm or leg transmission. Also, the devices that are in diameter enter the mobility. Interference between WBAN interference sensors (WBAN) of WBAN interference, (ii) interference between the same network using the same frequency band, WBAN, noise, and (iii) interference classification. Interference between domains created by different network interferences with coexistence can be seen in Fig. (1). Multiple techniques exist for interference mitigation and to improve the network parameters like data rate, energy, and latency. Existing techniques will be discussed in section II. Section III describes the conclusion from all existing techniques. Section IV presents a discussion, and finally, references are given in section V.
66 Computational Intelligence for Data Analysis, Vol. 2
Ahmad et al.
Fig. (1). Coexistence WBANs creating interference.
Difference Between WBAN vs. WSN Concerning IoMT WBAN, as described by Le et al. [3], Yang et al. [4], and Sana Ullah et al. [5], is an effective standard for eHealth applications and telemedicine applications in the introduction starts with IEEE standards 802.15.4 and 802.15.6. It observes and records the main signs and significant changes in the patient's condition. There is a consensus among scientists that WSN and WBAN are not exactly the same. As mentioned by Kailas et al. [6], Chen et al. [7], and Khan et al. [8], there are some differences between WSN and WBAN. Protocols developed for WSN and ADHoc networks may not match WBAN properties. WBAN size and energy fields are stricter than WSN. In addition, the WBAN information includes medical data that causes reliability, security, and time problems for WSN applications. Due to health risk concerns, transmission control is extremely limited in WBAN. Therefore, the range and transmission range of WBAN is considerably smaller than WSN. Finally, WBAN sensor nodes change the requirements and characteristics because the WSN's sensors are uniform and play a role in comparison. As mentioned by Lai et al. [9], the IEEE 802.15.6 WBAN standard was proposed as a talented wireless invention for very fewer power devices, for example, IEEE 802.15.4 or IEEE 802.15.6 utilized as a part of a human body. This innovation is particularly proposed for WBANs process and arrangement condition, creating it a reasonable choice for various applications in medical and different fields. The different frequency bands used for the transmission contained
Wireless Body Area Networks
Computational Intelligence for Data Analysis, Vol. 2 67
in the IEEE 802.15.6 standard used: 400 groups collected, narrowband (NB), 800, 900 MHz and 2.3 GHz, and 2.4. Human body communication Ultra-wideband (UWB) (HBC) uses 3.1 to 11.2 GHz with a frequency of 10 to 50 MHz. Therefore, analysts agree that the 2.4 GHz frequency band, adding more to the aptitude to provide head-to-head interference, is a challenging area used in medical applications Wireless Sensor Network (WSN) The first difference is that the width and number of sensors sent to the WSN are much higher than the WBAN. This provides various requirements for WBAN and communication protocols used at different levels. The IEEE 802.15.4 (ZigBee) standard, as described by Kwak et al. [10] and Fei et al. [11], is utilized broadly for WSN execution where it suggests a bigger scope zone notwithstanding having the best execution under interference when matched with Bluetooth. This survey was not conducted to study the problem of close coexistence with this WBAN research on wireless technology research. It aligns with numerical modeling research and one of the consequences of using low-power Wi-Fi in the WBAN and ZigBee and IEEE 802.15.6 standards. The proposed approaches in the literature review can be examined by modifying the parameters based on the interrogated radio technology. WBAN ARCHITECTURE Fig. (2) depicts a person using an IoT device that transmits data to his medical team for remote monitoring and real-time vital sign fluctuations. On-body and implanted IoT devices are the two types of IoT devices in IoMT. Implantable IoT devices include a heart arrhythmia monitor and an endoscopic capsule. A wearable IoT device, such as a watch, remains attached to the body. WBAN contains sensor nodes that are implanted inside the body, on the body, or around the body for behavioral examination. These sensors frequently look at human fundamental data and generally utilize wireless communication for connecting to the sink node. The sink node first gathers information from every single other node and transmits data to the monitoring staff. Based on the information, the medical staff can take the appropriate medical action.
68 Computational Intelligence for Data Analysis, Vol. 2
Ahmad et al.
EEG Sensor
ECG Sensor
PPG Sensor
cellular
Hub
4G
Internet
Medical Server
Sp02 Sensor Wifi
Motion Sensor
Wireless Body Area Network
Medical Service Level
Fig. (2). WBAN architecture.
WBAN APPLICATIONS Rehabilitation and Therapy This includes the person most likely to cause a stroke and, to the extent possible, the disease if the patient continues to develop during treatment. Therefore, the difference between human development research and housing reconstruction is clearly an important film. Sensor extension, multi-sensor information combination, data integration and virtual patient development, such as the structure and definition of limitations and specific requirements in the field of research. As mentioned by Liu et al. [12] and Hadjidj et al. [13], recovery is not helpful for patients to regain the ability to use it after discharge to rearrange the drug.
Wireless Body Area Networks
Computational Intelligence for Data Analysis, Vol. 2 69
Wearable Health Monitoring System The wearable health monitoring system (WHMS) is a recent example of an ambulatory system created by scientists Zhou et al. at the University of Alabama [14]. This examination progresses a big system, health status checking, and telemedicine. WHMS utilizes old Wi-Fi technology. And cellular systems are used to transmit information from BANs to an external system and gather information by utilizing various kinds of devices, for example, PCs. Medical experts can get information through the Internet, which assists in problem caution when a medically related abnormality is recognized. Disaster Aid Network Harvard University developed a system called The Disaster Aid Network (AID-N) to assess the health of workers in major emergencies. The system utilizes an online interface to encourage connections among first responders. Like WHMS, AID-N utilizes Wi-Fi and cellular systems to build up communication between the individual, servers-based smartphone, and the system's database servers. TECHNOLOGIES Bluetooth As described by Milenković et al. [15], Bluetooth technology was introduced as a short-range radio frequency for security. It uses a piconet node that should be synchronized with the system clock using the transition method specified by the receiving node. Each device can simultaneously interact with seven different devices in a piconet, a given network, and a device as a receiving node exceeding seven for the entire period of devices. Each device can work with several piconets when it enters the radio next to another receiver. It has the ability to communicate without displaying the network device. Therefore, it is usually used to connect various stand-alone devices to assist voice information and applications. Scientific and medical ISM 2.4 GHz band uses 1600 low frequencies per second to condense interference. This standard defines three plans of equipment with dissimilar transmission control functions and requires attention from 1 to 100 meters. The most extreme data transfer rate is 3 Mbit/s. Low Energy Bluetooth Bluetooth Low Energy (BLE), as described by Gao et al. [16], is ideal for WBAN applications. Low maintenance BLE is designed to route wireless networks from small devices to general-purpose devices. Moreover, these devices are too small to work with energy, they also have costs connected with a standard Bluetooth
70 Computational Intelligence for Data Analysis, Vol. 2
Ahmad et al.
radio receiver, but they are perfect for health surveillance applications. Sustained data transmission speeds of up to 1 Mbit. The frequency shift of the universal spectrum allows the coexistence of BLE and Wi-Fi, but because it works in the GHz ISM 2.4 band, the cohabitation with the various devices becomes a problem. Bluetooth supports applications with two identical devices connected to the advertising network, with different data transfer speeds and power that is beneficial for high data rates, for example, between two separate servers, two WBANs, or between WBANs and PCs. ZigBee ZigBee is a wireless networking technology widely used in low-energy environments. ZigBee not only supports 128-bit security for a low communication data rate, but it also supports long battery life and secure network authentication, designed for high-frequency applications that require integrity and confidentiality. The ZigBee Association determines the level of demand and determines security software at the network and application levels. ZigBee's innovation is divided into two parts. First of all, ZigBee organizes the security levels of applications and software applications 802.15.4. Another is described in IEEE 802.15.4 and used in the physical media access control layer, where CSMA/CA radio tube technology does not follow the ZigBee device. ZigBee WBAN problems are WLAN noise transmission, especially at 2.4 GHz, compared to other wireless networks. Another weakness of ZigBee is the low data rate (250 kbps). The fact that the data rate is low and easily expressed in all medical centers is used. IEEE 802.11 IEEE 802.11, as described in the “Mobile Health” book [17], is a special standard used for WLAN. Using IEEE 802.11, WiFi allows the high-speed Wi-Fi network to surf the Internet, in combination with an access point or in ad-hoc mode. The modified data character offers many features, making it ideal for wireless broadband video conferencing that praises video streaming. Wi-Fi is an important aspect that has affected all counterfeit mobile phones, Macs, and tablets. However, Alam et al. [18] mentioned that high energy consumption is a major disadvantage. IEEE 802.15.4 IEEE 802.15.4 has been designed for low power, low data rate, and low-cost applications. The IEEE 802.15.4 was released in 2003. This standard was operating 869 MHz, 915 MHz, and 2400MHz bands all over the world. This protocol has worked in the star and peer-to-peer topology.
Wireless Body Area Networks
Computational Intelligence for Data Analysis, Vol. 2 71
IEEE 802.15.6 IEEE 802.15.6, as described by Mühlbacher et al. [19], is a new standard that allows various internal and external communications as well as human body control for medical and non-medical WBAN applications. The IEEE 802.15.6 standard is a 400, 800, 900 MHz, 2.3, and 2.4 GHz narrowband (NB) frequency band and an ultra-wideband (UWB) with 3.550 MHz HBC mentioned in the “IEEE Standard for Local and metropolitan area networks - Part 15.6: Wireless Body Area Networks” [20] and 3.1 GHz, 11.2). This standard is revolutionary in mobile phone sensor networks and is particularly suitable for various wireless networks. Many NAG sensor nodes (256) and enough information for lymphatic work, less energy consumption are defined by the requirements of the application. With the IEEE 802.15.6 standard, data transfer rates of up to 10 Mbps and very low power consumption are possible. Moreover, you can visualize the mobility of the WBAN (that is, coexistence moves directly to another WBAN zone). This is not applicable for WBAN applications that require the following situations: At a maximum level, this standard can meet most bandwidth requirements at 680 kbps for WBAN applications. However, it is impossible to eliminate the limitations of new applications requiring bright audio and video transmission. TECHNIQUES AND COMPARISON Some of the researchers worked on the coexistence of many sensors using the same frequency, called homogeneous coexistence. The homogeneous problem can be treated by central or distributed techniques. Central entities govern the use of time and frequency under the coexistence networks. Yang et al. [4] suggest a central appliance in which a resource-linked server requests from WPAN coordinators and the resources to be applied. In addition, the proposed technique assigns parallel transmission and synchronous intervals, and significantly minimizes the number of designated channels. The approach regards the interference at the sensor level and the time interval, which significantly reduces the latter, as well as improves the energy saving of the WBAN. At the same time, the approach mentioned by Chen et al. [21] uses Bayesian gaming power management technology to mitigate intervals between WBANs. Other approaches include reusable schemes to mitigate the effects of interference. To perceive and diminish interference of coexistence, WBANs Lightweight and Robust Interference Mitigation scheme (LRIM) was proposed by Zou et al. [22] based on Beacon Delivery Ratio (BDR). The author use Transmission Efficiency (TE), which includes the number of packets received by the receiver, many retransmissions, and how many times perform back-offs. TE and BDR threshold values were performed to specify the interference. Decisions are taken on channel
72 Computational Intelligence for Data Analysis, Vol. 2
Ahmad et al.
hopping by the coordinator, and channel hopping requests on sensors. If the request is sent by more than half of the sensors, the coordinator takes the decision for channel hopping on a candidate (based on RSSI and traffic rate). If a free channel is found, switch the whole operation to that channel. The authors proposed Asynchronous Inter-network Interference Avoidance (AIIA) proposed by Ma et al. [23] using CSMA/CA and TDMA. Listening to the controlling messages from other WBANs and then updating its scheduling to avoid interference.
Fig. (3). AIIA superframe structure.
Fig. (3) shows the AIIA superframe, which contains beacon (B), CR, and SR. AIIA superframe made some changes in IEEE 802.15.6 superframe, which include: (i) CR is parallel to the emergency access period (EAP), random access period (RAP), and contention access period (CAP). (ii) SR in AIIA is divided into many periods, comparable to the Medium Access Period (MAP) in IEEE 802.15.6. In AIIA, each coordinator performs two operations: 1st operation, the coordinator maintains the AIIA table by repeating OWD periodically. The coordinator uses a message to exchange information with each other. In the case of a homogeneous coordinator of each WBAN floods SR_AD message before its SR period. The coordinator creates an AIIA table containing six columns. A gateway identifier (GUID) is used to collect the received device identifier. The next column is hop count. The third one is superframe offset to save time between the coordinator and neighbor timer. The next fields are the start time and duration of TDMA, which shows the starting time and SR period. Seq-No indicates the updated entries. In a 2nd operation called SRR, the coordinator rearranges its inconsistent SR. SRR may have three phases. In the first phase, it refers to its own AIIA table, looks for idle
Wireless Body Area Networks
Computational Intelligence for Data Analysis, Vol. 2 73
time in the CR, recreates incompatible SR periods, and translates the SR_AD message. The coordinator finds the roaming period (RP) between the CRs of the coexisting WBAN and assigns the key to the RP in the second phase. The coordinator determines the new SR position and the transmission time of the SR_AD message in the third phase, then redirects the SR and forwards it to the homogeneous WBAN coordinator. Kim et al. [24] expressed a demonstration of a fast Fuzzy Power Connection (FPC) scheme as a blurry structure with a response channel to reduce transmission control and increase the interface bandwidth. The genetic algorithm (GA) is associated with the definition of the ideal parameters of a fuzzy controller. The demonstration of the structure includes an I/O source and a fuzzy power controller. The information signal is SINR, current interference control and the input signal. The output signal is the current transmission control and is filled as a critical channel for the structural contribution. The FPC examines its sources of information and determines the level of control of transmission at the exit. FPC consists of fuzzy learning infrastructure, a new I/O, a viewfinder, a fuzzy noise mechanism, and an epilator. GA is used in a fuzzy information base to drive a fuzzy engine. GA is divided into four phases. The first is known as a chromosome, in which the parametric quality and the management capacity are mentioned individually in the capacity of the fuzzy set and in the fuzzy theoretical basis of the data source. In the second phase, we will use the hybrid manager and conversions to acquire new ones. The third level is known as the learning system that chooses the best. In the final phase, we try to strengthen the fitness function and adapt to the connection limit, the transmission power, and the number of accents to be combined. Le et al. [25] explained how the power distribution algorithm minimizes noise and the need for QoS. This algorithm uses an optimization model to limit the transmission power of the sensors to the WBAN by performing the necessary transmission control with the basic requirements of the QoS sensor requirements. GA connects to WBAN to solve the problem of improvement. The GA model is a process of organic development in which the population is sensitively managed, and the fitness function is the optimal method for allocating energy with a minimum total capacity with QoS restrictions. Interference between non-conjugated WBAN channels is analyzed by Rezvani and Ghorashi [26]. The network topology was modeled in a two-dimensional geometric representation in which all the WBAN transmission intervals were modeled as a single-diagram diagram. In addition, the network topology developed a tagged WBASN and an interactive WBASN, which was designed and developed into a larger geometric probability function. Based on this model,
74 Computational Intelligence for Data Analysis, Vol. 2
Ahmad et al.
the gamma distribution function is used to evaluate the overall suitability of WBAN. The interference effect was analyzed to maintain the peripheral distance for an acceptable SINR ratio (signal-to-noise ratio plus noise). For example, the minimum network interval requires more than 7-12.5 m to provide SINR devices of at least 9-18 m to provide the average SINR required for WBAN. The interference algorithm analyzed in the WBASN cyber-physical system was developed by Wang et al. [27] by identifying the network topology and applying PCG. In a scenario where people trust WBAN, their positions have changed the network topology, interfering with other WBANs. This mechanism registers two algorithms: the management of games and a network of social contacts consisting of WBAN. The social network WBASN has developed a social network to create a social interaction network. An interference pattern of the overlapped communication area was created, where the buttons were nodes and the edges were the links between the two errors. The recognition of dynamic social interaction was developed using Bluetooth technology and acoustic waves. The distance between the two WBANs was determined by the acoustic wave and the time required by WBASN to estimate the distance between them. To measure the interaction and the dynamic topology along the Markov chain, a dynamic algorithm was used for the four-state interaction. The fourth law of the predictor can be considered a strong interaction, a weak interaction, a non-interactive, and non-invasive-weak interaction. Zhang et al. [28] described how a MAC-based QoS MAC planning scheme for several co-existing WBANs derives from a health monitoring scenario. Coordinators-coordinators exchange information before data transfer. Therefore, the coordinator determines which sensor can access the channel. The channel is divided into beacon periods or super MAC symbols. In each WBAN, the coordinator classifies the network traffic according to the priorities of the IEEE 802.15.6 1 standard and assigns a reduction of the time slots to the general fishery policy. The user with the highest priority gets the first available position. The coordinator of the first WBASN tag sends identification and time information. If other coordinators are in the service area and listen to this program, they store this information in the appropriate table. If the frequency range is exceeded and the transmission intervals are distributed, the priority level of the intervals in the interference zone is compared, and the lower priority interval is delayed for the next transmission. The next day, the coordinator sends the traffic information to all the devices. Peripherals know when to send the next superframe to avoid interference between WBAN. As mentioned in the IEEE ICC 2014 Report [29], information about the data exchange between WBANs is recorded at a distributed distance WBAN.
Wireless Body Area Networks
Computational Intelligence for Data Analysis, Vol. 2 75
According to this survey, WBAN will detect differences between the rests in the transmission interval before being sent to each WBAN in the average range. The first advantage of this comes from the condition of interaction. The WBAN accepting this information must recognize the possibility of avoiding interference. Another confidence is that WBAN can update the network connection and adjust the transmission effect. However, we must respect the moment when we discover and change the information about the community. This data should be fast and complete, as WBAN must exchange information about your channel, SINR, network usage, and prefix of the company before processing. Dynamic Coexistence Management (DCM), as described by Zhang et al., is a new Mac system algorithm designed for general WBASN based on the IEEE 802.15.4 standard. WBASN with DCM can detect and treat unacceptable infections among people. The signal that many WBASNs can record can also cause disruption of lead accidents and accidents. To resolve these conflicts, DCM uses beacon exchange and channel change. In the beacons, a CCC did not work, so those who deal with spies are trying to send the biggest victims. At the end of the assessment rate (PCP), Loft can detect the administrator's luck. If you have just acquired a reputation, see the Administrator for the current Super Frame feature and the current channel. If the guillotine concludes that beer is only closed once, the dialogue will start normally. On the contrary, it is necessary to lose the second position, or when using a WBASN, to replace the beacons at the clock. When beacon loss in DCM, the current file can be run while the beacons (like) big and terrible decisions about bacon appear. There are two events that protect the guard and the lighthouse with time. Or replace the headlamp with the largest division. In the data transfer data, the administrator shows the data at the end of the CFP. Perhaps this control framework consists of responses or negative identification status. In DCM mode, Waechter looks at the inspectors and looks for the possible conversion channel. Two options are full B search and functional scanning. These are the following. In detail, the administrator has detailed information about WBASN via the channel and on their own time. In case of inactive time, you can find the files for the gaps without information about the discovery channel, but without the naval construction. When antivirus is cracked on the channel, other channels are enabled during the next booking. If there is no delay in Super Squad (SD = BI), In the graphic, WBAN is represented by buttons and channel research (e.g., Time interval, frequency bands) through the graph of the graph. The boundary between the two keys is the neighborhood (coexistence) of the WBAN. In this technique, the different colors of different channels (different channels) of WBAN can be avoided. The affirmation force is the focus of the WBAN voting research, as there are several WBAN communication lines.
76 Computational Intelligence for Data Analysis, Vol. 2
Ahmad et al.
Deylami et al. [30] point out that independent original images for mobile WBAN are expected to recur and have small temporal complexity. Two channels look at TDMA, one for WBAN and the other for WBAN communication. For the first WBAN graphics, signing, communicating, and removing WBANs with the WBAN channel are required. The Intra WBAN channel is then used by WBAN sensors to transmit the critical data and the intervals of the respective centers. The color chart is found for all types of periods. In addition to frequency-based methods, frequent agility approaches also have great demand. ZigBee interference and WiFi confirmation method (ACK). The authors concluded that more than 30% of the ZigBee band lobes had lost cocaine due to the rebellion. The rate at which ACK is lost means the strength of the collision. By asking for the appropriate timestamp for the ACK, the network can actually decrease. Inoue and Inoue's explanation. To deploy WLAN transmission using WLAN (send you) and CTS (Clear to send). At this point, implementing WPAN and WLAN Node is a method of connecting the WLAN Channel to the ZigBee network and switching to the book. However, dual-mode buttons of high-density multi-hop WSN are usually WSN multi-hop and cannot operate, such as the transmission between the coordinator and the router, and the transmission between the router and the sensor node. To this end, the dual node mode bits are used by jokes and CTS for interference between the sensor node, the sending node between the coordinator and the router, and the wireless network, and to avoid incidents by the router and multiple nodes, it is between. More schemes that do not have an intermediate active period for sending traffic to the WSN and are unemployed to send crossWi-Fi so that the wireless WSN can rehabilitate reversal. However, this method cannot be applied to multi-hop WSN, as the router's unemployment period possesses communication between the routing sensor and the router and such unemployment period. Instead, our contribution from the difference between the last lab and the two aspects: (1) while operating twice per floor to provide the country of the node gateway, previous research has shown that such interference network I was concentrating on it.; (2) 6 LoWPAN and WLAN are associated as the latter data source so that the former corresponds. When 6 LoWPAN is transmitting, a collision between networks arrives. This sets the mood between the 6 LoWPAN networks and the WLAN network. This is far more complicated than discussed in the literature. However, the BB algorithm can effectively reduce the interruption problem between networks and can greatly improve the performance of heterogeneous WSN. Cheng et al. [31] proposed the design and implementation of an ultra-wideband (UWB) digital baseband transceiver for the purpose of WBAN. The authors had claimed that power dissipation, as well as the region of the proposed design, was reducing the use of Bose-Chaudhuri-Hocquenghem (BCH) and design level adjustment. This research has reduced the area by 42% and 38% has reduced the
Wireless Body Area Networks
Computational Intelligence for Data Analysis, Vol. 2 77
power dissipation with respect to the conventional structure. The transmitter has been designed of 90 Nanometer using CMOS technology with the energy used in the transmitter of 73 PJ/bit and 225 PJ/bit in receiver mode. The authors used IR-UWB technology because the authors claimed that IR-UWB low-power discrete time-based narrow pulses. IR-UWB has a lot of merits with respect to continuous carrier-based technology, using this technology transceiver architecture structure is very simple, has high data rate, and low energy per bit. Manchi et al. [32] proposed an analysis of CSMA/CA under a non-saturation load. In this research, the author found four customer priorities instead of eight priorities to achieve the most stringent requirement for WBAN performance. The authors had shown in the results better performance using the authors' proposed research technique. In IEEE 802.15.6 EAP and RAP, time periods are inefficient using bandwidth. At the start of the RAP time period, the frame collision probability has increased. Rashwand et al. [33] proposed aware interference and energy in WBAN. This research was based on nodes' mobility within the network with respect to the comparative movement to each other. The authors claimed that they reduced the transmission time, interference, energy consumption, and network power consumption when nodes are moving and increased the signal strength, and traffic load, and higher packet delivery ratio. The authors proposed an adaptive interference mitigation scheme when nodes move. Moreover, two schemes (interference reduction and avoidance) have been used in this research. In the Interference reduction scheme, data can be sent to the receiver using different power, and data rates. In the interference avoidance scheme, the coordinator assigned orthogonal channels to every node in a network. Movassaghi et al. [35] researched in output as well as the presentation of IEEE 802.15.6. The authors have researched the mathematical formula and found the highest output with lesser delay limits in IEEE 802.15.6. The limitation has been derived using a lot of frequency bands as well as data rates. The authors claimed that this research has been very helpful to the new researcher who works on IEEE 802.15.6. The authors used 420-450 MHz, 863-870 MHz, 902-928 MHz, and 2360-2400 MHz to complain about the results. Annexure A has explained a comparison of the interference mitigation scheme (Table 1).
78 Computational Intelligence for Data Analysis, Vol. 2
Ahmad et al.
Table 1. Comparison of various interference. Technique
Strategy
Mitigation Capacity & Collison Ratio
Processing Overhead & Energy Consumption
Advantages
Robust interference mitigation scheme (LRIM) [23].
Mitigation
Partial, Average
High, High
Provides reliability and increases the data rate.
High, Low
Energy Efficient.
Asynchronous inter-network Avoidance Partial, High interference Avoidance (AIIA) [24]. Fuzzy Power controller (FPC) [25].
Mitigation
Partial, Low
High, Average
coexisting WBAN receives the best signal using SINR
Power Allocation Algorithm Mitigation [26].
Partial, Low
High, High
Provide high throughput and QoS.
Cyber-Physical System [27]. Mitigation
Partial, Low
Low, High
High performance. According to the urgency level of the observed traffic.
Complete, Low
High, Average
Better energy lifetime of WBAN.
Mac Programming Scheme [28].
Avoidance
Dynamic Co-Existence Management (DCM) [29].
Avoidance Partial, High Average, Average
Energy Efficient.
Low Power Digital Baseband Avoidance Partial, High High, Low power, The transceiver circuit had a Transceiver Design [32]. simple design, high data rate, Low energy consumed per bit. Analysis of the CSMA/CA mechanism [33].
Mitigation
Partial, High
High, Low
In this research, the IEEE 802.15.6 model was based on user priorities.
A Secure System for Pervasive Social Network [34].
Avoidance
Partial, Low
High, Partial
In this research, the author had proposed securely communicating data in a network.
Enabling Interference-Aware Avoidance and Energy-Efficient [35].
Partial, Low
Partial, Low
In this research, the authors had increased the throughput, delay, packet delivery ratio, and Throughput.
Throughput and Delay [36]. Avoidance Partial, High
High, Partial
This research has been very helpful to the new researcher who works upon IEEE 802.15.6.
Wireless Body Area Networks
Computational Intelligence for Data Analysis, Vol. 2 79
CONCLUSION The Internet of Medical Things (IoMT) uses internet computer networks to connect medical equipment and apps to healthcare IT systems. In IoMT, coexistence can be fatal to delay or packet loss, indicating serious cases of WBAN applications, especially in health systems that make patient reporting potentially significant. In this white paper, we conduct a detailed analysis of the issues of a comprehensive review and coexistence and mitigation of the WBAN technologies. Three major wireless WBAN technologies, ZigBee, IEEE 802.15.6, and low-power Wi-Fi, are considered. We studied the WBAN mitigation method. Coexisting WBANs have been modified to interfere with the classification and accurate correlation between existing WBAN interference mitigation schemes. Although many interference mitigation plans have been reported, none of the advantageous systems is much superior to other systems. The dynamic channel WBASN requires a compromise between network bandwidth and power WBASN. When designing interference, you also need to consider channel dynamics, QoS, latency, reliability, and network lifetime. Since WBASN performance and QoS are very noise-sensitive, different algorithms to mitigate interference are to be developed in the future. Inter-block interference prevention design is one-way. QoS requirements have not been adequately investigated, and it is important to note performance, which is reduced using existing methods to reduce the noise of some biomedical applications. CONSENT FOR PUBLICATON Declared none. CONFLICT OF INTEREST The author declares no conflict of interest, financial or otherwise. ACKNOWLEDGEMENT Declared none. REFERENCES [1]
D.P. Tobon, T.H. Falk, and M. Maier, "Context awareness in WBANs: a survey on medical and nonmedical applications", IEEE Wirel. Commun., vol. 20, no. 4, pp. 30-37, 2013. [http://dx.doi.org/10.1109/MWC.2013.6590048]
[2]
S. González-Valenzuela, Min Chen, and V.C.M. Leung, "Mobility support for health monitoring at home using wearable sensors", IEEE Trans. Inf. Technol. Biomed., vol. 15, no. 4, pp. 539-549, 2011. [http://dx.doi.org/10.1109/TITB.2010.2104326] [PMID: 21216718]
[3]
T. Le, and S. Moh, "An Interference-Aware Traffic-Priority-Based Link Scheduling Algorithm for Interference Mitigation in Multiple Wireless Body Area Networks", Sensors (Basel), vol. 16, no. 12, p. 2190, 2016.
80 Computational Intelligence for Data Analysis, Vol. 2
Ahmad et al.
[http://dx.doi.org/10.3390/s16122190] [PMID: 27999385] [4]
D. Yang, Y. Xu, and M. Gidlund, "Wireless Coexistence between IEEE 802.11- and IEEE 802.15.4Based Networks: A Survey", Int. J. Distrib. Sens. Netw.,. vol. 7, no. 1, p. 912152, 2011. [http://dx.doi.org/10.1155/2011/912152]
[5]
S. Ullah, H. Higgins, B. Braem, B. Latre, C. Blondia, I. Moerman, S. Saleem, Z. Rahman, and K.S. Kwak, "A Comprehensive Survey of Wireless Body Area Networks", J. Med. Syst., vol. 36, no. 3, pp. 1065-1094, 2012. [http://dx.doi.org/10.1007/s10916-010-9571-3] [PMID: 20721685]
[6]
A. Kailas, and M.A. Ingram, "Wireless Aspects of Telehealth", Wirel. Pers. Commun., vol. 51, no. 4, pp. 673-686, 2009. [http://dx.doi.org/10.1007/s11277-009-9763-7]
[7]
M. Chen, and A.V. Vasilakos, "Ubiquitous Body Sensor Networks", Mob. Netw. Appl., vol. 16, no. 6, pp. 661-662, 2011. [http://dx.doi.org/10.1007/s11036-011-0341-3]
[8]
J.Y. Khan, M.R. Yuce, G. Bulger, and B. Harding, "Wireless Body Area Network (WBAN) design techniques and performance evaluation", J. Med. Syst., vol. 36, no. 3, pp. 1441-1457, 2012. [http://dx.doi.org/10.1007/s10916-010-9605-x] [PMID: 20953680]
[9]
X. Lai, Q. Liu, X. Wei, W. Wang, G. Zhou, and G. Han, "A survey of body sensor networks", Sensors (Basel), vol. 13, no. 5, pp. 5406-5447, 2013. [http://dx.doi.org/10.3390/s130505406] [PMID: 23615581]
[10]
K. Kwak, S. Ullah, N. Ullah, "An overview of IEEE 802.15.6 standard", 2010 3rd International Symposium on Applied Sciences in Biomedical and Communication Technologies (ISABEL 2010), 2010. [http://dx.doi.org/10.1109/ISABEL.2010.5702867]
[11]
Y. Fei, L. Zhong, and N.K. Jha, "An energy-aware framework for dynamic software management in mobile computing systems", ACM Trans. Embed. Comput. Syst., vol. 7, no. 3, pp. 1-31, 2008. [http://dx.doi.org/10.1145/1347375.1347380]
[12]
Z. Liu, B. Liu, and C.W. Chen, "Transmission-Rate-Adaption Assisted Energy-efficient Resource Allocation with QoS Support in WBANs", IEEE Sens. J.. pp. 1, 2017 [http://dx.doi.org/10.1109/JSEN.2017.2725906]
[13]
A. Hadjidj, M. Souil, A. Bouabdallah, Y. Challal, and H. Owen, "Wireless sensor networks for rehabilitation applications: Challenges and opportunities", J. Netw. Comput. Appl., vol. 36, no. 1, pp. 1-15, 2013. [http://dx.doi.org/10.1016/j.jnca.2012.10.002]
[14]
H. Zhou, and H. Hu, "Human motion tracking for rehabilitation—A survey", Biomed. Signal Process. Control, vol. 3, no. 1, pp. 1-18, 2008. [http://dx.doi.org/10.1016/j.bspc.2007.09.001]
[15]
A. Milenković, C. Otto, and E. Jovanov, "Wireless sensor networks for personal health monitoring: Issues and an implementation", Comput. Commun., vol. 29, no. 13-14, pp. 2521-2533, 2006. [http://dx.doi.org/10.1016/j.comcom.2006.02.011]
[16]
T. Gao, T. Massey, L. Selavo, D. Crawford, B. Chen, K. Lorincz, V. Shnayder, L. Hauenstein, F. Dabiri, J. Jeng, A. Chanmugam, D. White, M. Sarrafzadeh, and M. Welsh, "The advanced health and disaster aid network: a light-weight wireless medical system for triage", IEEE Trans. Biomed. Circuits Syst., vol. 1, no. 3, pp. 203-216, 2007. [http://dx.doi.org/10.1109/TBCAS.2007.910901] [PMID: 23852414]
[17]
S. Adibi, "Mobile Health", Springer Series in Bio-/Neuroinformatics, 2015. [http://dx.doi.org/10.1007/978-3-319-12817-7]
[18]
M.M. Alam, and E. Ben Hamida, "Strategies for Optimal MAC Parameters Tuning in IEEE 802.15.6
Wireless Body Area Networks
Computational Intelligence for Data Analysis, Vol. 2 81
Wearable Wireless Sensor Networks", J. Med. Syst., vol. 39, no. 9, p. 106, 2015. [http://dx.doi.org/10.1007/s10916-015-0277-4] [PMID: 26266628] [19]
A.C. Mühlbacher, A. Kaczynski, P. Zweifel, and F.R. Johnson, "Experimental measurement of preferences in health and healthcare using best-worst scaling: an overview", Health Econ. Rev., vol. 6, no. 1, p. 2, 2016. [http://dx.doi.org/10.1186/s13561-015-0079-x] [PMID: 26743636]
[20]
"IEEE Standard for Local and metropolitan area networks - Part 15.6: Wireless Body Area Networks," in IEEE Std. 802.15.6-2012 , vol., no., pp. 1-271, 2012. [http://dx.doi.org/10.1109/IEEESTD.2012.6161600]
[21]
M. Chen, J. Wan, S. Gonzalez, X. Liao, and V.C.M. Leung, "A Survey of Recent Developments in Home M2M Networks", IEEE Commun. Surv. Tutor., vol. 16, no. 1, pp. 98-114, 2014. [http://dx.doi.org/10.1109/SURV.2013.110113.00249]
[22]
L. Zou, B. Liu, C. Chen, and C.H. Chen, "Bayesian game based power control scheme for inter-wban interference mitigation", Global Communications Conference (GLOBECOM), pp. 240– 245, 2014. [http://dx.doi.org/10.1109/GLOCOM.2014.7036814]
[23]
L. Ma, Y. Ge, and Y. Zhu, "TinyZKP: A Lightweight Authentication Scheme Based on ZeroKnowledge Proof for Wireless Body Area Networks", Wirel. Pers. Commun., vol. 77, no. 2, pp. 10771090, 2014. [http://dx.doi.org/10.1007/s11277-013-1555-4]
[24]
E.J. Kim, S. Youm, T. Shon, and C.H. Kang, "Asynchronous inter-network interference avoidance for wireless body area networks", J. Supercomput., vol. 65, no. 2, pp. 562-579, 2013. [http://dx.doi.org/10.1007/s11227-012-0840-4]
[25]
T. Le, and S. Moh, "Interference Mitigation Schemes for Wireless Body Area Sensor Networks: A Comparative Survey", Sensors (Basel), vol. 15, no. 6, pp. 13805-13838, 2015. [http://dx.doi.org/10.3390/s150613805] [PMID: 26110407]
[26]
S. Rezvani, and S.A. Ghorashi, "Context aware and channel-based resource allocation for wireless body area networks", IET Wirel. Sens. Syst., vol. 3, no. 1, pp. 16-25, 2013. [http://dx.doi.org/10.1049/iet-wss.2012.0100]
[27]
X. Wang and L. Cai, "Interference Analysis of Co-Existing Wireless Body Area Networks", 2011 IEEE Global Telecommunications Conference - GLOBECOM 2011, 2011. [http://dx.doi.org/10.1109/GLOCOM.2011.6133624]
[28]
Z. Zhang, H. Wang, C. Wang, and H. Fang, "Interference Mitigation for Cyber-Physical Wireless Body Area Network System Using Social Networks", IEEE Trans. Emerg. Top. Comput., vol. 1, no. 1, pp. 121-132, 2013. [http://dx.doi.org/10.1109/TETC.2013.2274430] [PMID: 25436180]
[29]
"IEEE ICC 2014 explores next wave digital communications in wondrous Sydney, Australia [Conference Report]", IEEE Communications Magazine, vol. 52, no. 8, pp. 8-10, 2014. [http://dx.doi.org/10.1109/TETC.2013.2274430] [PMID: 25436180]
[30]
M.N. Deylami, and E. Jovanov, "A distributed scheme to manage the dynamic coexistence of IEEE 802.15.4-based health-monitoring WBANs", IEEE J. Biomed. Health Inform., vol. 18, no. 1, pp. 327334, 2014. [http://dx.doi.org/10.1109/JBHI.2013.2278217] [PMID: 24403431]
[31]
S.H. Cheng, and C.Y. Huang, "Coloring-Based Inter-WBAN Scheduling for Mobile Wireless Body Area Networks", IEEE Trans. Parallel Distrib. Syst., vol. 24, no. 2, pp. 250-259, 2013. [http://dx.doi.org/10.1109/TPDS.2012.133]
[32]
P.K. Manchi, R. Paily, and A.K. Gogoi, "Low-Power Digital Baseband Transceiver Design for UWB Physical Layer of IEEE 802.15.6 Standard", IEEE Trans. Industr. Inform., vol. 13, no. 5, pp. 24742483, 2017.
82 Computational Intelligence for Data Analysis, Vol. 2
Ahmad et al.
[http://dx.doi.org/10.1109/TII.2017.2717882] [33]
S. Rashwand, J. Misic, and V.B. Misic, "Analysis of CSMA/CA Mechanism of IEEE 802.15.6 under Non-Saturation Regime", IEEE Trans. Parallel Distrib. Syst., vol. 27, no. 5, pp. 1279-1288, 2016. [http://dx.doi.org/10.1109/TPDS.2015.2447528]
[34]
S. Movassaghi, D.B. Smith, M. Abolhasan, and A. Jamalipour, "Opportunistic Spectrum Allocation for Interference Mitigation Amongst Coexisting Wireless Body Area Networks", ACM Trans. Sens. Netw., vol. 14, no. 2, pp. 1-22, 2018. [http://dx.doi.org/10.1145/3139257]
[35]
M. Abul Hassan, S. Irfan Ullah, A. Salam, A. Wajid Ullah, M. Imad, and F. Ullah, "Energy efficient hierarchical based fish eye state routing protocol for flying Ad-hoc networks", Indonesian Journal of Electrical Engineering and Computer Science, vol. 21, no. 1, pp. 465-471, 2021. [http://dx.doi.org/10.11591/ijeecs.v21.i1.pp465-471]
[36]
M.A. Hassan, S. I Ullah, I.U. Khan, S.B.H. Shah, A. Salam and A.W. Ullah Khan” Unmanned Aerial Vehicles Routing Formation using fisheye state routing for flying ad-hoc networks”, In: The 4th International Conference on Future Networks and Distributed Systems (ICFNDS), pp. 1-7, 2020
Computational Intelligence for Data Analysis, 2023, Vol. 2, 83-106
83
CHAPTER 6
Artificial Intelligence-Based IoT Applications in Future Pandemics Tarun Virmani1,*, Anjali Sharma2, Ashwani Sharma3, Girish Kumar3 and Meenu Bhati3 School of Pharmaceutical Sciences, MVN University, Palwal, Haryana, 121105, India Pharmacovigilance Expert, Uttar Pradesh, India 3 School of Pharmaceutical Sciences, MVN University, Palwal, Haryana, 121105, India 1 2
Abstract: One of the greatest issues confronting the globe now is the pandemic disease calamity. Since December 2019, the world has been battling with COVID-19 pandemic. The COVID-19 crisis has made human life more difficult. Decision-making systems are urgently needed by healthcare institutions to deal with such pandemics and assist them with appropriate suggestions in real-time and prevent their spreading. To avoid and monitor a pandemic outbreak, healthcare delivery involves the use of new technologies, such as artificial intelligence (AI), the internet of things (IoT) and machine learning (ML). AI is reshaping the healthcare system to tackle the pandemic situation. AI is the science and engineering of creating intelligent machines to give them the ability to think, attain and exceed human intelligence. The advancement in the use of AI and IoT-based surveillance systems aids in detecting infected individuals and isolating them from non-infected individuals utilizing previous data. By assessing and interpreting data using AI technology, the IoT-based system employs parallel computing to minimize and prevent pandemic disease. In a pandemic crisis, the ability of ML or AI-based IoT systems in healthcare has provided its capacity to monitor and reduce the growth of the spread of pandemic disease. It has even been shown to reduce medical expenditures and enhance better therapy for infected individuals. This chapter majorly focuses on the applications of AI-based IoT systems in tracking pandemics. The ML-based IoT could be a game-changer in epidemic surveillance. With the proper implementation of proposed inventions, academicians, government officials and experts can create a better atmosphere to tackle the pandemic disease.
Keywords: Artificial Intelligence, COVID-19, Healthcare, Internet of things, Machine learning, Pandemic. Corresponding author Tarun Virmani: School of Pharmaceutical Sciences, MVN University, Palwal, Haryana, 121105, India; E-mail: [email protected] *
Mariya Ouaissa, Mariyam Ouaissa, Zakaria Boulouad, Inam Ullah Khan, Sailesh Iyer (Eds.) All rights reserved-© 2023 Bentham Science Publishers
84 Computational Intelligence for Data Analysis, Vol. 2
Virmani et al.
INTRODUCTION The viruses that originate from wild birds or animals are the main cause of pandemic diseases. The virus enters the human body through mutation. India has endured a wide range of epidemics and pandemics over the years. An epidemic is an uncommon occurrence of specific health-related problems, events, and diseases in a region or community. It involves a sudden, severe outbreak of a disease that pre-existed in the community. The term pandemic is used to describe the rapid spread of diseases throughout the world. Thus, there is a need to confront and eradicate the problems of global health along with adequate measures to prevent further transmission. The history has been replete with reports about cholera, smallpox, dengue, influenza, polio, plague, and several other diseases. Throughout history, many pandemics have occurred. Their control has been difficult in many instances because a functioning global surveillance system has not been in place. In India, Cholera had been a major issue throughout the 19th century, increasing death tolls each year. However, the influenza pandemic came later in the early 20th century [1, 2]. It has been reported that the primary cause of the sudden and rapid outbreaks of the disease is a lack of sanitation, proper public health, and malnutrition [3 - 5]. To manage the pandemic in India, the government and health organizations had been taken effective measures which are critical in controlling the pandemic [6, 7]. COVID-19, a recent outbreak that began in Wuhan City in December 2019, is one example of a pandemic. Humanity is now living in a limited area due to the pandemic situation. Pandemics have disrupted the normal life of humans. This situation has also influenced social, business, and regional activities and thus forced them to live within certain limits. COVID-19 was declared a worldwide pandemic based on its severity on January 30, 2020 [8, 9]. Recent updates on COVID-19 cases were approximately 452 million confirmed cases, with 6.02 million deaths reported [10]. The COVID-19 virus can be transmitted to others from infected individuals in numerous ways and is considered hazardous [11]. Therefore, a complete lockdown has been imposed throughout the country to prevent the virus from spreading further. Several countries have curbed their outbreaks by enforcing a strict lockdown. Still, despite a complete lockdown, the disease is not completely eradicated. Numerous nations have worked jointly on developing medicine to treat COVID-19 [12, 13]. However, as of now, there has been no known treatment to completely eradicate this disease. Nevertheless, few medicines are being investigated as prospective therapies. Moreover, the WHO has recommended clinical trials for the proposed medicine [14, 15]. A timely and accurate diagnosis of COVID-19 patients is essential to their medical treatment and thus helps in preventing the spread of outbreaks.
Future Pandemics
Computational Intelligence for Data Analysis, Vol. 2 85
During the worldwide crisis, the medical industry is seeking new technologies that can monitor and control this pandemic of the COVID-19 virus. One such technology is artificial intelligence (AI), which identifies patients at risk, offers them medical help, and controls the spread of infection in real-time. Moreover, helps in easily tracking the transmission of the virus. To track cases, locate disease clusters, diagnose COVID-19, keep records, and predict future outbreaks, an AI system has been effectively used [16]. Machine learning (ML) and deep learning (DL) are the two types of AI subfields. Both ML and DL are part of AI. ML is a branch of computer vision that uses algorithms to assess, make decisions, and learn from raw data. ML is also applicable to support medical discovery and clinical decisions [17]. They have limited capabilities when it comes to processing raw data. DL, a recent and fastest-growing subfield of machine learning, has inspired great global interest in the previous few years. Unlike machine learning, DL does not necessitate the creation of handmade features or the manual extraction of data. It's a more advanced form of machine learning that allows computers to extract, understand, and analyze meaningful facts or information from raw data, then process it and make decisions based on it. Different representations of learning mechanisms are attempted by DL. DL assists in the abstraction of process input data from large-scale data by applying a multi-layered deep neural network (DNN) [18, 19]. Radio imaging technology, such as X-ray, clinical blood test data, and computed tomography (CT), are utilized to screen patients utilizing ML and AI. Radiological images such as CT filters and X-rays can be used by doctors to supplement traditional diagnosis and screening methods [20]. Internet of things (IoT) is the most up-to-date, popular, and advanced technology that can automate remote control, intelligent management, data monitoring, and management via a real-time network. As a result, using AIsupported IoT technology is of great importance to clinical medicine and particularly in the context to prevent and control the spread of future pandemics. Fig. (1) represents the benefits of AI and IoT in the field of the healthcare field. The IoT powered by AI allows the emulation of intelligent behavior that aids in decision-making with a minimum amount of human interventions. IoT systems are used to collect data remotely from COVID-19 patients. Healthcare workers are given this information to diagnose COVID-19. In addition, using the previous patient’s data, the software can forecast the mortality risk. AI is about devices learning from their experience and stored data, whereas IoT involves devices communicating using the internet [21 - 23]. Any device that is capable of connecting to the internet for transferring and monitoring data is an IoT device. IoT has gained traction in industry and academic disciplines, particularly in healthcare, in recent years. Modern healthcare systems are being transformed by the IoT revolution, including economic, societal, and technological implications. It is transforming the healthcare system from traditional to become more
86 Computational Intelligence for Data Analysis, Vol. 2
Virmani et al.
individualized, making it easier to diagnose, treat and monitor patients easily. IoT is gradually becoming a critical technology in healthcare, where it is used in improving service quality, may help reduce costs, and provide sophisticated user experiences [24 - 27]. IoT can also be used to track infected patients in epidemic situations. This technology is designed in such a way that it may be enabled when a pandemic happens and provides tactics for tacking pandemic outbreaks as well as the ability to provide computerized data. A large initiative for early detection of any illness, whether irresistible or non-irresistible is crucially important for early treatment to save more lives. Rapid screening and detection methods are effective in preventing the outspread of pandemic illnesses such as COVID-19 [28, 29]. Control of COVID-19 by the AI-assisted IoT system is vital for both patients and the public. For instance, wearable devices allow people to keep an eye on and record their heart rate, body temperature, respiration rate, and other physiological parameters. The person can immediately detect abnormalities in their vital signs even if they are in an isolated state also [30]. This chapter majorly focuses on the applications of AI, IoT, and ML-based systems that are being used in the healthcare field, combating the epidemic, and will also be used for predicting and managing future pandemic diseases.
Fig. (1). Benefits of AI and IoT in the healthcare field.
Future Pandemics
Computational Intelligence for Data Analysis, Vol. 2 87
IOT AND AI IN HEALTH CARE IoT refers to the interconnection of numerous sensors, devices, and other technologies. A dynamic analysis model calls for IoT technologies to be used to treat various illnesses in several medical circumstances. By incorporating IoT into healthcare systems, health practitioners can gain valuable data that provides insight into the symptoms and course of infection while also allowing for longterm consideration [12, 13, 15]. It amplifies the rapid expansion of a vast amount of data and a wide range of different data types. As, a result, IoT-based systems generate a lot of data, which is referred to as big data. The vast amount of data produced by IoT devices is too much for humans and computer software to comprehend and process effectively. To control them, artificial intelligence and machine learning techniques are used [31]. AI is the most important factor in the imitation of human tasks. It is capable of performing activities that were previously only performed by human intelligence. AI can learn aspects from a huge quantity of data using complex algorithms, and then use the results to aid clinical practice. It could also have self-correcting capabilities to enhance accuracy depending on the input. Physicians can benefit from AI systems that provide up-to-date medical information from clinical practices and help them provide effective patient care. Moreover, an AI system can aid in the reduction of diagnostic and treatment errors, which are unavoidable in clinical practice. Furthermore, an AI system collects usable data from a huge patient population to aid in developing real-time conclusions for health risk alerts and prediction [32 34]. When a pandemic occurs, the crucial step is the identification of symptoms, proper monitoring, and implementation of the treatment. When it comes to merging AI and IoT in healthcare, there's a good possibility they'll boost operational efficiency. The important steps in the smart and efficient deployment of AI algorithms in IoT devices are tracking (gathering), monitoring (analyzing), automation (modeling, predicting), control, and optimization (training) [35]. AI combined with IoT is more potent in terms of intelligent decision-making and can aid in the fight against the pandemic. IOT AND AI: APPLICATIONS IoT is more of a concept that builds the entire architectural framework that permits the integration and effective interchange of information between the people in need and the service providers. In the current usual situation, the majority of issues arise as a result of ineffective patient reachability, which is also the most important issue after vaccine development [36, 37]. The implementation of the IoT idea improves patient accessibility, allowing them to receive important care and eventually recover from their illness [38]. In recent days, the IoT has become more well-known for its use in healthcare monitoring and surveillance.
88 Computational Intelligence for Data Analysis, Vol. 2
Virmani et al.
During an epidemic, the IoT can potentially be used to track infected patients. IoT technology was created in such a way that it can be used in the event of a pandemic, and it has pandemic-fighting strategies as well as computerized and transparent treatment during the epidemic [23, 38]. IoT and AI have become key tools in monitoring and analyzing instances since the COVID-19 occurrence has changed our day-to-day lives. As a result of these advancements, the requirement to combat infection has been balanced against the competing need to provide individual protection [39]. While there is an urgent need to contribute large amounts of energy to combat the pandemic, it is critical to recognize that such devices should be strictly limited in their use, both in terms of reason and time, and that individual rights to security, non-segregation, the protection of editorial sources, and other opportunities be fully protected [40 - 42]. Disease transmission is a major concern for patients in medical clinics. Cleanliness monitoring devices enabled by the IoT help keep patients safe. IoT devices also assist resource managers in areas such as pharmacy store stock control and nature observation, such as analyzing cooler temperatures and temperature and moisture control [43 45]. In recent decades and for future perception, AI has been the most emerging and demanding scientifically engineered technique. Through this, the computational understanding of machines can be obtained by incorporating intelligent behavior to innovate intelligent and smart machines. AI consists of various pieces of techniques, tools, and algorithms such as neural networks, symbolic AI, DL, ML, and genetic algorithms. These tools are growing and showing impact in different fields like military, space, robotics, and health [46 48]. In recent years, the advanced featuring tool of AI is ML reorganized in different fields of engineering and science. Nowadays, it is largely adopted in our daily lives, but the ability to find out the conceptual abstract from the large volume of data and feature learning is the most powerful contribution of ML as a tool of AI [18, 49, 50]. The use of AI in imaging processes has attracted substantial attention in the healthcare industry. Fig. (2) illustrates the use of IoT and AI in the various pharmaceutical industry. IoT and AI have been pushed hard to provide efficient and rapid healthcare services, particularly in the context of COVID-19, to automate and facilitate a variety of activities for healthcare workers.
Future Pandemics
Computational Intelligence for Data Analysis, Vol. 2 89
Fig. (2). Illustrates the use of IoT and AI in the various pharmaceutical industry.
AI AND IOT-ENABLED REMOTE SCREENING In the case of the COVID-19 outbreak, screening infected people as soon as feasible is a critical step in preventing and controlling the spread of the virus. Manual screening is indeed slow, however remote screening technologies can improve screening speed and efficiency. As a result, AI and IoT-based remote screening must be established. Schinkothe et al. 2020 [51] suggested a free AIoT-based caregiver cockpit (C19CC). One user is first linked to a healthcare worker, who immediately categorizes the individual based on the description provided. The harshness of the examinees is shown by color codes. COVID-19 patients, for example, are identified by red codes, and examinees with red codes are automatically screened out, with their biographical information and recent range of activity highlighted. C19CC can be utilized not only for remote screening but also for remote monitoring, hospital wards, and other applications. C19CC conducts a contactfree pre-screening of patients and promptly identifies those who require immediate treatment. Lung ultrasound imaging classifiers are utilized for screening or diagnosis of COVID-19 in addition to traditional remote screening approaches. Tan and Liu, 2020 [52] offer a rudimentary facial recognition algorithm. The major purpose of this technology is to use thermal imaging to remotely screen out suspect patients and then use a facial recognition system to recover those who have had close contact with them, to isolate people in time, and
90 Computational Intelligence for Data Analysis, Vol. 2
Virmani et al.
prevent the virus from spreading [53]. The gathered lung ultrasonic images will be input into the platform via a portable scanner proposed by [54]. After the platform has processed the data, it will classify it in the subspace network. This sequence of procedures allows patients to be screened for COVID-19. The classifier can be employed on a big scale in nursing homes with significant funds, fully eliminating the risk of infection among the elderly who rarely visit the hospital. As a COVID19 patient frequently exhibits the symptom of fever, facial recognition technology will be utilized to detect the patient and notify the hospital platform via the Internet or mobile devices, allowing the patient to be separated and further diagnosed. The filter application process is intended for use in smartphone apps and web pages. It can address the need for multiple people to be screened at the same time. The data-gathering module may be used by a huge number of people at the same time, and the machine-learning module can answer queries about the risk of infection. A user notification module is used to send specific messages to the user on-demand to alert them of the result. The notification module will notify persons with COVID-19 indications of morbid progression. However, how to prevent platform abuse in the event of hostile assaults and theft of user information remains an essential challenge to be tackled [30]. Some of the remote screening methods are summarized in Table 1. Table 1. Remote screening methods and outcomes. Approach
Characteristic
Result
References
Soft wearable devices
Used in COVID-19 rapid screening.
Lowering the risk of pneumonia among healthcare personnel.
[55, 56]
The thermal imaging-based facial recognition system
Capability to track patients.
Aid in the prevention of COVIDtransmission.
[52]
Improving energy efficacy.
[54]
Subspace Obtaining test data accuracy multilayer+Portable scanner of over 96%. C19CC (Cockpit)
Identifying telemedicine options that can improve patient care immediately.
Improving COVID-19 patient care and safety.
[51]
One-shot learning framework
Effectively diagnosing coronavirus from tonsillitis for mass screening.
Used to identify patients who are at risk of becoming infected at an early stage.
[57]
In addition, wearable device technologies can be used for remote screening. Lonini et al. 2021 [55] proposed a wearable device that captures data by monitoring the tiny vibrations produced by heartbeat and respiration. Wearable based on AI and IoT systems is being used to assess COVID-19-related indications such as body temperature and respiratory rate as the novel coronavirus causes stress around the globe. Mohammed et al. 2019 [58] suggested the usage
Future Pandemics
Computational Intelligence for Data Analysis, Vol. 2 91
of a smart helmet with a thermal imaging system to identify coronavirus automatically using the thermal imaging process, reducing human-to-human interaction. Smart helmets have a positioning system. When the system detects temperatures that are greater than normal, it will immediately react. Wearable gadgets track a patient's whereabouts using Global Positioning System (GPS) data, allowing doctors to closely monitor their status. Fig. (3) is the architecture of the platform.
Fig. (3). Architecture of the platform.
Finding infected people in a crowd is critical for early COVID-19 diagnosis and control. Another typical technique to speed up the process of discovering contaminated people and zones during the pandemic is to use unmanned aerial vehicles (UAV) and, in particular, IoT-based drones. Drone technology can eliminate human contact while also reaching regions that are difficult to reach. During the COVID-19 pandemic, smartphone applications enabled by IoT that utilized information such as the Geographic Information System (GIS) and GPS for tracking purposes were widely employed to boost the chances of discovering infected people. Patients will benefit from smartphone applications that use the Internet of Medical Things (IoMT) to provide suitable therapies while at home. It also makes it easier for healthcare staff and authorities to keep track of patients and disease outbreaks [59, 60]. Healthcare staff can acquire real-time data on people's health in this way. The different types of tools and technologies used during COVID-19 with their capability have been briefed in Table 2.
92 Computational Intelligence for Data Analysis, Vol. 2
Virmani et al.
Table 2. IoT-linked devices used in COVID-19. Tools and Technology
Characterization
Types
Uses
References
Wearables
An app-enabled gadget that is worn on or adheres to the body to receive and process information.
Smart helmets, Smart thermometers
Robots
A programmed machine that can perform complex tasks in the same way that a living being does.
Telerobots Social robot
Applications (Smartphone)
A smartphone app built to perform specific activities.
Aarogya Setu Social Media Whatsapp
Used for monitoring, [38, 65 - 67] tracking improving the connection between people and healthcare services Provide medical assistance without having to go to the hospital.
Drones
An aircraft that is equipped with cameras, GPS, sensors, and communication systems and that is flown with little or no human intervention.
Medical drone, Surveillance drone, multipurpose drone
Execute a range of activities, including finding, monitoring, and delivering packages Less human interactions Preventing the spread of disease among healthcare workers.
Monitoring regularly [58, 61, 62] Improving the overall quality of patient care Reducing the number of hospital visits More efficient and safer hospital Used to monitor temperature. Reduce the number of people suffering from mental illnesses. By using remote diagnosis and treatment, we can reduce the number of interactions.
[63, 64]
[68, 69]
The following are the applications of IoT with patients, doctors, and hospitals:
Future Pandemics
Computational Intelligence for Data Analysis, Vol. 2 93
Patients and IoT Patients have access to tailored consideration through wearables such as wellness groups and other remotely related gadgets such as glucometers, oximeters, pulse monitoring sleeves, and so on. These devices can be programmed to remind you of carbohydrate content, practice checks, configurations, pulse types, and a lot more [70]. People, particularly elderly patients, have been altered by the Internet of Things, which has enabled the continuous monitoring of medical issues. Individuals who live alone, as well as their families, are impacted the most. A ready system sends signals to families and concerned health providers in the event of any unpleasant influence or changes in an individual's normal activities. IoT for Doctors By integrating wearables and other home monitoring devices with IoT, doctors can better monitor their patients' health. They can monitor patients' adherence to treatment regimens or any other criterion for clinical consideration. IoT enables medical professionals to be more observant and proactive in their relationships with patients. Information obtained from IoT devices can help doctors decide the best treatment option for their patients and reach the expected outcomes. IoT in Hospitals In addition to monitoring patients' well-being, IoT devices are useful in a variety of areas in clinics. Sensor-enabled IoT devices are used to track the location of clinical equipment such as wheelchairs, defibrillators, nebulizers, oxygen syphons, and other monitoring devices. Clinical staffing arrangements in various sectors can also be analyzed in real-time [20, 28, 71, 72]. Diagnosis CT scans and X-rays are the two most used imaging modalities for diagnosing in the context of COVID-19. However, in the setting of a novel coronavirus outbreak, using CT scans and X-rays may put patients and doctors at risk of cross-infection. However, reading a large number of imaging scans and manually tracing lung lesions will delay COVID-19 detection [73]. Hence, it is critical to establish an IoT-based intelligent diagnostic system to help frontline clinicians tackle COVID-19. In recent years, DL's many methodologies have proven to be highly successful for a wide range of big data analytics. CNN is widely used in a variety of applications, including speech recognition, computer vision, natural language processing, picture identification, and text detection and recognition [74]. The basic design of a CNN is shown in Fig. (4). The three types of CNN layers that are typically utilized are convolutional layers, fully connected layers,
94 Computational Intelligence for Data Analysis, Vol. 2
Virmani et al.
and pooling layers [75]. CNN variations and CNN itself can capture the overall structural appearance of feature-rich images. Medical image research for electrocardiogram (ECG) monitoring, image diagnosis, and picture segmentation have advanced significantly [76]. A 1D illustration of CNN architecture is shown in Fig. (4). CNN can deduce the whole image's context by combining the same number of down-sampling and up-sampling layers. ResNet, which is utilized for medical images, is another well-known CNN design. It aids in the acceleration of optimization convergence as well as the flow of gradient backpropagation [77, 78].
Fig. (4). (a) The basic design of a CNN. (b) A 1D illustration of CNN architecture.
With a small sample data set, Jiang's team [79] trained a VGG-16 convolutional neural network migration which is used to distinguish between the early, late, and severe stages of COVID-19 in patients and used to build an intelligent COVID-19 diagnostic model [30]. As shown in Fig. (5), Gomes et al. 2020 [80] offer an intelligent system to facilitate the diagnosis of X-ray scan images and build IKONOS (a desktop programme) using X-ray images to diagnose COVID-19. The doctor uploads the image to the app, which extracts features using classical classifiers, texture, and shape descriptors, and further analyses them using the intelligent system to identify COVID-19. The above-mentioned models are examples of AI and IoT-based systems that help in providing reference information to healthcare workers and increase efficiency throughout the pandemic prevention. Hence, AI-based IoT systems are expected to become increasingly popular in intelligent medical care in the future.
Future Pandemics
Computational Intelligence for Data Analysis, Vol. 2 95
Fig. (5). IKONOS may be used to load chest X-rays taken from symptomatic patients. The application is comprised of an intelligent system capable of extracting features and categorizing images. The findings can be viewed on the computer on which the software has been installed.
MONITORING AND CONTROL OF EPIDEMIC VIA ML-BASED IOT IoT based on machine learning is an application-specific, low-force, feasible, and simple-to-use solution to any persistent issues. Sensors are data collectors from the real world that are routed via an organization, while actuators allow things to act or respond in response to the data collected by sensors. Fig. (6) depicts the proposed IoT design for preventing pandemic illness spread. The information is exchanged via a passage device, which is then sent to the cloud passage. Separation of information, i.e., complete information, is extricated in the vast information distribution center. As it contained ordered information, it was a significant information distribution center. AI-powered by machine learning is used to create framework models based on requirements and acquired data. The evaluation of data can be used to depict findings and to establish correlations between executions. In open lavatories, IR sensors can be used to program the action of entrances and use of the water flow appropriately. By utilizing optical cameras at the passing purposes of doors of airports, train stations, transportation hubs, and retail centers, infrared thermometers may be used to verify the internal heat level to identify the contaminated among groups and face recognition. Essentially, sensors, as proposed in engineering, can be used to monitor the inside temperature, planned entryway activities, water gracefully controlled at public areas and latrines, and online meeting to keep a strategic distance from direct touch with the real world
96 Computational Intelligence for Data Analysis, Vol. 2
Virmani et al.
and people connect. Simulated intelligence-based deep learning can aid in the comprehension of health care patterns, the modeling of danger associations, and the prediction of outcomes. For small applications or individuals, a single temperature sensor, a single Node-MCU, or an Arduino board equipped with sensors and the Internet can be used [81]. Therefore, applying an ML-based approach, can reduce the spread of the pandemic and help in combating it.
Fig. (6). Proposed IoT design for preventing pandemic illness spread.
Drug Discovery and Vaccine Research Since the outbreak of the COVID-19 pandemic, it's been critical to find medications that can be used to treat the pandemic. ML can aid in the identification of existing medications that may be beneficial in the treatment of COVID-19. ML can learn from the structures of drugs and proteins and anticipate their interactions, allowing clinical trials to be conducted. Scientists and researchers have utilized a variety of methods, including repurposing existing drugs (therapeutic) and developing new ones to get the correct and effective medications [82]. The use of machine learning and the development of new models have prompted researchers to concentrate on the use of the machine and deep learning models to find medications that could cure COVID-19. It could take a long time to develop a new vaccination based on the available clinical data. However, using machine learning techniques, the total procedure can be greatly shortened while maintaining vaccination quality. For example, Ekins et al. 2015
Future Pandemics
Computational Intelligence for Data Analysis, Vol. 2 97
[83] describe how the Bayesian machine learning model was used in a study to produce an Ebola vaccine. While working on the H7N9 virus, Zhang et al. 2017 [84] used the random forest technique to improve the accuracy of the scores. Currently, the scientific community is putting a lot of work into using machine learning to find a design for the COVID-19 vaccine. Gonzalez-Dias et al. 2020 [85] described the stages of predicting vaccination immunogenicity and reactogenicity signatures using machine learning. The stages include data preparation, vaccine selection, and relevant gene selection, as well as selecting the appropriate machine learning algorithm for modeling and evaluating the predictive model's performance. Thus, AI aids in the development of vaccines and drug therapies at a far faster rate as well as clinical trials during vaccine development. Applicability of AI-Enabled System The smart city, as an interconnected urban culture, collects data from multiple embedded devices every second, implying that smart cities can work well with machine learning methodologies during the COVID-19 epidemic. For improved learning and predictive models, machine learning techniques rely on data, which can provide some intrinsic and essential insights to enable smart city decisionmakers to take preventive steps during the COVID-19 pandemic. Different machine learning algorithms work in conjunction with other domains of artificial intelligence (AI), allowing the model to create a robust self-learning platform. As data availability is restricted and we must deal with real-time data streaming, it is critical to examine the role of AI and machine learning in combating COVID-19. As a result, the importance of self-learning systems increases in smart cities. The entire flow of how AI and ML technologies can be used to combat the COVID-19 epidemic in smart cities is depicted in Fig. (7). Some of the several forms of data are created by the information and communication technology equipment incorporated in smart cities, as shown in Fig. (7) are as follows: 1. Statistical data, which typically includes the cumulative daily number of detected cases, new positive cases, deaths, and recovered cases, among other things, could be used to predict future instances and plan for emergencies. 2. The epidemiological data primarily relates to all clinical patient test data, including data about tests on various medications, various drug trials, patients' medical histories, and patient responses to various medications, among other things. 3. Real-time surveillance data provided by smart city sensors and cameras could
98 Computational Intelligence for Data Analysis, Vol. 2
Virmani et al.
be useful in tracking and preventing the spread of COVID-19. One of COVID19's initial identifiers, for example, is based on fever symptoms; hence, body temperature can be monitored using facial recognition and other personal information [82]. Thus, ML approaches are used to process and analyze the data to extract insights that can be applied in a variety of applications.
Fig. (7). Applications of AI and ML-based technologies used to fight COVID-19.
FUTURE PANDEMIC PREDICTION Despite being the greatest epidemic in modern memory, this pandemic occurred during the digital age. As a result, every component of the analysis can now be represented in terms of data, including at the macro level, logistically, and physiologically; this will undoubtedly be useful in forecasting the behavior of this current pandemic or unknown future pandemics. It has been documented that AI techniques have the potential to be applied to COVID-19 genome sequences to retrieve valuable info about the organism. When it comes to pattern analysis, effective computer-based tools are available that allow people, especially bioinformaticians, to examine difficult genetic and genomic data sets that are large and complex. Algorithms have been devised for the analysis of mutations in genome sequences, which are used to locate sites in genome sequences where the nucleotide bases change and quantify the mutation rate. Thus, Sequential pattern mining (SPM) data can be used to reveal new about virus pathogenicity, clinical
Future Pandemics
Computational Intelligence for Data Analysis, Vol. 2 99
symptomology, mutations, and strains information [16, 86]. Eng et al. 2014 [87] used a machine-learning approach (random forest) to identify probable zoonotic influenza strains, which are viruses that normally only infect animals but could also be dangerous to humans. As a result, machine learning could aid in the prediction of future pandemics caused by any species. The sole drawback is that the data could come from a different zone, such as the source of COVID-19, which could come from “bats” hence, pandemic sources could be different from those seen previously (different genome structure). Furthermore, classical machine learning requires that the data distribution in training and testing be from the same domains. Transfer learning (TL), a kind of machine learning, may, on the other hand, efficiently manage scenarios in which the training and testing data come from different data distributions. That is, knowledge gained from previous pandemics could be applied to new domains in the future, even with lower amounts of data. Fig. (8) depicts a scenario in which the pre-trained model from the present COVID-19 pandemic (with big data and labels) may be used, with substantially fewer data and labels, to anticipate future pandemics, prepare the smart city for such an event, and swiftly assist in the disease's spread. It represents the components of the smart city-based framework for countering COVID-19, which has a lot of potential use.
Fig. (8). Transfer learning for predicting future pandemics.
CONCLUSION One of the most serious challenges faced by the globe today is a threat of a pandemic illness. Therefore, to monitor and prevent the spread of pandemic
100 Computational Intelligence for Data Analysis, Vol. 2
Virmani et al.
disease, advanced intelligent monitoring technology is now required. Particularly during the COVID-19 epidemic, the entire healthcare system was confronted with new issues that needed to be addressed. However, intelligent technology, i.e., AIbased IoT systems, has proven to be a highly important resource with many applications ranging from remote screening, diagnosis, treatment and vaccine development, reducing the risk of contamination. These technologies are also beneficial in maintaining quality supervision through the use of real-time information. As intelligent technologies and devices have shown their potential during major outbreaks, i.e., COVID-19, and are also being discussed above. Considering the next pandemic is something that no one wants to think about. However, we must be prepared, and early detection and intervention are critical steps in the prevention of a disease outbreak. Hence, implementing advanced technologies can help to forecast the occurrence of the disease by employing a statistical-based strategy. The designed ML-based IoT module has the potential to be a game-changer in epidemic monitoring, and applying various methods could be fruitful for getting information regarding different strains and mutations. Researchers, doctors, government officials, and academicians can all benefit from the effective deployment of this technology, which will help to establish a more favorable atmosphere for tackling further pandemic diseases. In the future, more AI-based IoT systems will be developed to combat pandemics. By using technology and integrated algorithms, it will be quite beneficial to screen, monitor, diagnose, and predict the recurrence or occurrence of the pandemic so that prompt actions will be taken. Thus, the use of advanced technology will serve as improved medical treatments in the future. CONSENT FOR PUBLICATON Declared none. CONFLICT OF INTEREST The author declares no conflict of interest, financial or otherwise. ACKNOWLEDGEMENT Declared none. REFERENCES [1]
J.T. Watson, M. Gayer, and M.A. Connolly, "Epidemics after natural disasters", Emerg. Infect. Dis., vol. 13, no. 1, pp. 1-5, 2007. [http://dx.doi.org/10.3201/eid1301.060779] [PMID: 17370508]
[2]
J.M. Hughes, M.E. Wilson, B.L. Pike, K.E. Sailors, J.N. Fair, and M. LeBreton, “The. origin and prevention of pandemics", Clin. Infect. Dis., vol. 50, no. 12, pp. 1636–40, 2010”
Future Pandemics
Computational Intelligence for Data Analysis, Vol. 2 101
[3]
A.L. Rice, L. Sacco, A. Hyder, and R.E. Black, "Malnutrition as an underlying cause of childhood deaths associated with infectious diseases in developing countries", Bull. World Health Organ., vol. 78, no. 10, pp. 1207-1221, 2000. [PMID: 11100616]
[4]
T.J. John, L. Dandona, V.P. Sharma, and M. Kakkar, "Continuing challenge of infectious diseases in India", Lancet, vol. 377, no. 9761, pp. 252-269, 2011. [http://dx.doi.org/10.1016/S0140-6736(10)61265-2] [PMID: 21227500]
[5]
S. Srabani, “Indian cholera: a myth Indian Journal of History of Science”, Retrieved April, vol. 20, p. 2018, 2012.
[6]
"Epidemics and Pandemics in India throughout History: A Review Article", Indian J. Public Health Res. Dev., vol. 10, no. 8, 2019.
[7]
T.K. Burki, "Coronavirus in China", Lancet Respir. Med., vol. 8, no. 3, p. 238, 2020. [http://dx.doi.org/10.1016/S2213-2600(20)30056-4] [PMID: 32027848]
[8]
W.B. Yu, G.D. Tang, L. Zhang, and R.T. Corlett, "Decoding the evolution and transmissions of the novel pneumonia coronavirus (SARS-CoV-2 / HCoV-19) using whole genomic data", Zool. Res., vol. 41, no. 3, pp. 247-257, 2020. [http://dx.doi.org/10.24272/j.issn.2095-8137.2020.022] [PMID: 32351056]
[9]
D. Arnold, "Pandemic India: coronavirus and the uses of history", J. Asian Stud., vol. 79, no. 3, pp. 569-577, 2020. [http://dx.doi.org/10.1017/S0021911820002272] [PMID: 34191871]
[10]
P. Hamet, and J. Tremblay, "Artificial intelligence in medicine", Metabolism, vol. 69, pp. S36-S40, 2017. [http://dx.doi.org/10.1016/j.metabol.2017.01.011] [PMID: 28126242]
[11]
J.F.W. Chan, S. Yuan, K.H. Kok, K.K.W. To, H. Chu, J. Yang, F. Xing, J. Liu, C.C.Y. Yip, R.W.S. Poon, H.W. Tsoi, S.K.F. Lo, K.H. Chan, V.K.M. Poon, W.M. Chan, J.D. Ip, J.P. Cai, V.C.C. Cheng, H. Chen, C.K.M. Hui, and K.Y. Yuen, "A familial cluster of pneumonia associated with the 2019 novel coronavirus indicating person-to-person transmission: a study of a family cluster", Lancet, vol. 395, no. 10223, pp. 514-523, 2020. [http://dx.doi.org/10.1016/S0140-6736(20)30154-9] [PMID: 31986261]
[12]
W. Liu, Z.W. Tao, L. Wang, M.L. Yuan, K. Liu, L. Zhou, S. Wei, Y. Deng, J. Liu, H.G. Liu, M. Yang, and Y. Hu, "Analysis of factors associated with disease outcomes in hospitalized patients with 2019 novel coronavirus disease", Chin. Med. J. (Engl.), vol. 133, no. 9, pp. 1032-1038, 2020. [http://dx.doi.org/10.1097/CM9.0000000000000775] [PMID: 32118640]
[13]
M. Goumenou, D. Sarigiannis, A. Tsatsakis, O. Anesti, A. Docea, D. Petrakis, D. Tsoukalas, R. Kostoff, V. Rakitskii, D. Spandidos, M. Aschner, and D. Calina, "COVID-19 in Northern Italy: An integrative overview of factors possibly influencing the sharp increase of the outbreak (Review)", Mol. Med. Rep., vol. 22, no. 1, pp. 20-32, 2020. [http://dx.doi.org/10.3892/mmr.2020.11079] [PMID: 32319647]
[14]
A. Mitra, B. Deutsch, F. Ignatovich, C. Dykes, and L. Novotny, "Nano-optofluidic detection of single viruses and nanoparticles", ACS Nano, vol. 4, no. 3, pp. 1305-1312, 2010. [http://dx.doi.org/10.1021/nn901889v] [PMID: 20148575]
[15]
P. Mehta, D.F. McAuley, M. Brown, E. Sanchez, R.S. Tattersall, and J.J. Manson, "COVID-19: consider cytokine storm syndromes and immunosuppression", Lancet, vol. 395, no. 10229, pp. 10331034, 2020. [http://dx.doi.org/10.1016/S0140-6736(20)30628-0] [PMID: 32192578]
[16]
N. Arora, A. K. Banerjee, and M. L. Narasu, “The role of artificial intelligence in tackling COVID19”, Future Virology, vol. 15, no. 11. Future Medicine, pp. 717–724, 2020. [http://dx.doi.org/10.2217/fvl-2020-0130]
102 Computational Intelligence for Data Analysis, Vol. 2
Virmani et al.
[17]
I. Kononenko, "Inductive and Bayesian learning in medical diagnosis", Appl. Artif. Intell., vol. 7, no. 4, pp. 317-337, 1993. [http://dx.doi.org/10.1080/08839519308949993]
[18]
Y. LeCun, Y. Bengio, and G. Hinton, “Deep learning". Nature, vol. 521, no. 7553, pp. 436-444, 2015.
[19]
D. Yu, and L. Deng, "Deep learning and its applications to signal and information processing [exploratory dsp]", IEEE Signal Process. Mag., vol. 28, no. 1, pp. 145-154, 2011. [http://dx.doi.org/10.1109/MSP.2010.939038]
[20]
A.A. Ardakani, A.R. Kanafi, U.R. Acharya, N. Khadem, and A. Mohammadi, "Application of deep learning technique to manage COVID-19 in routine clinical practice using CT images: Results of 10 convolutional neural networks", Comput. Biol. Med., vol. 121, p. 103795, 2020. [http://dx.doi.org/10.1016/j.compbiomed.2020.103795] [PMID: 32568676]
[21]
Q. Du, H. Song, and X. Zhu, "Social-feature enabled communications among devices toward the smart IoT community", IEEE Commun. Mag., vol. 57, no. 1, pp. 130-137, 2019. [http://dx.doi.org/10.1109/MCOM.2018.1700563]
[22]
P. Partila, J. Tovarek, G.H. Ilk, J. Rozhon, and M. Voznak, "Deep learning serves voice cloning: how vulnerable are automatic speaker verification systems to spoofing trials?", IEEE Commun. Mag., vol. 58, no. 2, pp. 100-105, 2020. [http://dx.doi.org/10.1109/MCOM.001.1900396]
[23]
L. Garg, E. Chukwu, N. Nasser, C. Chakraborty, and G. Garg, "Anonymity preserving IoT-based COVID-19 and other infectious disease contact tracing model", IEEE Access, vol. 8, pp. 159402159414, 2020. [http://dx.doi.org/10.1109/ACCESS.2020.3020513] [PMID: 34786286]
[24]
S.M. Riazul Islam, Daehan Kwak, M. Humaun Kabir, M. Hossain, and Kyung-Sup Kwak, "The internet of things for health care: a comprehensive survey", IEEE Access, vol. 3, pp. 678-708, 2015. [http://dx.doi.org/10.1109/ACCESS.2015.2437951]
[25]
J. Qi, P. Yang, G. Min, O. Amft, F. Dong, and L. Xu, "Advanced internet of things for personalised healthcare systems: A survey", Pervasive Mobile Comput., vol. 41, pp. 132-149, 2017. [http://dx.doi.org/10.1016/j.pmcj.2017.06.018]
[26]
C.A. da Costa, C.F. Pasluosta, B. Eskofier, D.B. da Silva, and R. da Rosa Righi, "Internet of Health Things: Toward intelligent vital signs monitoring in hospital wards", Artif. Intell. Med., vol. 89, pp. 61-69, 2018. [http://dx.doi.org/10.1016/j.artmed.2018.05.005] [PMID: 29871778]
[27]
H. HaddadPajouh, A. Dehghantanha, R. M. Parizi, M. Aledhari, and H. Karimipour, “A survey on internet of things security: Requirements, challenges, and solutions”, Internet Things, vol. 14, p. 100129, 2021. [http://dx.doi.org/10.1016/j.iot.2019.100129]
[28]
T. Ai, Z. Yang, H. Hou, C. Zhan, C. Chen, W. Lv, Q. Tao, Z. Sun, and L. Xia, "Correlation of chest CT and RT-PCR testing for coronavirus disease 2019 (COVID-19) in China: a report of 1014 cases", Radiology, vol. 296, no. 2, pp. E32-E40, 2020. [http://dx.doi.org/10.1148/radiol.2020200642] [PMID: 32101510]
[29]
R. Vaishya, M. Javaid, I.H. Khan, and A. Haleem, "Artificial Intelligence (AI) applications for COVID-19 pandemic", Diabetes Metab. Syndr., vol. 14, no. 4, pp. 337-339, 2020. [http://dx.doi.org/10.1016/j.dsx.2020.04.012] [PMID: 32305024]
[30]
S. W. Chen, X. W. Gu, J. J. Wang, and H. S. Zhu, “AIoT Used for COVID-19 Pandemic Prevention and Control”, Contrast Media Mol. Imaging, vol. 2021, 2021.
[31]
M. S, and R. Latha, "Artificial intelligence with the internet of things on healthcare systems: A survey", International Journal of Advanced Trends in Computer Science and Engineering, vol. 8, no. 6, pp. 2847-2854, 2019.
Future Pandemics
Computational Intelligence for Data Analysis, Vol. 2 103
[http://dx.doi.org/10.30534/ijatcse/2019/27862019] [32]
D.B. Neill, "Using artificial intelligence to improve hospital inpatient care", IEEE Intell. Syst., vol. 28, no. 2, pp. 92-95, 2013. [http://dx.doi.org/10.1109/MIS.2013.51]
[33]
C.S. Lee, P.G. Nagy, S.J. Weaver, and D.E. Newman-Toker, "Cognitive and system factors contributing to diagnostic errors in radiology", AJR Am. J. Roentgenol., vol. 201, no. 3, pp. 611-617, 2013. [http://dx.doi.org/10.2214/AJR.12.10375] [PMID: 23971454]
[34]
F. Jiang, Y. Jiang, H. Zhi, Y. Dong, H. Li, S. Ma, Y. Wang, Q. Dong, H. Shen, and Y. Wang, "Artificial intelligence in healthcare: past, present and future", Stroke Vasc. Neurol., vol. 2, no. 4, pp. 230-243, 2017. [http://dx.doi.org/10.1136/svn-2017-000101] [PMID: 29507784]
[35]
S.M.A.A. Abir, S.N. Islam, A. Anwar, A.N. Mahmood, and A.M.T. Oo, "Building resilience against COVID-19 pandemic using artificial intelligence, machine learning, and IoT: A survey of recent progress", IoT, vol. 1, no. 2, pp. 506-528, 2020. [http://dx.doi.org/10.3390/iot1020028]
[36]
Y. Wang, M. Hu, Q. Li, X.-P. Zhang, G. Zhai, and N. Yao, “Abnormal respiratory patterns classifier may contribute to large-scale screening of people infected with COVID-19 in an accurate and unobtrusive manner”, ArXiv Prepr. ArXiv200205534, 2020.
[37]
A. Haleem, M. Javaid, R. Vaishya, and S.G. Deshmukh, "Areas of academic research with the impact of COVID-19", Am. J. Emerg. Med., vol. 38, no. 7, pp. 1524-1526, 2020. [http://dx.doi.org/10.1016/j.ajem.2020.04.022] [PMID: 32317202]
[38]
R.P. Singh, M. Javaid, A. Haleem, and R. Suman, "Internet of things (IoT) applications to fight against COVID-19 pandemic", Diabetes Metab. Syndr., vol. 14, no. 4, pp. 521-524, 2020. [http://dx.doi.org/10.1016/j.dsx.2020.04.041] [PMID: 32388333]
[39]
R. Kumar, F. Al-Turjman, L. Anand, A. Kumar, S. Magesh, K. Vengatesan, R. Sitharthan, and M. Rajesh, "Genomic sequence analysis of lung infections using artificial intelligence technique", Interdiscip. Sci., vol. 13, no. 2, pp. 192-200, 2021. [http://dx.doi.org/10.1007/s12539-020-00414-3] [PMID: 33558984]
[40]
R. Geetha, S. Sivasubramanian, M. Kaliappan, S. Vimal, and S. Annamalai, "Cervical cancer identification with synthetic minority oversampling technique and PCA analysis using random forest classifier", J. Med. Syst., vol. 43, no. 9, p. 286, 2019. [http://dx.doi.org/10.1007/s10916-019-1402-6] [PMID: 31312985]
[41]
S. K. Nataraj, F. Al-Turjman, A. H. Adom, R. Sitharthan, M. Rajesh, and R. Kumar, “Intelligent robotic chair with thought control and communication aid using higher order spectra band features”, IEEE Sens. J., vol. 22, no. 18, pp. 17362-17369, 2022. [http://dx.doi.org/10.1109/JSEN.2020.3020971]
[42]
S. Pradeepa, K.R. Manjula, S. Vimal, M.S. Khan, N. Chilamkurti, and A.K. Luhach, "DRFS: detecting risk factor of stroke disease from social media using machine learning techniques", Neural Process. Lett., pp. 1-19, 2020. [http://dx.doi.org/10.1007/s11063-020-10279-8]
[43]
R. Sitharthan, M. Rajesh, K. Madurakavi, J. Raglend, and R. Kumar, “Assessing nitrogen dioxide (NO2) impact on health pre-and post-COVID-19 pandemic using IoT in India”, Int. J. Pervasive Comput. Commun., vol. 9, 2020.
[44]
M. Ramamurthy, I. Krishnamurthi, S. Vimal, and Y. H. Robinson, “Deep learning based genome analysis and NGS-RNA LL identification with a novel hybrid model”, Biosystems, vol. 197, p. 104211, 2020. [http://dx.doi.org/10.1016/j.biosystems.2020.104211] [PMID: 32795485]
104 Computational Intelligence for Data Analysis, Vol. 2
Virmani et al.
[45]
R. Sitharthan, S. Krishnamoorthy, P. Sanjeevikumar, J. B. Holm-Nielsen, R. R. Singh, and M. Rajesh, “Torque ripple minimization of PMSM using an adaptive Elman neural network-controlled feedback linearization-based direct torque control strategy”, Int. Trans. Electr. Energy Syst., vol. 31, no. 1, p. e12685, 2021.
[46]
J. McCarthy, Artificial intelligence, logic and formalizing common sense.Philosophical logic and artificial intelligence. Springer, 1989, pp. 161-190. [http://dx.doi.org/10.1007/978-94-009-2448-2_6]
[47]
A.M. Turing, Computing machinery and intelligence.Parsing the turing test. Springer, 2009, pp. 2365. [http://dx.doi.org/10.1007/978-1-4020-6710-5_3]
[48]
S. Russell and P. Norvig, “Artificial intelligence: A modern approach. third edit”, Up. Saddle River N. J., vol. 7458, 2010.
[49]
D. Grover and B. Toghi, “MNIST dataset classification utilizing k-NN classifier with modified sliding-window metric”, in Science and Information Conference, pp. 583–591, 2019.
[50]
Y. Sun, D. Liang, X. Wang, and X. Tang, “Deepid3: Face recognition with very deep neural networks”, ArXiv Prepr. ArXiv150200873, 2015.
[51]
T. Schinköthe, M.R. Gabri, M. Mitterer, P. Gouveia, V. Heinemann, N. Harbeck, and M. Subklewe, “A web-and app-based connected care solution for COVID-19 in-and outpatient care: qualitative study and application development”, JMIR Public Health Surveill., vol. 6, no. 2, p. e19033, 2020. [http://dx.doi.org/10.2196/19033] [PMID: 32406855]
[52]
W. Tan and J. Liu, “Application of face recognition in tracing COVID-19 fever patients and close contacts”, in 2020 19th IEEE International Conference on Machine Learning and Applications (ICMLA), 2020, pp. 1112–1116. [http://dx.doi.org/10.1109/ICMLA51294.2020.00179]
[53]
Y. Zhang, Q. Wang, and S.H. Yuan, "Introduction to the special issue on computer modelling of transmission, spread, control and diagnosis of COVID-19", Comput. Model. Eng. Sci., vol. 127, no. 2, pp. 385-387, 2021. [http://dx.doi.org/10.32604/cmes.2021.016386]
[54]
D. Hou, R. Hou and J. Hou, "Interpretable Saab Subspace Network for COVID-19 Lung Ultrasound Screening," 2020 11th IEEE Annual Ubiquitous Computing, Electronics & Mobile Communication Conference (UEMCON), New York, NY, USA, pp. 0393-0398, 2020. [http://dx.doi.org/10.1109/UEMCON51285.2020.9298069]
[55]
L. Lonini, N. Shawen, O. Botonis, M. Fanton, C. Jayaraman, C.K. Mummidisetty, S.Y. Shin, C. Rushin, S. Jenz, S. Xu, J.A. Rogers, and A. Jayaraman, "Rapid screening of physiological changes associated with COVID-19 using soft-wearables and structured activities: a pilot study", IEEE J. Transl. Eng. Health Med., vol. 9, pp. 1-11, 2021. [http://dx.doi.org/10.1109/JTEHM.2021.3058841] [PMID: 33665044]
[56]
P.K. Singh, S. Nandi, K.Z. Ghafoor, U. Ghosh, and D.B. Rawat, "Preventing COVID-19 spread using information and communication technology", IEEE Consum. Electron. Mag., vol. 10, no. 4, pp. 18-27, 2021. [http://dx.doi.org/10.1109/MCE.2020.3047608]
[57]
H. Mulchandani, P. Dalal, O.A. Ramwala, P. Parikh, U. Dalal, M. Paunwala and C. Paunwala, "Tonsillitis based Early Diagnosis of COVID-19 for Mass-Screening using One-Shot Learning Framework," 2020 IEEE 17th India Council International Conference (INDICON), New Delhi, India, 2020, pp. 1-6. [http://dx.doi.org/10.1109/INDICON49873.2020.9342371]
[58]
M.N. Mohammed, H. Syamsudin, and S. Al-Zubaidi, "R. R. AKS, and E. Yusuf, “Novel COVID-19 detection and diagnosis system using IOT based smart helmet”,", Int. J. Psychosoc. Rehabil., vol. 24,
Future Pandemics
Computational Intelligence for Data Analysis, Vol. 2 105
no. 7, pp. 2296-2303, 2020. [59]
M. A. El Khaddar and M. Boulmalf, “Smartphone: the ultimate IoT and IoE device”, Smartphones Appl. Res. Perspect., vol. 137, 2017.
[60]
V. Chamola, V. Hassija, V. Gupta, and M. Guizani, "A comprehensive review of the COVID-19 pandemic and the role of IoT, drones, AI, blockchain, and 5G in managing its impact", IEEE Access, vol. 8, pp. 90225-90265, 2020. [http://dx.doi.org/10.1109/ACCESS.2020.2992341]
[61]
T. Tamura, M. Huang, and T. Togawa, "Current developments in wearable thermometers", Advanced Biomedical Engineering, vol. 7, no. 0, pp. 88-99, 2018. [http://dx.doi.org/10.14326/abe.7.88]
[62]
S. Ghosh, “Police in China, Dubai, and Italy are using these surveillance helmets to scan people for COVID-19 fever as they walk past and it may be our future normal”, Bus. Insid., vol. 17, 2020.
[63]
M. Tavakoli, J. Carriere, and A. Torabi, “Robotics, smart wearable technologies, and autonomous intelligent systems for healthcare during the COVID-19 pandemic: An analysis of the state of the art and future vision”, Adv. Intell. Syst., vol. 2, no. 7, p. 2000071, 2020. [http://dx.doi.org/10.1002/aisy.202000071]
[64]
K. Knibbs, “There’s no cure for COVID-19 loneliness, but robots can help”, Wired Mag., 2020.
[65]
“aarogya setu app: How to use Aarogya Setu app and find out if you have coronavirus symptoms - The Economic Times. https://economictimes.indiatimes.com/tech/software/how-to-use-aarogya-set-app-and-find-out-if-you-have-covid-19-symptoms/articleshow/75023152.cms (accessed Mar. 26, 2022).
[66]
T. India, “Indian government launches Aarogya Setu app to track coronavirus infections.” 2020.
[67]
“Aarogya Setu Mobile App”, MyGov.in. https://www.mygov.in/aarogya-setu-app/ (accessed Mar. 25, 2022).
[68]
M. Sharma, “How drones are being used to combat COVID-19”, Geospatial World, vol. 4, 2020.
[69]
J. Reagan, “COVID-related Drone Delivery Soars Into Canadian First Nation”, DRONELIFE, Jun. 10, 2020. https://dronelife.com/2020/06/10/ddc-drone-covid-drone-delivery/ (accessed Mar. 25, 2022).
[70]
M. Ndiaye, S.S. Oyewobi, A.M. Abu-Mahfouz, G.P. Hancke, A.M. Kurien, and K. Djouani, "IoT in the wake of COVID-19: A survey on contributions, challenges and evolution", IEEE Access, vol. 8, pp. 186821-186839, 2020. [http://dx.doi.org/10.1109/ACCESS.2020.3030090] [PMID: 34786294]
[71]
T. Ozturk, M. Talo, E.A. Yildirim, U.B. Baloglu, O. Yildirim, and U. Rajendra Acharya, "Automated detection of COVID-19 cases using deep neural networks with X-ray images", Comput. Biol. Med.,. vol. 121, p. 103792, 2020. [http://dx.doi.org/10.1016/j.compbiomed.2020.103792] [PMID: 32568675]
[72]
B. D. Deebak and F. Al-Turjman, “Secure-user sign-in authentication for IoT-based eHealth systems”, Complex Intell. Syst., pp. 1–21, 2021.
[73]
S.H. Wang, S.L. Fernandes, Z. Zhu, and Y.D. Zhang, "AVNC: attention-based VGG-style network for COVID-19 diagnosis by CBAM", IEEE Sens. J., vol. 22, no. 18, pp. 17431-17438, 2022. [http://dx.doi.org/10.1109/JSEN.2021.3062442] [PMID: 36346097]
[74]
M. Jaderberg, A. Vedaldi, and A. Zisserman, "Deep features for text spotting", European conference on computer vision, pp. 512–528, 2014.
[75]
Y. Lecun, L. Bottou, Y. Bengio, and P. Haffner, "Gradient-based learning applied to document recognition", Proc. IEEE, vol. 86, no. 11, pp. 2278-2324, 1998. [http://dx.doi.org/10.1109/5.726791]
[76]
W. Yin, X. Yang, L. Zhang, and E. Oki, "ECG monitoring system integrated with IR-UWB radar
106 Computational Intelligence for Data Analysis, Vol. 2
Virmani et al.
based on CNN", IEEE Access, vol. 4, p. 1, 2016. [http://dx.doi.org/10.1109/ACCESS.2016.2608777] [77]
A. Fakhry, T. Zeng, and S. Ji, "Residual deconvolutional networks for brain electron microscopy image segmentation", IEEE Trans. Med. Imaging, vol. 36, no. 2, pp. 447-456, 2017. [http://dx.doi.org/10.1109/TMI.2016.2613019] [PMID: 28113967]
[78]
R. Guerrero, C. Qin, O. Oktay, C. Bowles, L. Chen, R. Joules, R. Wolz, M.C. Valdés-Hernández, D.A. Dickie, J. Wardlaw, and D. Rueckert, "White matter hyperintensity and stroke lesion segmentation and differentiation using convolutional neural networks", Neuroimage Clin., vol. 17, pp. 918-934, 2018. [http://dx.doi.org/10.1016/j.nicl.2017.12.022] [PMID: 29527496]
[79]
Z. Jiang, and X. Xu, "“Intelligent assisted diagnosis of COVID-19 based on CT images”, J", Mol. Imaging, vol. 43, no. 2, pp. 264-269, 2020.
[80]
J.C. Gomes, "IKONOS: an intelligent tool to support diagnosis of COVID-19 by texture analysis of Xray images", Res. Biomed. Eng., pp. 1-14, 2020.
[81]
R. Sitharthan, and M. Rajesh, "Application of machine learning (ML) and internet of things (IoT) in healthcare to predict and tackle pandemic situation", Distrib Parallel Databases, vol. 40, no. 4, p. 887, 2021. [PMID: 34393377]
[82]
A.E. Ezugwu, I.A.T. Hashem, O.N. Oyelade, M. Almutari, M.A. Al-Garadi, I.N. Abdullahi, O. Otegbeye, A.K. Shukla and H. Chiroma, “A novel smart city-based framework on perspectives for application of machine learning in combating covid-19,” BioMed Res. Int., vol. 2021, 2021.
[83]
S. Ekins, J. S. Freundlich, A. M. Clark, M. Anantpadma, R. A. Davey, and P. Madrid, “Machine learning models identify molecules active against the Ebola virus in vitro,” F1000Research, vol. 4, 2015.
[84]
L. Zhang, H.X. Ai, S.M. Li, M.Y. Qi, J. Zhao, Q. Zhao, and H.S. Liu, "Virtual screening approach to identifying influenza virus neuraminidase inhibitors using molecular docking combined with machinelearning-based scoring function", Oncotarget, vol. 8, no. 47, pp. 83142-83154, 2017. [http://dx.doi.org/10.18632/oncotarget.20915] [PMID: 29137330]
[85]
P. Gonzalez-Dias, E.K. Lee, S. Sorgi, D.S. de Lima, A.H. Urbanski, E.L. Silveira, and H.I. Nakaya, "Methods for predicting vaccine immunogenicity and reactogenicity", Hum. Vaccin. Immunother., vol. 16, no. 2, pp. 269-276, 2020. [http://dx.doi.org/10.1080/21645515.2019.1697110] [PMID: 31869262]
[86]
M.S. Nawaz, P. Fournier-Viger, A. Shojaee, and H. Fujita, "Using artificial intelligence techniques for COVID-19 genome analysis", Appl. Intell., vol. 51, no. 5, pp. 3086-3103, 2021. [http://dx.doi.org/10.1007/s10489-021-02193-w] [PMID: 34764587]
[87]
C.L.P. Eng, J.C. Tong, and T.W. Tan, "Predicting host tropism of influenza A virus proteins using random forest", BMC Med. Genomics,. vol. 7, no. 3, pp. 1–11, 2014. [http://dx.doi.org/10.1186/1755-8794-7-S3-S1] [PMID: 25521718]
Computational Intelligence for Data Analysis, 2023, Vol. 2, 107-119
107
CHAPTER 7
Cyber Secure AIoT Applications in Future Pandemics Maria Nawaz Chohan1,* and Sana Nawaz Chohan2 1 2
National Defence University, Islamabad, Pakistan Foundation University Institute of Rehabilitation Sciences, Islamabad, Pakistan Abstract: In the era of digitalization, artificial intelligence and IoT play an important role in COVID-19. Collecting real-time data using the internet of things has removed barriers and improved end-to-end delays between patients & doctors. During COVID19, IoT connected people through wireless communication technology. However, by utilizing AI, different diseases can be identified easily. This research article has merged IoT with AI, which is called the Artificial Internet of Things (AIoT). Monitoring of patient health can be made possible due to the sub-class of AI known as machine learning. Industry 5.0 has combined big data, IoT, AI, 5G and cognitive ICT technologies to exchange information. Due to the widespread of dangerous diseases, people face several challenges, including inadequate preparation, shortage of medicines and poor resources, and increasing death rates. Data collection is the initial step toward research and innovation. Therefore, many applications are discussed properly, which include tele-medicine, early warning systems, wearable devices, and UAVs that help to support the healthcare industry.
Keywords: AI, COVID-19, IoT, UAVs. INTRODUCTION Integration of AI & IoT has gained a lot of attention, improving standards, especially in the healthcare industry. Internet of things uses sensors to collect information. Zigbee is a contemporary protocol that is considered the backbone of IoT [1]. Every country tried a lot to reduce COVID-19 by giving awareness using social and mainstream media. However, there exist many applications related to healthcare which include remote patient monitoring, fighting against COVID-19, heart disease detection, electronic medical report cards and tracking the infectious disease. Utilizing machine learning, doctors predict different diseases easily [2]. IoT enhances machine-to-machine and human-to-machine cognition to learn from Corresponding author Maria Nawaz Chohan: National Defence University, Islamabad, Pakistan; E-mail: [email protected]
*
Mariya Ouaissa, Mariyam Ouaissa, Zakaria Boulouad, Inam Ullah Khan, Sailesh Iyer (Eds.) All rights reserved-© 2023 Bentham Science Publishers
108 Computational Intelligence for Data Analysis, Vol. 2
Chohan and Chohan
the environment [3]. Smart IoT-based machine learning concepts train algorithms to provide efficient results in the healthcare industry [4]. Industry 5.0 will provide combined knowledge of IoT, big data, AI and 5G to give possible solutions regarding healthcare. While multi-kernel is used in IoT applications for processing better-quality images [5]. Also, smart mobile applications are developed to monitor the health status of every patient [6]. Moreover, telehealth web and mobile applications are used all over the world to give optimal consultation to patients [7]. Traditional server-client environments can be utilized using wireless body area sensor networks to provide high accuracy levels in healthcare applications [8, 9]. The major contribution points of this research paper are as under: ● ● ● ● ●
Artificial Internet of things applications for healthcare Machine learning techniques for COVID-19 Industry 5.0 for smart healthcare systems Using flying vehicles in the health industry People’s facing challenges because of future pandemics
This research article has diverse studies about AIoT, machine learning techniques, industry 5.0, UAVs and future challenges because of pandemics. Fig. (1) shows wireless body area IoT sensor nodes connectivity with mobile devices using the internet for tele-medicines. Location (GPS) Communication Gateway
Base Station
Mobile
Wearables
Internet Blu
eto
oth
Motion
/Zig
Bee
Doctor
/WL
AN
Telehealth-care (Remote monitoring) Wellness
Digital Health Remote surgery
Connected ambulances
Fig. (1). Wireless body area IoT sensor networks for tele-medicine.
loT sensor monitoring
Future Pandemics
Computational Intelligence for Data Analysis, Vol. 2 109
LITERATURE STUDY AI-based models are designed to mimic the threat and unstable behavior of patients. Various types of AI technologies are initialized due to that COVID-19 can be tackled easily [10]. However, IoT can be used to secure communication mediums between patients & doctors [11]. Wearable systems need to be utilized, which give signals to satellites and base stations for further action [12]. Investigation of IoT-based applications is designed to improve the healthcare industry [13]. Industry 5.0 has revolutionized healthcare by utilizing IoT, AI, big data and communication networks. Tele-medicines applications are used a lot during COVID-19 to remotely discuss health-related issues with doctors [14]. Online smart clinics and IoT sensors or wearables are introduced to help people during COVID-19 [15]. The whole world has suffered COVID-19, which has exhausted my entire life. AI & and IoT have made a huge impact to facilitate humans [16]. A novel algorithm is formulated to optimize UAV energy levels using E-AntHocNet, which has many healthcare applications. UAV’s can be used for monitoring rescue operations and sending medical equipment’s from one place to another. Flying ad hoc networks is the combination of UAVs. Due to the high level of mobility, aerial vehicles use routing techniques to send information using the shortest routes [17]. Table 1. Various applications for patients monitoring. Author/Reference
Applications
Description
Timmers T. et al. [18]
Education based application
Providing information about patients
Bourdon H. et al. [19]
Hospital information system application
Pandemic COVID-19
Medina M. et al. [20]
Monitoring-based application designed for home
Collecting information on the telephone
Drew D & Nguyen L. et al. [21]
IoT-based monitoring app
Data collection from the internet of medical things
Ben Hassen H. et al. [22]
Students Mental Health Monitoring
Checking & evaluating behavior of students during COVID-19
Yamamoto K. et al. [23]
Data sharing using a simple mail transfer protocol
Email-based observation and generating a receipt
Huckins J.F. & daSilva A.W. et al. [24]
Tele-net communication using a phone
Specifically designed for eyes
110 Computational Intelligence for Data Analysis, Vol. 2
Chohan and Chohan
In addition, Table 1 describes different applications utilized during COVID-19 properly discussed. ARTIFICIAL INTERNET HEALTHCARE
OF
THINGS
APPLICATIONS
FOR
The Internet of things is quite helpful in many research areas, including wireless body area networks, tracking, monitoring, and more importantly, electronic health. Local and remote data collection about patient health provides better details through which doctors can make the right decision at the right time. During the COVID-19 pandemic, a novel approach is introduced called healthcare-AIoT, which mimics the spread of the virus [25 - 28]. H-AIoT consists of the following components which are as under: H-AIoT Based Hardware Tele-health monitoring systems used to have intelligent hardware to store patient data. IoT-based devices are designed to observe patient's daily routine activities. Therefore, IoT systems easily give information about heartbeat, temperature, walking & sleeping. Also, breathing and coughing can be detected by utilizing IoT models. These IoT-hardware components help doctors to make an optimal decision. In addition, web-based online patient monitoring systems are introduced, which are used mostly during COVID-19 [29]. H-AIoT Based Software Tele-medicine provides better connectivity between patients & doctors. For this purpose, Online web-based software’s comes in the category of the internet of medical things. Wireless Communication technologies can be used for optimal connectivity [30]. Communication/Routing Protocols In wireless body area sensor networks, various routing protocols are implemented in the healthcare industry to collect information on patients. Application-specific routing protocols are designed to increase the lifetime of the network [31]. UAV’s/Drones in the Healthcare Industry UAVs play a pivotal role in improving logistic services in the healthcare industry. Indeed, mini-drones are both more cost-effective and more efficient in rescue missions when compared to standard aerial vehicles. Communication between UAVs is also more efficient due to minimized end-to-end delay in communication networks [32].
Future Pandemics
Computational Intelligence for Data Analysis, Vol. 2 111
Wearable AI-IoT Sensors Wearable IoT devices have conveniently improved humans’ life. Fall detection systems use motion sensors that target the performance of the elderly person. Therefore, smart wearable clothing is introduced for integrating motion sensors in special clothing. However, China is having standard personal alert safety system [33]. AI-IoT-Based Monitoring System AI-IoT has advanced monitoring systems for giving optimal results. Therefore, on-time or early information about patient health is an important factor where doctors finalize the better decision. Various AI techniques are utilized which have detected dangerous diseases [34]. Also, AI-based applications evaluate past history, which is helpful for giving optimal predictions. However, in China, AIoT-based cameras are installed everywhere to have better observation and people follow quarantine rules. Although, in developing countries, UAVs are utilized to monitor whether peoples use to wear a mask or not [35]. For better communication between nodes' routing protocols, deployment is the best choice. Therefore, advanced routing protocols like Anthocnet, DSR, DSDV, ZRP, MDART and AOMDV are implemented in the topological environment of IoT-based FANETs. In addition, the supervised learning technique decision tree is used to improve signal strength from the ground base station to UAVs [36]. Fig. (2) depicts the flow chart where AIoT based treatment approach is designed especially for COVID-19 patients. Detection of Cyber-Attacks in IoMT The statistical modeling-based threshold is designed to detect DoS, DDoS and ping of death attacks. Therefore, the proposed intrusion detection system monitors data packets, creating a buffer overflow. The concept of queueing theory is implemented to evaluate the level of congestion in the network. Denial of service attack affects the entire process and sometimes jams the network. Also, novel IDS have minimized false alarm capabilities usually configured by network administrators [37].
112 Computational Intelligence for Data Analysis, Vol. 2
Chohan and Chohan
COVID-19 Symptoms
Artificial Internet of Things Based Treatment
AI Aided Image Segmentation Taking Samples for Decision Classification
COVID-19
COVID-19
Positive
Negative
Recovery Quarantine AIoT Treatment
Cured
Fig. (2). AIoT-based treatment for COVID-19 patient.
Machine Learning Techniques for COVID-19 Machine learning algorithms fulfill the resource allocation gap in healthcare technologies. However, ML is the emerging class of artificial intelligence. Machine learning techniques are widely utilized to give possible solutions for COVID-19. Even though, machine learning techniques are possibly divided into three main categories which include:
Future Pandemics ● ● ●
Computational Intelligence for Data Analysis, Vol. 2 113
Supervised Unsupervised Reinforcement Learning
Fig. (3) shows the techniques related to machine learning concepts. Also, ML reportedly focuses on prediction, classification, diagnoses, and drug invention. Some powerful approaches are used in combating COVID-19 which are as below: ● ● ● ●
Support Vector Machine Random Forest Decision Tree Logistic Regression
Machine Learning for COVID-19
Supervised Learning for COVID-19
UnSupervised Learning for COVID-19
Reinforcement Learning for COVID-19
Fig. (3). Machine learning techniques for COVID-19.
Researchers properly used the above-mentioned techniques, which are quite helpful in detecting, classifying, screening and discovering antibodies [38]. During implementation, the following phases are taken into consideration for making machine learning algorithms. ● ● ● ● ●
Raw form data Data processing Training dataset Testing of the external dataset Modeling & Evaluation
Industry 5.0 for Smart Healthcare Systems Industry 5.0 has given new directions to give a smart concept of digitalization and possible solutions for healthcare systems. However, the concept of industry 5.0 consists of optimal technologies using wireless communication as the backbone [39]. The fields utilized, especially in industry 5.0, are as below:
114 Computational Intelligence for Data Analysis, Vol. 2 ● ● ● ●
Chohan and Chohan
Big Data Internet of Medical Things Cognitive Machine learning & computing 5G communication network & beyond
Industry 5.0 Related Challenges The above-mentioned technologies support to facilitate advancement in industry 5.0 [39]. Every new idea faces a lot of challenges: ● ● ● ● ● ●
Monitoring of Peoples Medical equipments for patients Awareness & Education Research for a better future Smart hospital management system Tele-medicine approach/Online doctors’ clinic
Industry 5.0 will improve living standards in the entire world. Smart transportation using UAVs, rescue, smart hospitals and telemedicine will enhance the concept of smart cities. Therefore, the smart city approach is still in its infancy and needs major improvements. Even in developed countries, smart cities is quite new, but in the near future, AIoT will upgrade connectivity between peoples. Using Flying Vehicles in Health Industry UAV’s can be utilized on the non-military side as well, which facilitates humans a lot. Especially in COVID-19, drones are being investigated to reduce the burden on healthcare professionals. However, machine learning-based UAVs are quite efficient to complete every task on time. Aerial vehicles can be used for monitoring and surveillance. Recently, UAVs have been used in the medical field to rescue someone by sending information to the base station. Drones have diverse applications, which include spraying on COVID-19 patients; installation of loudspeakers in UAV’s which can be used to identify threats against the entire network [40]. Furthermore, aerial vehicles applications are as follows: ● ● ●
Monitoring people for wearing masks. Data/sample collection. UAV-based taxi for transportation.
Future Pandemics
Computational Intelligence for Data Analysis, Vol. 2 115
In addition, human and mini-drone interaction can be optimized by formulating standards and protocols. Future Challenges In the near future, novel disease prevention and detection will be easily possible due to artificial intelligence. Also, the techniques can be improved by utilizing machine learning techniques. AIoT devices will be incorporated into the human body, which will give predictions regarding disease detection like heart attack, cancer, COVID-19 and many more. Moreover, the world is now prepared for pandemics like COVID-19, where people can switch their activities online easily. Although, UAVs will provide help to rescue and provide accurate location coordinates. Apart from that, deep learning, cognitive IoT and optimization techniques can be utilized to improve living standards in smart cities. CONCLUSION AI-based techniques with the combination of IoT will improve connectivity in smart cities. Especially the use of AI-based applications increased due to COVID19, where people are restricted to homes. Initially, all country governments were having problems; later, they converted their education, industry and everything online. This paper provides a detailed understating of AIoT techniques to combat future pandemics. Machine learning techniques are incorporated to improve applications related to COVID-19. The future smart cities and industry 5.0 is introduced in this research paper, which will improve the standard of living. Also, future challenges are deeply discussed at the end of the paper. In addition, EAnthocNet hybrid protocol can be deployed in various applications for further innovation. CONSENT FOR PUBLICATION Not applicable. CONFLICT OF INTEREST The author declares no conflict of interest, financial or otherwise. ACKNOWLEDGEMENT Declared none.
116 Computational Intelligence for Data Analysis, Vol. 2
Chohan and Chohan
REFERENCES [1]
V.K. Pappakrishnan, R. Mythili, V. Kavitha, and N. Parthiban. "Role of Artificial Intelligence of Things (AIoT) in Covid-19 Pandemic: A Brief Survey." In Proceedings of the 6th International Conference on Internet of Things, Big Data and Security (IoTBDS 2021), Prague, Czech Republic, pp. 23-25, 2021.
[2]
M. Muzammal, R. Talat, A.H. Sodhro, and S. Pirbhulal, "A multi-sensor data fusion enabled ensemble approach for medical data from body sensor networks", Inf. Fusion, vol. 53, pp. 155-164, 2020. [http://dx.doi.org/10.1016/j.inffus.2019.06.021]
[3]
A. Barnawi, P. Chhikara, R. Tekchandani, N. Kumar, and B. Alzahrani, "Artificial intelligenceenabled Internet of Things-based system for COVID-19 screening using aerial thermal imaging", Future Gener. Comput. Syst., vol. 124, pp. 119-132, 2021. [http://dx.doi.org/10.1016/j.future.2021.05.019] [PMID: 34075265]
[4]
Imran, U. Zaman, M. Waqar, " Using Machine Learning Algorithms for Housing Price Prediction: The Case of Islamabad Housing Data", Soft Comput. Mach. Intell. J., vol. 1, pp. 11-23, 2021.
[5]
J. Jebadurai, and J. Dinesh Peter, "Super-resolution of retinal images using multi-kernel SVR for IoT healthcare applications", Future Gener. Comput. Syst., vol. 83, pp. 338-346, 2018. [http://dx.doi.org/10.1016/j.future.2018.01.058]
[6]
J. Chuang, L. Maimoon, S. Yu, H. Zhu, C. Nybroe, O. Hsiao, S.H. Li, H. Lu, and H. Chen, SilverLink: Smart home health monitoring for senior care.ICSH. Springer: Berlin, Germany, 2015, pp. 3-14.
[7]
C.S. Namahoot, M. Brückner, and C. Nuntawong, "Mobile diagnosis system with emergency telecare in Thailand (MOD-SET)", Procedia Comput. Sci., vol. 69, pp. 86-95, 2015. [http://dx.doi.org/10.1016/j.procs.2015.10.009]
[8]
D.G. Korzun, A.V. Borodin, I.A. Timofeev, I.V. Paramonov, and S.I. Balandin, "Digital assistance services for emergency situations in personalized mobile healthcare: Smart space based approach", Proceedings of the 2015 International Conference on Biomedical Engineering and Computational Technologies (SIBIRCON),. Novosibirsk, Russia, pp. 62–67, 2015. [http://dx.doi.org/10.1109/SIBIRCON.2015.7361852]
[9]
I. Bisio, F. Lavagetto, M. Marchese, and A. Sciarrone, "A smartphone-centric platform for remote health monitoring of heart failure", Int. J. Commun. Syst., vol. 28, no. 11, pp. 1753-1771, 2015. [http://dx.doi.org/10.1002/dac.2778]
[10]
A. Brem, E. Viardot, and P.A. Nylund, "Implications of the coronavirus (COVID-19) outbreak for innovation: Which technologies will improve our lives?", Technol. Forecast. Soc. Change, vol. 163, p. 120451, 2021. [http://dx.doi.org/10.1016/j.techfore.2020.120451] [PMID: 33191956]
[11]
A.A. Hussain, O. Bouachir, F. Al-Turjman, and M. Aloqaily, "AI Techniques for COVID-19", IEEE Access, vol. 8, pp. 128776-128795, 2020. [http://dx.doi.org/10.1109/ACCESS.2020.3007939] [PMID: 34976554]
[12]
F. Wu, T. Wu, and M. Yuce, "An internet-of-things (IoT) network system for connected safety and health monitoring applications", Sensors (Basel), vol. 19, no. 1, p. 21, 2018. [http://dx.doi.org/10.3390/s19010021] [PMID: 30577646]
[13]
P. Afshar, S. Heidarian, F. Naderkhani, A. Oikonomou, K.N. Plataniotis, and A. Mohammadi, "COVID-CAPS: A capsule network-based framework for identification of COVID-19 cases from Xray images", Pattern Recognit. Lett., vol. 138, pp. 638-643, 2020. [http://dx.doi.org/10.1016/j.patrec.2020.09.010] [PMID: 32958971]
[14]
A.A.R. Alsaeedy, and E.K.P. Chong, "Detecting regions at risk for spreading COVID-19 using existing cellular wireless network functionalities", IEEE Open J. Eng. Med. Biol., vol. 1, no. 1, pp. 187-189, 2020. [http://dx.doi.org/10.1109/OJEMB.2020.3002447] [PMID: 34192281]
Future Pandemics
Computational Intelligence for Data Analysis, Vol. 2 117
[15]
Y. Siriwardhana, C. De Alwis, G. Gur, M. Ylianttila, and M. Liyanage, "Te _ght against the COVID19 pandemic with 5G technologies", IEEE Eng. Manage. Rev., vol. 48, no. 3, pp. 72-84, 2020. [http://dx.doi.org/10.1109/EMR.2020.3017451]
[16]
P.K. Kollu, K. Kumar, P.R. Kshirsagar, S. Islam, Q.N. Naveed, M.R. Hussain, and V.P. Sundramurthy, "Development of Advanced Artificial Intelligence and IoT Automation in the Crisis of COVID-19 Detection", J. Healthc. Eng., vol. 2022, pp. 1-12, 2022. [http://dx.doi.org/10.1155/2022/1987917] [PMID: 35281536]
[17]
I.U. Khan, I.M. Qureshi, M.A. Aziz, T.A. Cheema, and S.B.H. Shah, "Smart IoT Control-Based Nature Inspired Energy Efficient Routing Protocol for Flying Ad Hoc Network (FANET)", IEEE Access, vol. 8, pp. 56371-56378, 2020. [http://dx.doi.org/10.1109/ACCESS.2020.2981531]
[18]
T. Timmers, L. Janssen, J. Stohr, J.L. Murk, and M A H. Berrevoets, "Using eHealth to support COVID-19 education, self-assessment, and symptom monitoring in the Netherlands: observational study", JMIR Mhealth Uhealth, vol. 8, no. 6, p. e19822, 2020. [http://dx.doi.org/10.2196/19822] [PMID: 32516750]
[19]
H. Bourdon, R. Jaillant, A. Ballino, P. El Kaim, L. Debillon, S. Bodin, and L. N’Kosi, "Teleconsultation in primary ophthalmic emergencies during the COVID-19 lockdown in Paris: Experience with 500 patients in March and April 2020", J. Fr. Ophtalmol., vol. 43, no. 7, pp. 577-585, 2020. [http://dx.doi.org/10.1016/j.jfo.2020.05.005] [PMID: 32564983]
[20]
M. Medina, C. Babiuch, M. Card, R. Gavrilescu, W. Zafirau, E. Boose, K. Giuliano, A. Kim, R. Jones, and A. Boissy, "Home monitoring for COVID-19", Cleve. Clin. J. Med., 2020. [http://dx.doi.org/10.3949/ccjm.87a.ccc028] [PMID: 32409432]
[21]
D.A. Drew, L.H. Nguyen, C.J. Steves, C. Menni, M. Freydin, T. Varsavsky, C.H. Sudre, M.J. Cardoso, S. Ourselin, J. Wolf, T.D. Spector, A.T. Chan, A.T. Chan, D.A. Drew, L.H. Nguyen, A.D. Joshi, C-G. Guo, W. Ma, C-H. Lo, R.S. Mehta, S. Kwon, D.R. Sikavi, M.V. Magicheva-Gupta, Z.S. Fatehi, J.J. Flynn, B.M. Leonardo, C.M. Albert, G. Andreotti, L.E. Beane-Freeman, B.A. Balasubramanian, J.S. Brownstein, F. Bruinsma, A.N. Cowan, A. Deka, M.E. Ernst, J.C. Figueiredo, P.W. Franks, C.D. Gardner, I.M. Ghobrial, C.A. Haiman, J.E. Hall, S.L. Deming-Halverson, B. Kirpach, J.V. Lacey Jr, L. Le Marchand, C.R. Marinac, M.E. Martinez, R.L. Milne, A.M. Murray, D. Nash, J.R. Palmer, A.V. Patel, L. Rosenberg, D.P. Sandler, S.V. Sharma, S.H. Schurman, L.R. Wilkens, J.E. Chavarro, A.H. Eliassen, J.E. Hart, J.H. Kang, K.C. Koenen, L.D. Kubzansky, L.A. Mucci, S. Ourselin, J.W. RichEdwards, M. Song, M.J. Stampfer, C.J. Steves, W.C. Willett, J. Wolf, and T. Spector, "Rapid implementation of mobile technology for real-time epidemiology of COVID-19", Science, vol. 368, no. 6497, pp. 1362-1367, 2020. [http://dx.doi.org/10.1126/science.abc0473] [PMID: 32371477]
[22]
H. Ben Hassen, N. Ayari, and B. Hamdi, "A home hospitalization system based on the Internet of things, Fog computing and cloud computing", Informatics in Medicine Unlocked, vol. 20, p. 100368, 2020. [http://dx.doi.org/10.1016/j.imu.2020.100368] [PMID: 32537483]
[23]
K. Yamamoto, T. Takahashi, M. Urasaki, Y. Nagayasu, T. Shimamoto, Y. Tateyama, K. Matsuzaki, D. Kobayashi, S. Kubo, S. Mito, T. Abe, H. Matsuura, and T. Iwami, "Health observation app for COVID-19 symptom tracking integrated with personal health records: proof of concept and practical use study", JMIR Mhealth Uhealth, vol. 8, no. 7, p. e19902, 2020. [http://dx.doi.org/10.2196/19902] [PMID: 32568728]
[24]
J.F. Huckins, A.W. daSilva, W. Wang, E. Hedlund, C. Rogers, S.K. Nepal, J. Wu, M. Obuchi, E.I. Murphy, M.L. Meyer, D.D. Wagner, P.E. Holtzheimer, and A.T. Campbell, "Mental health and behavior of college students during the early phases of the COVID-19 pandemic: longitudinal smartphone and ecological momentary assessment study", J. Med. Internet Res., vol. 22, no. 6, p. e20185, 2020.
118 Computational Intelligence for Data Analysis, Vol. 2
Chohan and Chohan
[http://dx.doi.org/10.2196/20185] [PMID: 32519963] [25]
D.D. Olatinwo, A. Abu-Mahfouz, and G. Hancke, "A survey on LPWAN technologies in WBAN for remote health-care monitoring", Sensors (Basel), vol. 19, no. 23, p. 5268, 2019. [http://dx.doi.org/10.3390/s19235268] [PMID: 31795483]
[26]
V. Singh, H. Chandna, A. Kumar, S. Kumar, N. Upadhyay, and K. Utkarsh, "IoT-Q-Band: A low cost internet of things based wearable band to detect and track absconding COVID-19 quarantine subjects", EAI Endorsed Transactions on Internet of Things, vol. 6, no. 21, p. 163997, 2020. [http://dx.doi.org/10.4108/eai.13-7-2018.163997]
[27]
C. Raj, C. Jain, and W. Arif, ‘‘HEMAN: Health monitoring and nous: An IoT based e-health care system for remote telemedicine,’’ in Proc. Int. Conf. Wireless Commun., Signal Process. Netw. (WiSPNET), pp. 2115–2119, 2017. [http://dx.doi.org/10.1109/WiSPNET.2017.8300134]
[28]
J. Mohammed, C.-H. Lung, A. Ocneanu, A. Thakral, C. Jones, and A. Adler, ‘‘Internet of Things: Remote patient monitoring using Web services and cloud computing,’’ in Proc. IEEE Int. Conf. Internet Things (iThings), IEEE Green Comput. Commun. (GreenCom) IEEE Cyber, Phys. Social Comput. (CPSCom), pp. 256–263, 2014. [http://dx.doi.org/10.1109/iThings.2014.45]
[29]
T. Malche, S. Tharewal, P.K. Tiwari, M.Y. Jabarulla, A.A. Alnuaim, W.A. Hatamleh, and M.A. Ullah, "Artificial Intelligence of Things- (AIoT-) Based Patient Activity Tracking System for Remote Patient Monitoring", J. Healthc. Eng., vol. 2022, pp. 1-15, 2022. [http://dx.doi.org/10.1155/2022/8732213] [PMID: 35273786]
[30]
M.M. Abdellatif, and W. Mohamed, "Telemedicine: An IoT Based Remote Healthcare System", International Journal of Online & Biomedical Engineering, vol. 16, no. 6, 2020. [http://dx.doi.org/10.3991/ijoe.v16i06.13651]
[31]
M.R. Rahman, M.M. Islam, A.I. Pritom, and Y. Alsaawy, "ASRPH: Application Specific Routing Protocol for Health care", Comput. Netw., vol. 197, p. 108273, 2021. [http://dx.doi.org/10.1016/j.comnet.2021.108273]
[32]
H.E. Comtet, and K.A. Johannessen, "A Socio-Analytical Approach to the Integration of Drones into Health Care Systems", Information (Basel), vol. 13, no. 2, p. 62, 2022. [http://dx.doi.org/10.3390/info13020062]
[33]
X. Chai, R. Wu, M. Pike, H. Jin, W.Y. Chung, and B.G. Lee, "Smart Wearables with Sensor Fusion for Fall Detection in Firefighting", Sensors (Basel), vol. 21, no. 20, p. 6770, 2021. [http://dx.doi.org/10.3390/s21206770] [PMID: 34695983]
[34]
A. Kishor, and C. Chakraborty, "Artificial intelligence and internet of things based healthcare 4.0 monitoring system", Wirel. Pers. Commun., pp. 1-17, 2021.
[35]
R.K.R. Kummitha, "Smart technologies for fighting pandemics: The techno- and human- driven approaches in controlling the virus transmission", Gov. Inf. Q., vol. 37, no. 3, p. 101481, 2020. [http://dx.doi.org/10.1016/j.giq.2020.101481] [PMID: 32313361]
[36]
I.U. Khan, S.B.H. Shah, L. Wang, M.A. Aziz, T. Stephan, and N. Kumar, "Routing protocols & unmanned aerial vehicles autonomous localization in flying networks", Int. J. Commun. Syst., p. e4885, 2021. [http://dx.doi.org/10.1002/dac.4885]
[37]
A. Abdollahi, and M. Fathi, "An Intrusion Detection System on Ping of Death Attacks in IoT Networks", Wirel. Pers. Commun., vol. 112, no. 4, pp. 2057-2070, 2020. [http://dx.doi.org/10.1007/s11277-020-07139-y]
[38]
M. Khan, M.T. Mehran, Z.U. Haq, Z. Ullah, S.R. Naqvi, M. Ihsan, and H. Abbass, "Applications of artificial intelligence in COVID-19 pandemic: A comprehensive review", Expert Syst. Appl., vol. 185, p. 115695, 2021.
Future Pandemics
Computational Intelligence for Data Analysis, Vol. 2 119
[http://dx.doi.org/10.1016/j.eswa.2021.115695] [PMID: 34400854] [39]
M. Javaid, A. Haleem, R.P. Singh, M.I.U. Haq, A. Raina, and R. Suman, "Industry 5.0: Potential Applications in COVID-19", Journal of Industrial Integration and Management, vol. 5, no. 4, pp. 507-530, 2020. [http://dx.doi.org/10.1142/S2424862220500220]
[40]
Á. Restás, "Drone Applications Fighting COVID-19 Pandemic—Towards Good Practices", Drones (Basel), vol. 6, no. 1, p. 15, 2022. [http://dx.doi.org/10.3390/drones6010015]
120
Computational Intelligence for Data Analysis, 2023, Vol. 2, 120-136
CHAPTER 8
Machine Learning Solution for Orthopedics: A Comprehensive Review Muhammad Imad1,*, Muhammad Abul Hassan1, Shah Hussain Bangash1 and Naimullah1 1
Abasyn University, Peshawar, Pakistan Abstract: Bone provides support to the skeletal system, associated with joints, cartilage, and muscles attached to bones to help move the body and protect the human internal organs. Bone fracture is a common ailment in the human body due to external force on the bone. The structure of the bone is disturbed, which causes pain, frailness, and bone not functioning properly. Avulsion fracture, Greenstick fracture, Comminuted fracture, Compression fracture, Simple fracture, and Open fracture are different types of fractures. The literature presents a significant number of research papers covering the detection of different kinds of fractures (wrist, hand, leg, skull, spine, chest, etc.). There are different medical imaging tools available such as X-ray, Magnetic Resonance Imaging (MRI), Computed Tomography (CT) and ultrasound, which detect different types of fractures. This paper represents a review study to discuss various bone fracture detection and classification techniques between fracture and non-fracture bone.
Keywords: Bone Fracture, Classification, CT, Detection, Machine Learning, MRI, Orthopedics, X-ray. INTRODUCTION Medical image processing is an evolving field of science growing immensely in the healthcare industry because of its advancements in technology and software breakthroughs. It plays an essential role in disease diagnosis, improves patient care, and helps medical practitioners with the treatment type. Detection, assessment and adequate fracture care are considered important as a misdiagnosis frequently contributes to inadequate medical intervention, heightened frustration and expensive lawsuits, as described by Umadevi et al. [1]. In medical image processing, bone is the structural framework of the body called a skeleton, which protects the internal organs such as the ribs, making a shield around the heart, liver, and lungs, as shown in Fig. (1). The skull protects the human brain, and the *
Corresponding author Muhammad Imad: Abasyn University, Peshawar, Pakistan; E-mail: [email protected] Mariya Ouaissa, Mariyam Ouaissa, Zakaria Boulouad, Inam Ullah Khan, Sailesh Iyer (Eds.) All rights reserved-© 2023 Bentham Science Publishers
Machine Learning Solution for Orthopedics
Computational Intelligence for Data Analysis, Vol. 2 121
spine protects the nerves in the human body's spinal column. Bone, as mentioned by Abubacker et al. [2], provides the environment for bone marrow for blood cell creation and acts as a storage area for minerals and calcium.
Fig. (1). Bones in human body.
At birth, there are around 270 soft bones which are reduced to 206 in adulthood.
122 Computational Intelligence for Data Analysis, Vol. 2 .
Imad et al.
The largest bone in the human body is the thighbone or femur bone, and the smallest is the stapes in the middle ear, which is 3mm long. As shown in Fig. (2), the internal structure of bone provides strength to the body, as described by Vijayakumar et al. [3].
Fig. (2). Internal structure of bone.
In the human body, a fracture is the main ailment of human bone due to external force and crack. The images are further analyzed using medical image processing to identify the bone fracture. According to Al-Ayyoub et al. [4], medical imaging can automatically diagnose the image using an advanced computer. Human organs in digital form, as described by Imad et al. [5], are created by using different equipment, such as X-ray, Mammography, MRI, Ultrasound, Endoscopy, Positron Emission, Tomography, Fluoroscopy, Clinical Imaging, and bone scanning. According to Irfan Ullah et al. [6], X-ray is the latest noninvasive, painless, and economical solution. An X-ray can make the bone image in the body, such as the wrist, knee, ankle, and leg. A crack, also known as a fracture, can be any bone in the body. The fracture may include the hip, wrist, ankle, spine, Jaw, and ribs. Fractures are classified as open (if the skin is damaged) or closed (if the skin is intact). A break to the bone that does not harm the surrounding tissue or rip through the skin is a closed fracture.
Machine Learning Solution for Orthopedics
Computational Intelligence for Data Analysis, Vol. 2 123
An open fracture, as mentioned by Mustra et al. [7] and Kazeminia et al. [8], describes a situation in which the bone penetrates the skin. As described in Fig. (3), many types of fractures can occur in a bone, which is defined by Cao et al. [9] as follows:
Fig. (3). Types of fractures in a bones.
●
● ● ●
Pathological fracture (spontaneous fracture) – A fracture of a bone due to weakening the bone, such as neoplasia osteomalacia. Longitudinal fracture – Extending along the length of the bone. Spiral fracture - A fracture where the line of fracture bone is twisted. Greenstick fracture – The bone fracture from one side, on the other side, is
124 Computational Intelligence for Data Analysis, Vol. 2
● ●
● ● ●
Imad et al.
bent. Commonly occurs in children. Simple fracture (closed fracture) - One that does not produce an open wound. Compound fracture (open fracture): The compound fracture defines the location of the bone penetrating the skin. Oblique fracture - A fracture that is in diagonal form. Comminuted fracture - The bone is splintered into more than two pieces. Transverse: The fracture along the right angle of the long bone.
LITERATURE REVIEW The researcher has worked on bone fractures and implemented numerous techniques for different types of bone fractures using X-ray and CT images. Dennis Wen-Hsiang Yap [10] presented a femur bone fracture detection method to analyze the trabecular texture by calculating the difference between the orientation map of fractured bone and the mean orientation map of healthy bone samples. The orientation map of the femur bone's upper extremity is extracted and changed into the scalar map of difference. The Bayesian and SVM classifier is used to categorise the variance map. Martin Donnelley [11] presents a method for femur bone fracture in the X-ray image, consisting of several steps. In the first steps, the non-linear anisotropic diffusion method extracts the edge and smooths the image without losing any critical information from the boundary location. The second is to modify the Hough transformation to evaluate parameters for straight lines that better approximate the long bones' edges. The gradient magnitude and direction are created for the line parameters and illustrate irregular areas such as fractures. Vineta Lai Fun Lum [12] analyzes different methods for identifying bone fractures in an X-ray image. Gabor orientation, Markov random field, and gradient intensity direction are used for texture extraction. A different patient has different fractures in shape and size. The apps were sampled using an adaptive sampling method to accommodate these variations. This approach created feature maps of the same scale for the same bone form and feature type. Test results indicate that the usefulness of a procedure to improve both precision and sensitivity relies on both the quality of the process and the percentage of incorrect samples in the study collection. Jia [13] introduces a form of segmentation that describes fractured bones within the casting materials in an X-ray picture of a patient's arm and reveals the connection between the fractured bones. This segmentation function is often complicated due to low contrast and high noise ratio in X-ray images generated by the casting materials. A geodesic model of active contour with global constraints
Machine Learning Solution for Orthopedics
Computational Intelligence for Data Analysis, Vol. 2 125
is implemented for this function. A prior form is gathered and used as our model's regional limit. Victor Ortiz [14] presents a new deformable model based on physics to track physical deformations (fractures) using image matching. The model results from minimizing a field of deformation that simultaneously meets the constraints of an elastic body and a measure of local image similarity. The model gives us a scientifically accurate deformation area and helps us analyze the deformed material characteristics. This can be very helpful when examining pressures induced by the deformation of certain objects in their setting, in our case, the identification of skull fractures. Xuejing Jin [15] presents a new method for medical processing to express the different ranges of fractal-dimensional variations based on the multiscale fractal feature (DMF). The DMF values in different regions are usually different in a medical image. According to D MF values, small lesions can be detected in a medical X-ray image. A digital X-ray fracture is applied to this process. The experimental results specify that the background regions have D MF values of 0.005-0.020, the normal regions have D MF values of 0.040-0.070, and the knot regions have D MF values ranging between 0.100 and 0.200. Ravia Shabnam Parveen [16] compared two methods in this article, the Histogram Equalization and Contrast Limited Adaptive Histogram Equalization. Histogram equalization assigns pixel intensity values in the input image to provide a uniform distribution of intensities in the output image. CLAHE operates in the image known as tiles in small regions. The Adaptive Histogram Equalization function improves the contrast. The bilinear interpolation method is used to remove the inserted boundaries. CLAHE provide the same performance as the Adaptive Contrast Enhancement technique and overcome the limitation of HE. The result shows that the methodology of equalization of CLAHE works better than the process of equalization of Histogram. Zheng Wei [17] proposed an algorithm for fracture bone fracture extraction in an X-ray image, using marker-controlled watershed transform to segment X-ray fracture images dependent on gradient modulation and homotopy alteration. The characteristics of area number, region area, region centre, and a protuberant fracture image polygon are extracted by marker processing and area function. First, the Hough transform in the protuberant polygon of the X-ray fracture image identifies and removes shapes. The lines consist of line fractures and centerline to parallel lines. Finally, measure the distance between the line of separation and the axis of the centerline perpendicular. Song Tie-Rui [18] presented several descriptive image enhancement algorithms:
126 Computational Intelligence for Data Analysis, Vol. 2
Imad et al.
contrast enhancement, histogram enhancement, smoothing filter, median filter, ideal low-pass filter, low-filter butter, ideal high-pass filter, high-filter butter and homomorphic filter. Finally, the improved images were compared with the results, using various segmentation threshold and edge extraction methods to identify appropriate methods for enhancing X-ray image algorithms. Venkatesh Mahadevan [19] derived the correct proximal femur density by the design of a computer-aided detection system for 50 Indian pre and postmenopausal women from 2-dimensional visual, radiographic hip pictures, contrasting these findings with proximal femoral BMD values calculated by the Dual Energy Radiation Absorptiometry (DXA) for the same women. This theoretical method explored how BMD and anthropometric variables such as height and body weight have been related. The trabecular density technique can be helpful in clinical practice and research to better address the risk of osteoporotic fracturing. DXA's results show that osteoporosis and osteopenia were 20% and 34%, respectively, of Indian women. Hum Yan Chai [20] proposed an automated fracture detection method for femur bone using GLCM techniques. The method distinguishes the bone fracture loss and appearance based on the parameter value of GLCM. The criterion for bone fracture absence and appearance is set at 0.95. A total of 86.67% of the precision of the developmental algorithm appears to be an effective method for the automated detection of bone fractures. The findings indicate that the program can produce reliable and reproducible results. Greg Wood's [21] aim of the research was to better understand the processes underlying perceptual expertise in interpreting skeletal radiographs. Thirty participants were tested and examined for the clinical cases of different patients' normal and abnormal skeletal X-rays. The result for the presented techniques shows the technique achieved the accurate result for the detection of skeletal. Umadevi [1] considered an application of ensemble classification in X-ray images to fracture detection. For X-ray pictures, a minimum of 12 features have been taken from two groups of objects, namely texture features and form characteristics. Three binary classification models, namely SVM, BPNN, and KNN, were built to ensemble classification models. The research addresses ensemble structures with two classifications and three classifications. During the different feature sets, classifiers were built. Nirase Fathima Abubacker [22] focuses on automated normal and abnormal skull images. The skull fracture throughout Digital Imaging and Communications in Medicine (DlCOM) is observed in a simple automated method to remove the skull bone with histogram-based thresholds and connect the neighbouring pixel to
Machine Learning Solution for Orthopedics
Computational Intelligence for Data Analysis, Vol. 2 127
identify the fracture. The experimental results show that the object is identified with a high detection rate. Mahmoud Al-Ayyoub [4] addresses the issue of finding X-ray fractures in the hand bone. Numerous methods for the pre-processing step of the image are tested. Also, various sets of features are computed and tested. Finally, both foundation and multi-level meta-classifiers are being tested. By implementing boosting and then bagging on the Bayesian Network classifiers, an accuracy level of 91.8% is obtained where the feature set includes features computed using Wavelets, Curvelets and GLCM. Vijayakumar [3] proposed a method for automatically detecting femur bone fractures using computerized GLCM techniques. From this approach, the existence of bone fracture depends on the GLCM value of the parameter extracted. The accuracy of the built algorithm is at least 87% achieved, which promises an efficient method for automatically identifying bone fractures. Mario Mustra [7] presented an automated CR bone alignment system that shows forearms and legs. The proposed method uses thresholding for optimal threshold determination based on Otsu's system and Hough transform for the identification of straight-line segments from which the angle of rotation can be determined. This technique is not oriented to correct bone segmentation but rather to locating the borders of the bone so that it can determine the direction and angles of rotation of the bone. Hough transforms and the correct neighborhood within the Hough accumulator after defining the various levels of accuracy in angle estimation according to the angular resolution. Kazeminia [8] presents a new bone segmentation technique that pre-processes the image, such as noise cancellation and edge detection. The picture is denoising using a guided filter; then, the bone boundary is detected using a multi-and singlethreshold canny edge detector. Rotate the image in vertical directions on the prominent edges to improve the borders. Then, each image row checks for strength peaks that match the bone edges. Yu Cao [9] presents the generalized detection method for bone fractures, common throughout the body to various bone fracture forms and structures. In a novel descriptive computing system named “Stacking Random Forest Fusion,” the method uses features derived from the nominee patches of X-ray pictures. Random forest learners use this method at a lower level to extract the simplified distribution labels at the next step in the class likelihood labeling.
128 Computational Intelligence for Data Analysis, Vol. 2
Imad et al.
San Myint [23] aims to identify bone fractures in human lower legs from X-ray images. The method proposed has three steps: pre-processing, segmentation, and detection of fractures. In the pre-processing step removing noise, image brightness adjustment, and colour adjustment are the main procedures for image enhancement using a Gaussian filter. The feature extraction uses the technique of Hough transform to detect lines in the image. Ling Wang [24] proposed per trochanteric fractures, an automated and effective classification method based on image segmentation techniques. The difference between fracture parts is first defined per trochanteric femoral fracture types. In the 3D femur, there is 4 direction to reduce the computational complexity. The level set process is to segment the fracture from the context. After segmentation, the Canny edge detector extracts the edge and differentiates between the normal and testing images. Ayumi Yamada [25] develops a method for detecting skull fractures using CT images. The image is preprocessed using the morphological operation to extract the bone region. From the image of the CT Scout, the cranial vault can be determined. The cranium has a low-density cancellous bone between the small, two-layer high-density bones. Although compact bone fracture lines are more easily identified than cellular bones, three-dimensional (3D) Laplacian filtering is then removed from the bone surface. The application of the black-hat transform removes linear structures from the bone surface image. Tanudeep Kaur [26] represents bone X-ray fracture detection using segmentation, Fuzzy c-means and Multilevel Wavelet algorithms. The fuzzy c-mean segmentation, the morphological operator, has been used for bone fracture detection. The wavelet approach is used to find the region area of detection and detect the fracture. Ying Liang Ma [27] presents an automatic method for the detection of imagebased tracking. The geometric model is used, which is based on matching the Xray image between the external tracking system and the image overlay system with a critical coordinate system. The picture overlay will change the bone fragment's location and orientation, which is controlled by the fracture robot in real-time. Ankur Mani Tripathi [28] presents image processing techniques to find the fracture's location in the femur bone. The morphological operation and edge detection has been used to suppress the background and illuminate the foreground. The support vector machine (SVM) has been used to differentiate between the fracture of the femur bone and the non-fracture femur bone.
Machine Learning Solution for Orthopedics
Computational Intelligence for Data Analysis, Vol. 2 129
Mariam M. Saii [29] presents a novel method for detecting and classifying fractures. The basic stages of fracture detection include pre-processing bone images and morphological operations to obtain the ROI region manipulated to remove non-fracture pixels through a post-processing phase. The suggested approach extracts three features from the bone image: transverse, cracks, and divergence to define the type of fracture or bone image integrity. The designed systems correctly detect the different types of fractures alongside the type of hybrid fracture. Wint Wah Myint [30] fracture type and recognition in the lower leg bone (Tibia) are developed using different image processing techniques. This work aims to detect fracture or non-fracture and classify the lower leg bone (tibia) fracture type in the X-ray image. The detection system for tibia bone fractures is developed with three main steps. They are pre-processing extraction features and classification for classifying fracture types and locating fracture locations. Unsharp Masking (USM), the sharpening method, is applied in pre-processing to improve the image and highlight the edges in the image. Harris corner detection algorithm then processes the sharpened image to extract feature points at the corner to extract the feature. And then, two methods of classification are chosen to identify fracture or non-fracture, and classify forms of fracture. Simple Decision Tree (DT) is used for fracture or not classification, and K-Nearest Neighbor (KNN) is used to classify fracture. Eveling Castro-Gutierrez [31] discussed acetabulum fractures that affect people aged 21 to 30 and proposed a method for Acetabulum Fractures as a Local Binary Pattern (LBP) for feature extractors and the Support Vector Machine (SVM) for classification. Table 1. Bone fracture detection in X-ray images. Refs.
Work On
[10]
Femur Fractures
Methodology
Result
Advantage
Disadvantage
Gaussian filter, Based on a neck-shaft angle, SVM and Bayesian has Bayesian and the training and testing image Bayesian less detection Support Vector was kept at 12%, where 39 accurately detect rate .1 of 3 Machine, Gabor femur fractures were for and SVM methods highest filters training and 13 fractured for provides the detection but testing highest accuracy low accuracy Accuracy: N/A among the three- and a high false technique alarm rate. In 2 of 3 methods, accuracy is the high but low detection rate
130 Computational Intelligence for Data Analysis, Vol. 2
Imad et al.
(Table 1) cont.....
Refs.
Work On
Methodology
Result
Advantage
Disadvantage
[32]
Long Bone Fracture Detection
AMSS anisotropic diffusion, HOG Hough Transform,
The method accurately detects the mid-shaft long bone fracture Accuracy: N/A
SVM provides the highest accuracy in fracture detection among other methods like NSA
N/A
[12]
Bone Fracture Detection
Comprehensive 432 Cases Of Femur and The OR rule In the femur between different Radius.12% Femur bone and provides a better and radius classifiers, 30% Radius image. In femur result in both bone, the MRF Gini-SVM, bone, For Training, GO is accuracy and sensitivity is Bayesian method 100.0% MRF is 99.3% and sensitivity, high in the IGD is 100.0%. For Testing which is useful training and GO is 89.8%, MRF is 98.1%, in a standard testing set. and IGD is 90.7%. Inradius classifier. training set provides 100% accuracy, and testing is GO 90.5% MRF is 86.5%, and IGD is 96.0%.
[13]
Bone Fracture Detection
Model-Based Segmentation Method, Distance Transform, Tophat Transform
[14]
Bone Fracture Detection
Gaussian filter, Detected skull bone fracture Canny edge Accuracy: N/A detection, Method of the Finite Element.
[15]
[16]
Efficiently detect and segmentation of bone fracture Accuracy: N/A
Detect bone fracture outline from the lowcontrast image
Segmentation failed in one due to the poor contrast of the image.
Help the Physicians To Detect The Fractures From The Magnetic Resonance
More improvement is needed for better result
Bone Fracture Detection
Histogram The DMF value is between Guided the Equalization, 0.005 to 0.00, and the normal Physicians Spatial Filtering, region is 0.040 to 0.070. the In 2D X-ray Segmentation, knot region is 0.100 to 0.200 image in a realPrewitt Detector, Accuracy: N/A time Log Detector, environment. And Sobel Detector
The bone fracture region is not clear due to low quality and insufficient Clarity.
Bone Fracture Detection
Mean Filters, Histogram Equalization and contrast limited of adaptive histogram Equalization
Increased the noise in the image in the background during contrast
CLAHE provides better performance than HE. The original image pixels are 483493, and the HE Is 984185, and the CLAHE IS 634938 Accuracy: N/A
It will help in the noisy image to provide a better result
Machine Learning Solution for Orthopedics
Computational Intelligence for Data Analysis, Vol. 2 131
(Table 1) cont.....
Refs.
Work On
Methodology
Result
Advantage
Disadvantage
[17] Bone fracture detection and classification
Mark controlled watershed algorithm with homotopy modification, marker processing and Region props function, hough transform
The feature of regions has been correctly extracted Accuracy: N/A
N/A
Not very good in complex fractures like line fractured
[18]
Image enhanced using a different method
Contrast The homomorphic filter enhancement, HE, provides a better result than Filtering other filter techniques techniques like Accuracy: N/A median, low pass filter, high pass filter, and homomorphic filter
[19]
Detection of Osteoporosis in the Hip Region
Histogram specification, standard deviation, standard deviation, Gabor and wavelet (4level decomposition) operation, SPSS,
[20]
Detection of long-bone fractures.
Binary From equations 13 to 16, the Conversion, sensitivity rate is 80%, Gaussian specificity is 93.31%. operator, Accuracy: 86.67 Laplacian edge detector, Median filter and k-mean clusterering, GLCM method,
20% to 34% of Indian women have osteoporosis and osteopenia problem Accuracy: N/A
Several filtering Affect during techniques are the selection of compared in a a large extent of better way image enhancement
N/A
The osteoporosis and osteopenia disease risk factors not showed
N/A
N/A
132 Computational Intelligence for Data Analysis, Vol. 2
Imad et al.
(Table 1) cont.....
Refs.
Work On
Methodology
Result
Advantage
Disadvantage
[24]
The skeletal fracture detecting
Linear regression analysis, 30 participants with ten novices, intermediates and experts with clinical cases of normal and abnormal skeletal radiography while wearing eyetracking equipment
Accuracy: 34% of the variance in variable
Less time taking to diagnose the image of skeletal fracture
N/A
[25]
Leg bone fracture detection
Gaussian filter, Sobel, Prewitt, Robert and Canny edge detection, Hough transform technique,
Accuracy: N/A The system detects the fracture accurately
Canny edge detection provides a better result
The detection method affects the quality of the image
[26]
Detect bone fracture in 3D CT images
Distance regularization, canny edge detection, Gaussian filter with threshold values and level set method
Accuracy:84.2% recognition rate
The bone fracture is detected in 4 directions (anterior, anterolateral posterior, posterolateral)
Error in trochanter fracture, which is not diagnosed by the human eye also
[33]
Skin Cancer
Naïve Bayes, Logistic Regression, SVM, Random Forest, KNN
Accuracy: 92%,92%,89.6%,83%,50%
NB and LR Accuracy needs provide a better to be improved result than other results
[34]
Bullet detection
SVM, KNN, Decision Tree, Naïve Bayes,
Accuracy: 85%, 66.4%, 73%, 61%
SVM provides a Accuracy needs better result to be improved
[35, 36]
COVID-19
SVM, KNN, Random Forest, Naïve Bayes, Decision Tree, CNN, AlexNet, VGG
Accuracy: 96%, 92%, 90%, SVM, CNN, 82% VGG16 and CNN=95%, VGG16=94% AlexNet provide AlexNet=94% a better result
Accuracy of decision tree needs to be improved
Machine Learning Solution for Orthopedics
Computational Intelligence for Data Analysis, Vol. 2 133
(Table 1) cont.....
Refs.
Work On
Methodology
Result
Advantage
Disadvantage
[37]
COVID-19
Survey paper
Discussed Different Techniques
Discussed Ml, Deep Learning and IOT for COVID-19 prevention and control also discussed the future direction
NA
METHODOLOGY This section lists different methods applied to X-ray and CT images diagnosed bone fracture and discusses the corresponding paper’s results, advantages, and disadvantages. Table 1 presents short technical details of each paper, which help the researchers in this field. CONCLUSION This paper discussed the various methods applied to different bone fractures of the human body, as illustrated in Fig. (1). Bone fracture detection is a common problem. Even in such developed countries, it increased rapidly. However, the automatic detection of bone fractures is an important but complex task where limited algorithms are proposed. According to Jin et al. [15], different modalities were used to diagnose bone fractures, such as X-ray, CT, MRI and Ultrasound. Many imaging techniques have been studied in these articles to diagnose fractures in bone, such as pre-processing, image segmentation, Edge detection and classification. Based on the study conducted, Lum et al. [12] provide a better result of 93% in all other techniques. Machine learning techniques open many perspectives for radiologists and orthopaedics, saving time and assistance in diagnosing. For a clinician, such algorithms help prevent errors and improve the performance of diagnosing and detection. CONSENT FOR PUBLICATION Not applicable. CONFLICT OF INTEREST The author declares no conflict of interest, financial or otherwise.
134 Computational Intelligence for Data Analysis, Vol. 2
Imad et al.
ACKNOWLEDGEMENT Declared none. REFERENCES [1]
N. Umadevi, and S. Geethalakshmi, "Multiple classification system for fracture detection in human bone X-ray images", 2012 Third International Conference on Computing, Communication and Networking Technologies (ICCCNT’12), pp. 1-8, 2012. [http://dx.doi.org/10.1109/ICCCNT.2012.6395889]
[2]
N.F. Abubacker, A. Azman, M. Azrifah, and S. Doraisamy, "An approach for an automatic fracture detection of skull dicom images based on neighboring pixels," in 2013 13th International Conference on Intellient Systems Design and Applications, pp. 177-181, 2013. [http://dx.doi.org/10.1109/ISDA.2013.6920731]
[3]
R. Vijayakumar and G. Gireesh, "Quantitative analysis and fracture detection of pelvic bone X-ray images," in 2013 fourth international conference on computing, communications and networking technologies (ICCCNT), pp. 1-7, 2013. [http://dx.doi.org/10.1109/ICCCNT.2013.6726590]
[4]
M. Al-Ayyoub, I. Hmeidi, and H. Rababah, "Detecting Hand Bone Fractures in X-ray Images", J. Multim. Process. Technol., vol. 4, no. 3, pp. 155-168, 2013.
[5]
M. Imad, F. Ullah, and M.A. Hassan, "Pakistani Currency Recognition to Assist Blind Person Based on Convolutional Neural Network", Journal of Computer Science and Technology Studies, vol. 2, no. 2, pp. 12-19, 2020.
[6]
S.I. Ullah, A. Salam, W. Ullah, and M. Imad, "COVID-19 Lung Image Classification Based on Logistic Regression and Support Vector Machine", North African Conference on Management & Information Systems, pp. 13-23, 2021.
[7]
M. Mustra, M. Grgic, and B. Zovko-Cihlar, "Alignment of X-ray bone images," in 2014 X International Symposium on Telecommunications (BIHTEL), pp. 1-4, 2014. [http://dx.doi.org/10.1109/BIHTEL.2014.6987650]
[8]
S. Kazeminia, N. Karimi, B. Mirmahboub, S.M.R. Soroushmehr, S. Samavi, and K. Najarian, "Bone extraction in X-ray images by analysis of line fluctuations", 2015 IEEE International Conference on Image Processing (ICIP), pp. 882-886, 2015. [http://dx.doi.org/10.1109/ICIP.2015.7350926]
[9]
Y. Cao, H. Wang, M. Moradi, P. Prasanna, and T. F. Syeda-Mahmood, "Fracture detection in X-ray images through stacked random forests feature fusion," in 2015 IEEE 12th international symposium on biomedical imaging (ISBI), pp. 801-805, 2015. [http://dx.doi.org/10.1109/ISBI.2015.7163993]
[10]
D.W-H. Yap, Y. Chen, W.K. Leow, T.S. Howe, and M.A. Png, "Detecting femur fractures by texture analysis of trabeculae", Proceedings of the 17th International Conference on Pattern Recognition, vol. 3, pp. 730-733, 2004.
[11]
M. Donnelley, and G. Knowles, "Computer aided long bone fracture detection", Proceedings of the Eighth International Symposium on Signal Processing and Its Applications, vol. 1, pp. 175-178, 2005. [http://dx.doi.org/10.1109/ISSPA.2005.1580224]
[12]
V.L.F. Lum, W.K. Leow, Y. Chen, T.S. Howe, and M.A. Png, "Combining classifiers for bone fracture detection in X-ray images", IEEE International Conference on Image Processing 2005, vol. 1, p. I-1149, 2005.
[13]
Y. Jia, and Y. Jiang, "Active contour model with shape constraints for bone fracture detection", International Conference on Computer Graphics, Imaging and Visualisation (CGIV’06), pp. 90-95, 2006.
Machine Learning Solution for Orthopedics
Computational Intelligence for Data Analysis, Vol. 2 135
[http://dx.doi.org/10.1109/CGIV.2006.16] [14]
V. Ortiz, C. Yanez, A. Kuri, and I. Chairez, "Skull fractures detection by finite element method", Robotics and Automotive Mechanics Conference (CERMA 2007), pp. 271-276, 2007. [http://dx.doi.org/10.1109/CERMA.2007.4367698]
[15]
X. Jin, D. Qi, H. Wu, and F. Liu, "Detection of small lesions in medical X-ray image based on multiscale fractal feature," in 2009 2nd International Conference on Biomedical Engineering and Informatics, pp. 1-4, 2009.
[16]
N.R.S. Parveen, and M.M. Sathik, "Enhancement of bone fracture images by equalization methods", 2009 International Conference on Computer Technology and Development, vol. 2, pp. 391-394, 2009. [http://dx.doi.org/10.1109/ICCTD.2009.115]
[17]
Z. Wei, M. Na, S. Huisheng, and F. Hongqi, "Feature extraction of X-ray fracture image and fracture classification", 2009 International Conference on Artificial Intelligence and Computational Intelligence, vol. 2, pp. 408-412, 2009. [http://dx.doi.org/10.1109/AICI.2009.40]
[18]
S. Tie-Rui, and Z. Wei, "The research of X-ray bone fracture image enhancement algorithms", 2010 International Conference on Computer, Mechatronics, Control and Electronic Engineering, vol. 5, pp. 384-387, 2010.
[19]
V. Mahadevan, V. Sapthagirivasan, Management, and Technology, "Information processing of medical images for the detection of osteoporosis in hip region of interest," Int J Innov, Manage and Technol, vol. 1, no. 5, pp. 516, 2010.
[20]
H. Y. Chai, L. K. Wee, T. T. Swee, S.-H. Salleh, and A.K. Ariff, "Gray-level co-occurrence matrix bone fracture detection," Am. J. Appl. Sci., vol. 8, no. 1, p. 26, 2011. [http://dx.doi.org/10.3844/ajassp.2011.26.32]
[21]
G. Wood, K. M. Knapp, B. Rock, C. Cousens, C. Roobottom, and M. R. J. S. r. Wilson, "Visual expertise in detecting and diagnosing skeletal fractures," Skeletal Radiol., vol. 42, no. 2, pp. 165-172, 2013. [http://dx.doi.org/10.1007/s00256-012-1503-5]
[22]
N. F. Abubacker, A. Azman, M. Azrifah, and S. Doraisamy, "An approach for an automatic fracture detection of skull dicom images based on neighboring pixels," in 2013 13th International Conference on Intellient Systems Design and Applications, pp. 177-181, 2013. [http://dx.doi.org/10.1109/ISDA.2013.6920731]
[23]
S. Myint, A.S. Khaing, H.M. Tun, "Detecting Leg Bone Fracture In X-Ray Images", Int. J. Scientific Technol. Res., Vol. 5, No. 06, pp. 140-144, 2016.
[24]
L. Wang, H. Cheng, H. Lan, Y. Zheng, and K. Li, "Automatic recognition of per trochanteric bone fractures in femur using level sets," in 2016 38th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), 2016, pp. 3851-3854.
[25]
A. Yamada, A. Teramoto, T. Otsuka, K. Kudo, H. Anno, and H. Fujita, "Preliminary study on the automated skull fracture detection in CT images using black-hat transform," in 2016 38th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), pp. 6437-6440, 2016. [http://dx.doi.org/10.1109/EMBC.2016.7592202]
[26]
T. Kaur, A. Garg"Bone Fraction Detection using Image Segmentation", International Journal of Engineering Trends and Technology (IJETT), Vol. 36, No. 2, pp. 82-87, 2016.
[27]
Y. Ma, G. Dagnino, I. Georgilas, and S. Dogramadzi, "Automatic Tool Detection in X-ray Images for Robotic Assisted Joint Fracture Surgery," in 2017 IEEE International Conference on Internet of Things (iThings) and IEEE Green Computing and Communications (GreenCom) and IEEE Cyber, Physical and Social Computing (CPSCom) and IEEE Smart Data (SmartData), pp. 883-887, 2017. [http://dx.doi.org/10.1109/iThings-GreenCom-CPSCom-SmartData.2017.136]
136 Computational Intelligence for Data Analysis, Vol. 2
Imad et al.
[28]
A.M. Tripathi, A. Upadhyay, A.S. Rajput, A.P. Singh, and B. Kumar, "Automatic detection of fracture in femur bones using image processing", 2017 International Conference on Innovations in Information, Embedded and Communication Systems (ICIIECS), pp. 1-5, 2017. [http://dx.doi.org/10.1109/ICIIECS.2017.8275843]
[29]
M. M. Saii and A. M. Mayya, "Hand and Finger Fracture Detection and Type Classification Using Image Processing." Int. J. Computer Sci. Trends Technol., (IJCST). vol. 6, no. 4, 2018.
[30]
W.W. Myint, K.S. Tun, and H.M Tun, "Analysis on Leg Bone Fracture Detection and Classification Using X-ray Images", Machine Learning Research, vol. 3, no. 3, pp. 49-59, 2018. [http://dx.doi.org/10.11648/j.mlr.20180303.11]
[31]
E. Castro-Gutierrez, L. Estacio-Cerquin, J. Gallegos-Guillen, and J.D. Obando, "Detection of Acetabulum Fractures Using X-ray Imaging and Processing Methods Focused on Noisy Images", 2019 Amity International Conference on Artificial Intelligence (AICAI), pp. 296-302, 2019. [http://dx.doi.org/10.1109/AICAI.2019.8701297]
[32]
M. Donnelley and G. Knowles, "Computer aided long bone fracture detection," in ISSPA, pp. 175178, 2005. [http://dx.doi.org/10.1109/ISSPA.2005.1580224]
[33]
A. Salam, F. Ullah, M. Imad, and M. A. Hassan, "Diagnosing of Dermoscopic Images using Machine Learning approaches for Melanoma Detection," in 2020 IEEE 23rd International Multitopic Conference (INMIC), pp. 1-5, 2020.
[34]
M. Imad, S.I. Ullah, A. Salam, W.U. Khan, F. Ullah, and M.A. Hassan, "Automatic Detection of Bullet in Human Body Based on X-ray Images Using Machine Learning Techniques", Int. J. Comput. Sci. Inf. Secur., vol. 18, no. 6, 2020.
[35]
M. Imad, N. Khan, F. Ullah, M. A. Hassan, and A. Hussain, "COVID-19 classification based on Chest X-ray images using machine learning techniques," Journal of Computer Science and Technology Studies, vol. 2, no. 2, pp. 01-11, 2020.
[36]
A. Hussain, M. Imad, A. Khan, and B. Ullah, Multi-class Classification for the Identification of COVID-19 in X-ray Images Using Customized Efficient Neural Network. AI and IoT for Sustainable Development in Emerging Countries, 2022, pp. 473-486. [http://dx.doi.org/10.1007/978-3-030-90618-4_23]
[37]
M. Imad, A. Hussain, M. Hassan, Z. Butt, and N. Sahar, IoT Based Machine Learning and Deep Learning Platform for COVID-19 Prevention and Control: A Systematic Review. AI and IoT for Sustainable Development in Emerging Countries, 2022, pp. 523-536. [http://dx.doi.org/10.1007/978-3-030-90618-4_26]
Computational Intelligence for Data Analysis, 2023, Vol. 2, 137-148
137
CHAPTER 9
A Review of Machine Learning Approaches for Identification of Health-Related Diseases Muhammad Yaseen Ayub1,*, Farman Ali Khan1, Syeda Zillay Nain Zukhraf2 and Muhammad Hamza Akhlaq3 COMSATS University Islamabad, Attock Campus, Pakistan KIOS Research and Innovation Center of Excellence, Department of Electrical and Computer Engineering, University of Cyprus, Nicosia, Cyprus 3 Allama Iqbal Open University, Islamabad, Pakistan 1 2
Abstract: The field of medicine is one of the most respected and oldest professions in human history. It has a direct impact on human life. The main purpose of this chapter is to present a brief introduction to the use of advanced computer science technologies like machine learning in the process of disease detection. The chapter also discusses different machine learning algorithms which are used in the process of disease detection. It also points out which algorithms give better accuracy. This chapter lists all major and most commonly used machine learning libraries to detect various lifethreatening diseases. Lastly, a discussion on the future trends of technology which can be used in disease detection, and viral disease control is presented.
Keywords: Disease Detection, IoMT, Machine Learning, Machine Learning in IoMT. INTRODUCTION Currently, computational technologies are applied almost in every field of life. The medical field is considered to be one of them. During the past few years, a lot of research has been conducted in this domain to identify and treat diseases early on using machine learning approaches. Since machine learning can perform human cognitive functions effectively, thus it becomes a more suitable approach in the detection and treatment of disease. Recently, there has been lots of research done for the development of specialized methods in machine learning which can be highly accurate in the detection of a variety of diseases. However, we still struggle to detect and diagnose several life-threatening diseases because of the highly heterogeneous environment [1]. Several life-threatening diseases can be Corresponding author Muhammad Yaseen Ayub: COMSATS University Islamabad, Attock Campus, Pakistan; Email: [email protected] *
Mariya Ouaissa, Mariyam Ouaissa, Zakaria Boulouad, Inam Ullah Khan, Sailesh Iyer (Eds.) All rights reserved-© 2023 Bentham Science Publishers
138 Computational Intelligence for Data Analysis, Vol. 2
Ayub et al.
treated if they are detected at early stages, as shown in Fig. (1). Thus early detection of disease is essential.
Fig. (1). Early detection can save lives from fatal diseases.
In the past few decades, a lot of research has been conducted in the development of Machine learning algorithms that can perform human cognitive tasks. The medical field has greatly benefited from it; researchers developed machine learning models which can accurately detect or predict diseases like tumors, cysts, pneumonia, etc. They use MRI and CT Scan X-ray images coupled with computer vision technologies to detect or predict high-risk diseases at a very early stage. Machine learning algorithms also work on numeric and Boolean values obtained through different types of sensors [2]. Through the use of machine learning, computers can mimic human cognitive behavior and recognize different patterns. Computers can then effectively detect any abnormalities in those patterns and classify whether the patient has a specific disease. Machine learning has two major approaches: Supervised Learning and Unsupervised Learning (there also exists a hybrid of these two, semi-supervised learning). Fig. (2) shows a pictorial representation of different types of Machine learning techniques. Supervised Learning Supervised machine learning provides a labeled dataset during the training phase. The model learns to map an input to an output based on the examples provided during the training period. Examples of supervised learning are classification and logistic regression. Classification algorithms are used to assign a distinct label based on learning performed on test data. For example, a classification algorithm
Health-Related Diseases
Computational Intelligence for Data Analysis, Vol. 2 139
can differently classify a dog and a cat. There are various classification algorithms like Naïve Bayes, KNN, Decision Tree, Support Vector Machines, etc. [3]. Regression algorithms are used to estimate the relationship between the dependent and independent variables. They are used to predict numeric values. For example, many companies use linear regression models to predict costs, profit, loss, etc. There are several types of regression models: Linear Regression, Logistic Regression, Ridge Regression, Polynomial Regression, etc. [4]. Unsupervised Learning Unsupervised learning is a type of learning labeled dataset that is not provided to the computer. Instead, the computer finds different patterns on its own. It is based on pattern recognition without being provided with any target attribute, like in supervised learning. It finds hidden patterns and insights in data. It can be used when we do not have any labeled data for training and want to get useful insights and patterns in that data. Unsupervised learning’s example is clustering, where the model group data based on similarities. There are different types of clustering algorithms like Partition based clustering, Fuzzy clustering algorithms, Modelbased clustering algorithms, Density-based Clustering Algorithms, K-Means Clustering, etc. [5].
Fig. (2). Different machine learning algorithms and their uses.
140 Computational Intelligence for Data Analysis, Vol. 2
Ayub et al.
This book chapter is organized as follows, upcoming sections in the literature study present details about various life-threatening diseases. A literature study is followed by a section that discusses various algorithms exploited for disease detection. It lists down algorithms used in the detection of particular diseases. It is then followed by a section that shows various different tools and libraries used for disease detection. The last section provides a conclusion and future trends in disease detection. MOTIVATION Medical science directly impacts life. As the human population is growing at an explosive rate, we face a critical issue of resource limitation. To make disease detection easy, quick and affordable, we need to employ the latest artificial intelligence and machine learning algorithms in our current healthcare system. Recent research has shown that machine learning algorithms can detect diseases quickly and accurately. Some high-mortality rate diseases, like heart-related diseases, need an early diagnosis [6]. LITERATURE STUDY In the past two decades, lots of research has been done in tuning existing algorithms to work for the healthcare environment and developing new algorithms that specialize in feature extraction and perdition of different types of diseases. Research shows that a wide variety of diseases can be detected using machine learning techniques, such as brain tumors, heart-related diseases, lung cancer, liver functioning, Blood cancer, etc. Detection of disease is divided into two groups based on the method used. Some diseases are better and easily detected through pictorial data like X-ray images, CT Scans, Ultrasound, MRI scans, etc., while some diseases are not so easily detected through pictorial data; for them, we use specific sensors which then give us numeric data, which is then fed to a computer system for analysis and detection of any abnormalities. Sometimes a hybrid combination of both techniques is used to detect certain diseases. Fig. (3) shows how a machine learning-based system’s works. Machine learning-based systems require an input sample passed to AI or ML system, which analyzes it and detects any abnormalities, then gives output. Heart Diseases Detection According to WHO, heart-related issues are the leading cause of death, with millions of deaths globally each year [7]. Researchers have used a wide variety of machine learning algorithms on heart-related abnormality detection, for example, ANN SVM Logistic Regression, kNN, Classification Tree Naïve Bayes, etc.
Health-Related Diseases
Computational Intelligence for Data Analysis, Vol. 2 141
Results show that Logistic Regression and ANN have the best accuracy in predating heart-related abnormalities [8]. Most of the time, abnormalities in heart beating patterns are detected using readings obtained from an ECG sensor. Fig. (4) shows a normal ECG graph.
Fig. (3). Steps involved in disease detection using machine learning.
I
a VR
V1
V4
I
a VL
V2
V5
III
a VF
V3
V6
I
LOC 00000 - 0000
Speed : 25 mm / sec
Limb : 10 mm / mV
Chest : 10 mm / mV
Fig. (4). Shows a normal electrocardiogram (ECG).
50~ 0 . 15 - 150 Hz
16405
142 Computational Intelligence for Data Analysis, Vol. 2
Ayub et al.
Lung Diseases Detection Lung cancer is another major factor contributing to approximately 1.7 million deaths annually across the globe [9]. Lung cancer can be identified through X-ray and CT scan images of a patient’s chest. After some preprocessing on the image, feature extraction techniques are applied to it, like Contrast Limited Adaptive Histogram Equalization (CLAHE). CLAHE uses a gray-level co-occurrence matrix (GLCM) to extract features from the image. After feature extraction, image classification is performed, which is then followed by Performance Evaluation. Fig. (5) shows a pictorial representation of these steps. Using SVM Classifier, 96% accuracy is achieved.
Fig. (5). Flow chart of a machine learning model.
A more recent and widespread example of lung infection is pneumonia which can also be caused by COVID-19. Pneumonia can also be detected from X-ray or CT scans of a patient’s chest. Recent research has shown up to 95% accuracy in detecting pneumonia from chest X-rays [10]. Skin Disease Detection Due to ozone layer decomposition, skin cancer and related issues are sharply increasing, especially in countries with high deforestation rates. Different algorithms are used to classify a few most common skin diseases Algorithms like ANN, KNN, and Decision Tree. SVM and Random Forest are used to detect skin
Health-Related Diseases
Computational Intelligence for Data Analysis, Vol. 2 143
diseases like Atopic Dermatitis, Folliculitis, Leprosy, Herpes Zoter, etc., with 90% accuracy [11]. Brain Diseases Detection The brain is considered the most important and complex organ of the human body. The brain is the most complex organ with millions of interconnected neurons, and those neurons keep changing daily, forming new connections. Due to this complex structure of the brain, treating brain diseases like brain tumors is very difficult. The earliest possible brain tumor detection is required to save patients from brain damage and other complications [12]. In 2016, it was reported that brain tumor was the leading cause of cancer-related deaths among children of age group (0-14) in the United States. A brain tumor is also ranked the third most common type of cancer among teenagers [13]. Other than brain tumors, various other psychiatric disorders can be detected using MRI images of the brain. Various other types of brain-related diseases like dementia and Alzheimer’s can be detected using MRI scan images. Alzheimer’s is described as losing human intellectual abilities partially due to brain damage. It is more common in elderly people. Following machine learning algorithms are used in the diagnosis of Alzheimer’s, like SVM, Random forest, KNN, Neural Networks, etc. [14]. . Liver Diseases Detection The liver is the largest organ in the human body weighing around 3 pounds. It is divided into two parts which are the left lobe and the right lobe. The basic function of the liver is to filter toxic elements from the body and release certain chemicals which assist in digestion. According to stats provided by World Health Organization in 2018, liver cancer is the 5th most common type of cancer in men [15]. Some of the common liver disorders are as follows [16]: ●
●
●
●
Fatty liver: Fatty liver is a condition where a large number of vacuoles of triglyceride fat accumulate in liver cells. There can be multiple causes of it, however, it is a recoverable condition. Hepatitis: Hepatitis is inflammation of the liver which is generally caused by the virus. It is a treatable condition, but further complications of this disease can be fatal. Cirrhosis: It is considered to be one of the most serious and common liver conditions. In cirrhosis where the loss of liver cells is observed. It can be caused by many types of liver diseases where the liver is damaged. Liver cell regeneration is slower than the loss of liver cells in cirrhosis. Liver cancer: Liver cancer is a serious condition where abnormal growth is observed. There can be many types of cancer in the liver however, the most
144 Computational Intelligence for Data Analysis, Vol. 2
Ayub et al.
common one is known as hepatocellular carcinoma (HCC). The risk of liver cancer is quite high in patients with cirrhosis or hepatitis. There are 10 parameters that are commonly used to predict liver-related diseases. The parameters are Age, Gender, Total Bilirubin, Direct Bilirubin, Alkine Phosphate, Alanine Aminotransferase, Aspartate Aminotransferase, Total Proteins, Albumin, Albumin and Globulin Ratio. Liver disease detection is done using supervised machine learning algorithms like KNN, Naïve Bayes, Random Forest, Support Vector Machine and Logistic Regression [17]. ALGORITHMS EXPLOITED FOR VARIOUS DISEASES DETECTION As more than one algorithm is available for detecting one disease, their accuracy varies a little. The main question arises of which algorithm to use for what type of problem. Generally, disease detection is done either by scanned images (CT, MRI, and X-ray) or through sensors like ECG, EEG, and Glucometers. Literature study shows that algorithms like Convolutional Neural Network (CNN), Recurrent Neural Network (RNN) and Deep Neural networks, etc., perform better on picture-based datasets. They are used in combination with different feature extraction methods. While algorithms like Random Forest, Decision tree, Extreme Gradient Boosting, Support Vector Machine, Logistic regression, etc., work best with numeric data. Fig. (6) shows that in pictorial form.
Fig. (6). Algorithms based on classification type.
Health-Related Diseases
Computational Intelligence for Data Analysis, Vol. 2 145
Following is a table containing some common types of diseases and different algorithms used in detecting them; Table 1 shows the four most affected organs worldwide and the algorithms used in detecting issues related to them. Different algorithms have different tasks; some algorithms are for clustering (grouping on the basis of similarities), while other algorithms could be classification algorithms (True or False). More than one algorithm can be used to detect a single disease. Table 1. Shows algorithms used for disease detection in cancer, heart, brain, liver and the associated task of those algorithms. Disease
Algorithm
Task
Cancer
K-means Clustering
Clustering
DBSCAN
Clustering
Random Forest
Regression
SVM
Classification
KNN
Classification
Linear Regression
Regression
SVM
Classification
ID3
Classification
KNN
Classification
Heart
Brain
Liver
Naive Bayes
Classification
K-means Clustering
Clustering
SOMS
Clustering
KNN
Classification
Naïve Bayes
Classification
SVM
Classification
SVM
Classification
ID3
Classification
Linear Regression
Regression
KNN
Classification
Naive Bayes
Classification
TOOLS AND LIBRARIES USED FOR DISEASE DETECTION There are a number of Machine Learning libraries available which have all common Machine Learning algorithms pre-coded. Those algorithms can be just used where ever needed. There are lots of libraries for different purposes, from
146 Computational Intelligence for Data Analysis, Vol. 2
Ayub et al.
feature extraction to classification and plotting. Table 2 shows different libraries and their task [18 - 20]. Table 2. Shows the most commonly used Machine learning libraries and frameworks. Library
Task
Tensorflow
Used to build and train machine learning models.
Scikit-learn
Data analysis, data mining, classification, clustering and regression.
PyTorch
Natural Language Processing and Neural Networks.
Keras
Deep Learning.
RapidMiner
Data Pre-processing, Data Preparation, Machine learning, Text mining, Predictive analysis.
NLTK
Natural Language Processing.
MATLAB
Whole Environment for Machine Learning, Data Analysis, Regression, Classification, Clustering and solving complex math problems.
Pandas
Data manipulation.
Numpy
Multi-Dimensional data handling.
Matplotlib
Plotting graphs, charts or Data visualization.
Seaborn
Data Visualization.
The table above contains different libraries and tools. Often more than one of these tools and libraries are used in combination to handle data, clean data (remove outliers), train models and visualize results. CONCLUSION AND FUTURE TRENDS From the above discussion, it can be concluded that the use of Machine Learning provides a great boost to medical science, and helps doctors and medical staff in the early detection of diseases. It also reduces the cost of medical equipment used. Many tasks are automated, which saves the precious time of doctors and also helps save some critical patients who need urgent medical aid. Different experiments show that ANN, CNN, SVM, Logistic Regression and Random forest are some of the Machine learning algorithms that tend to perform best for medical diagnosis. The future focus should be further enhancing the accuracy of these algorithms while focusing on developing new algorithms. With the rapid use of IoT-based smart devices. Heath sector has seen a great boost. Smart IoT-based bands and wristwatches can keep person’s vitals under check and report if it finds any abnormalities. A smart Machine Learning-based personal health monitoring band can be effective and detect abnormal body behavior and notify people and their loved ones. Such systems can be of great use to elderly people. Recently, the
Health-Related Diseases
Computational Intelligence for Data Analysis, Vol. 2 147
world has faced COVID-19, which choked hospitals and medical supplies. It is essential to control and prevent the spread of viral diseases. IoT smart systems can play a vital role in monitoring the spread of disease. IoT and wearable smart gadgets can warn people about their vitals and can warn them if they breach social distancing rules. A network of smart IoT devices connected can share data and inform authorities if a greater number of people than a specific threshold is getting sick in an area. Governments and authorities can timely know the possible spread of pandemics or viral diseases, and act accordingly. CONSENT FOR PUBLICATON Declared none. CONFLICT OF INTEREST The author declares no conflict of interest, financial or otherwise. ACKNOWLEDGEMENT Declared none. REFERENCES [1]
N. Caballé-Cervigón, J.L. Castillo-Sequera, J.A. Gómez-Pulido, J.M. Gómez-Pulido, and M.L. PoloLuque, "Machine learning applied to diagnosis of human diseases: A systematic review", Appl. Sci. (Basel), vol. 10, no. 15, p. 5135, 2020. [http://dx.doi.org/10.3390/app10155135]
[2]
V. Jain, and J.M. Chatterjee, Machine learning with health care perspective. Springer: Cham, 2020, pp. 1-415. [http://dx.doi.org/10.1007/978-3-030-40850-3]
[3]
S. Ray, 2019, February. “A quick review of machine learning algorithm,” International conference on machine learning, big data, cloud and parallel computing (COMITCon), pp. 35-39, 2019. [http://dx.doi.org/10.1109/COMITCon.2019.8862451]
[4]
J.K. Holodinsky, A.Y.X. Yu, M.K. Kapral, and P.C. Austin, "Comparing regression modeling strategies for predicting hometime", BMC Med. Res. Methodol., vol. 21, no. 1, p. 138, 2021. [http://dx.doi.org/10.1186/s12874-021-01331-9] [PMID: 34233616]
[5]
P. Nerurkar, A. Shirke, M. Chandane, and S. Bhirud, "Empirical analysis of data clustering algorithms", Procedia Comput. Sci., vol. 125, pp. 770-779, 2018. [http://dx.doi.org/10.1016/j.procs.2017.12.099]
[6]
S.H. Jambukia, V.K. Dabhi, and H.B. Prajapati, "Classification of ECG signals using machine learning techniques: A survey", International Conference on Advances in Computer Engineering and Applications, 2015pp. 714-721 [http://dx.doi.org/10.1109/ICACEA.2015.7164783]
[7]
H. Ahmed, E.M.G. Younis, A. Hendawi, and A.A. Ali, "Heart disease identification from patients’ social posts, machine learning solution on Spark", Future Gener. Comput. Syst., vol. 111, pp. 714-722, 2020. [http://dx.doi.org/10.1016/j.future.2019.09.056]
148 Computational Intelligence for Data Analysis, Vol. 2
Ayub et al.
[8]
A.K. Dwivedi, "Performance evaluation of different machine learning techniques for prediction of heart disease", Neural Comput. Appl., vol. 29, no. 10, pp. 685-693, 2018. [http://dx.doi.org/10.1007/s00521-016-2604-1]
[9]
G.A.P. Singh, and P.K. Gupta, "Performance analysis of various machine learning-based approaches for detection and classification of lung cancer in humans", Neural Comput. Appl., vol. 31, no. 10, pp. 6863-6877, 2019. [http://dx.doi.org/10.1007/s00521-018-3518-x]
[10]
T.B. Chandra, and K. Verma, "Pneumonia detection on chest X-ray using machine learning paradigm", In: B. Chaudhuri, M. Nakagawa, P. Khanna, S. Kumar, Eds., Proceedings of 3rd International Conference on Computer Vision and Image Processing. Advances in Intelligent Systems and Computing, vol 1022. Springer, Singapore, 2020. [http://dx.doi.org/10.1007/978-981-32-9088-4_3]
[11]
M.S. Kolkur, D.R. Kalbande, and V. Kharkar, "Machine learning approaches to multi-class human skin disease detection", International Journal of Computational Intelligence Research, vol. 14, no. 1, pp. 1-12, 2018.
[12]
K. Usman, and K. Rajpoot, "Brain tumor classification from multi-modality MRI using wavelets and machine learning", Pattern Anal. Appl., vol. 20, no. 3, pp. 871-881, 2017. [http://dx.doi.org/10.1007/s10044-017-0597-8]
[13]
N. Abiwinanda, M. Hanif, S.T. Hesaputra, A. Handayani, and T.R. Mengko, 2019. “Brain tumor classification using convolutional neural network,” World congress on medical physics and biomedical engineering, Springer, Singapore, pp. 183-189, 2018. [http://dx.doi.org/10.1007/978-981-10-9035-6_33]
[14]
P. Lodha, A. Talele, and K. Degaonkar, "Diagnosis of alzheimer’s disease using machine learning", Fourth International Conference on Computing Communication Control and Automation (ICCUBEA), pp. 1-4, 2018. [http://dx.doi.org/10.1109/ICCUBEA.2018.8697386]
[15]
R.A. Khan, Y. Luo, and F.X. Wu, "Machine learning based liver disease diagnosis: A systematic review", Neurocomputing, vol. 468, pp. 492-509, 2022. [http://dx.doi.org/10.1016/j.neucom.2021.08.138]
[16]
M.B. Priya, P.L. Juliet, and P.R. Tamilselvi, "Performance analysis of liver disease prediction using machine learning algorithms", International Research Journal of Engineering and Technology, vol. 5, no. 1, pp. 206-211, 2018.
[17]
A.S. Rahman, F.J.M. Shamrat, Z. Tasnim, J. Roy, and S.A. Hossain, "A comparative study on liver disease prediction using supervised machine learning algorithms", International Journal of Scientific & Technology Research, vol. 8, no. 11, pp. 419-422, 2019.
[18]
G. Nguyen, S. Dlugolinsky, M. Bobák, V. Tran, Á. López García, I. Heredia, P. Malík, and L. Hluchý, "Machine Learning and Deep Learning frameworks and libraries for large-scale data mining: a survey", Artif. Intell. Rev., vol. 52, no. 1, pp. 77-124, 2019. [http://dx.doi.org/10.1007/s10462-018-09679-z]
[19]
G. Battineni, G.G. Sagaro, N. Chinatalapudi, and F. Amenta, "Applications of machine learning predictive models in the chronic disease diagnosis", J. Pers. Med., vol. 10, no. 2, p. 21, 2020. [http://dx.doi.org/10.3390/jpm10020021] [PMID: 32244292]
[20]
H. Ben Hassen, W. Dghais, and B. Hamdi, "An E-health system for monitoring elderly health based on Internet of Things and Fog computing", Health Inf. Sci. Syst., vol. 7, no. 1, p. 24, 2019. [http://dx.doi.org/10.1007/s13755-019-0087-z] [PMID: 31695910]
Computational Intelligence for Data Analysis, 2023, Vol. 2, 149-176
149
CHAPTER 10
Machine Learning in Detection of Disease: Solutions and Open Challenges Tayyab Rehman1, Noshina Tariq1, Ahthasham Sajid2,* and Muhammad Hamza Akhlaq3 Faculty of Computing, SZABIST, Islamabad, Pakistan Department of Computer Science, Faculty of ICT, BUITEMS, Quetta, Baluchistan, Pakistan 3 Department of Computer Science, Allama Iqbal Open University, Islamabad, Pakistan 1 2
Abstract: Disease diagnosis is the most important concern in the healthcare field. Machine Learning (ML) classification approaches can greatly improve the medical industry by allowing more accurate and timely disease diagnoses. Recognition and machine learning promise to enhance the precision of diseases assessment and treatment in biomedicine. They also help make sure that the decision-making process is impartial. This paper looks at some machine learning classification methods that have remained proposed to improve healthcare professionals in disease diagnosis. It overviews machine learning and briefly defines the most used disease classification techniques. This survey paper evaluates numerous machine learning algorithms used to detect various diseases such as major, seasonal, and chronic diseases. In addition, it studies state-of-the-art on employing machine learning classification techniques. The primary goal is to examine various machine-learning processes implemented around the development of disease diagnosis and predictions.
Keywords: Classifiers, Machine learning Disease classification, Machine Learning Methods. INTRODUCTION Everybody remains so concerned with their families and careers that they do not have time to care for themselves in today's era. Due to their hectic lives, most individuals suffer from anxiety, nervousness, sorrow, and various other ailments. Taking this into consideration key, they are acquiring factors, are severely ill, and have acute diseases. Numerous diseases cause individuals to die each year, such as heart disease, cancer, liver cancer, tuberculosis, and so on, but major or chronic diseases cause the biggest number of deaths in the healthcare domain [1]. Modern Corresponding author Ahthasham Sajid: Department of Computer Science, Faculty of ICT, BUITEMS, Quetta, Baluchistan, Pakistan; Email: [email protected] *
Mariya Ouaissa, Mariyam Ouaissa, Zakaria Boulouad, Inam Ullah Khan, Sailesh Iyer (Eds.) All rights reserved-© 2023 Bentham Science Publishers
150 Computational Intelligence for Data Analysis, Vol. 2
Rehman et al.
technology contributes to the expansion of advanced healthcare methods, such as smart human health monitoring systems, and individualized treatment to make a diagnosis for everyone with the utmost care. People are affected by many diseases every day, all around the world. Some of these diseases remain undiagnosed and reach the critical phase [2]. COVID-19, malaria, dengue fever, the common cold, and fever are all detected and treated along with basic laboratory examinations and readily obtainable drugs. However, featuring cancer, chronic, seasonal, psychiatric problems, and several diseases and infections can be controlled at a certain stage with more effort. In rare cases, they can even be treated if caught early enough [3]. Most previous research attempted to predict disease occurrence using patient laboratory testing and drugs [4, 5]. These models remained also widely applied to describe unknown risk factors, typically although simultaneously increasing detection specificity and sensitivity. Multiple approaches, including supporting vectors, machines logistical regression, random forests method, neural networks model and time cycle modeling methods, have been found to be useful in predicting disease in recent research [6, 7]. Before doing a thorough analysis of the problem, it is necessary to have a deeper understanding of basic ML techniques. The main objective of this work is to offer an overview of existing ML approaches [8]. The major contributions of this survey paper are listed below: 1. A novel state-of-the-art literature survey on diseases’ detection based on Machine and Deep Learning is presented. 2. Taxonomies based on supervised, unsupervised, reinforcement, and deep learning are proposed. 3. A critical analysis of state-of-the-art is presented in a tabular way. 4. Issues and challenges concerning disease type and ML approach are presented. The rest of the paper is organized as follows: Section 2 presents a classification and taxonomy of ML approaches used in disease prediction. Section 3 details disease detection mechanisms, and finally, Section 4 provides the conclusion and future directions. MACHINE LEARNING APPROACHES This research focuses on categorizing ML approaches into supervised, unsupervised learning, reinforcement learning, and DL, as seen in Fig. (1). The categories are mentioned below:
Machine Learning in Detection of Disease
Computational Intelligence for Data Analysis, Vol. 2 151
Fig. (1). Diagrammatic representation of machine learning techniques.
Supervised Learning (SL) Given a training number of instances with appropriate targets, processes respond accordingly to the complete possible inputs created on this training set. The SL method is also known as learning from exemplars. The two-step SL method is as follows and shown in Fig. (2). ● ●
A predictive model is a combination of available data and known answers. When showing new data, the predictive model offers appropriate responses.
Unsupervised Learning There are no appropriate responses or targets specified. Unsupervised learning explores commonalities between the stored data and classifies the data based on these commonalities. Data acquisition is yet another name for this. Clustering is a feature of unsupervised learning. When no outputs are provided, un-supervised ML is used to learn patterns in the data from the inputs. Clustering, NNs, and HMM are often used as unsupervised learning techniques. These methods allow the exploration of unlabeled data to find intrinsic or hiding patterns.
152 Computational Intelligence for Data Analysis, Vol. 2
Rehman et al.
Reinforcement Learning (RL) This technique of learning is promoted by behavioral psychology. The algorithm is notified when the answer remains inaccurate, although it is not just how to improve it. It must investigate and test a variety of choices before deciding on the best one. Another phrase is learning with such a reviewer, who provides any improvement ideas. RL differs from the SL method in that it does not provide precise input and output sets or explicitly specify sub-optimal behaviors. Furthermore, it concentrates on online performance. Data Mining (DM) It is a process and technique for transforming large amounts of data into useful knowledge. DM is classified as a nontrivial data analysis process extraction of implicit information that was previously unknown and might be valuably gleaned from a database's data [9]. DL in databases is part of a larger process termed knowledge discovery. Several pre-processing techniques are used to make the DM algorithm and post-processing easier to use processing procedures aiming at refining and increasing the knowledge that has been found [10]. DL, deep structured learning, and classified learning are other terms for the same thing. DL may learn from many stages of depictions or data statements to construct a hierarchical representation. As a result, DL is a “thing of the future” that leads to a rapid transition from shallow to deep architecture. Speech recognition, pattern recognition, computer vision, language processing, bioinformatics and statistical information remain all fields where DL is widely employed [11].
Fig. (2). Step-by-step supervised learning process.
Machine Learning in Detection of Disease
Computational Intelligence for Data Analysis, Vol. 2 153
ML algorithms were first developed and used to evaluate medical data sets. ML now recommends a variety of tools for effective data analysis. For the study of medical data, ML technologies are particularly successful, and much progress has been made in diagnosing disorders. In modern hospitals, reliable diagnostic data is presented in a health record or reports or their data section [12]. A valid diagnostic medical record is loaded into a machine as an intake to execute an algorithm. This generated classifier improves doctors in identifying new patients at a high rate and with greater accuracy [13]. ML has provided self-driving automobiles, voice recognition, and better human vision in the past. ML is now so prevalent that it is possible to utilize it numerous times a day without even realizing it. The theme of MLT is pattern detection, which can help predict and make decisions about diagnosis and therapy. ML algorithms can store and process data, combine data from several sources, and incorporate background information into research (Fig. 3) [14].
Fig. (3). Deep learning algorithm representation.
DETECTION OF DISEASE BY USING DIFFERENT MACHINELEARNING CLASSIFICATION Different ML approaches used by various researchers have produced disease diagnostic methods. Researchers have demonstrated that ML algorithms are useful in identifying a range of diseases. Fig. (4) shows a symbolic approach to disease diagnosis using Machine Learning Techniques.
154 Computational Intelligence for Data Analysis, Vol. 2
Rehman et al.
Fig. (4). Taxonomy of machine learning classification for disease detection and prediction.
CHRONIC DISEASE: DETECTION OF HEART DISEASE Heart and blood vessel diseases often called cardiovascular diseases (CVD), are just diseases that affect the heart vessels. Congestive heart failure and heart attacks are examples of coronary heart disease, a kind of cardiovascular disease. A certain kind of heart disease is coronary artery disease, which occurs when the blood vessels in the heart get congested; a fatty substance referred to as atherosclerosis forms privileged the coronary arteries [15]. Cardiovascular disease is one of the most unpredictable diseases in medicine because it involves many criteria and complexities. ML may be a good option for attaining high predictive precision, not just for heart disease but also for other diseases. This tool uses extracted features and data to predict heart diseases under various conditions. ML algorithms, such as NB [16], DT [17], KNN [18] and Neural Networks [19], are used to predict heart disease risk. Some of the most famous algorithms are discussed below. Naive Bayes (NB) A probabilistic classifier that does not ascribe dependencies between characteristics. The probability should be maximized to determine the class. The benefits include dealing with the naive Bayes model without applying Bayesian procedures. Naive Bayes classifiers are processes for determining the conventional order of grouping concerns, such as phishing detection, and are also
Machine Learning in Detection of Disease
Computational Intelligence for Data Analysis, Vol. 2 155
suitable for medical issues in general. It consists of attributes in a single class independent of other characteristics in the same type [20]. Decision Tree (DT) A DT is among the supervised learning techniques in machine learning algorithms. It may be used for classification as well as regression. With continuous and categorical parameters, it functions well. This algorithm separates the population into two or more related groupings based on the most important predictors. Many factors impact the patient in cardiac illness, including age, blood pressure, genetics, sugars, and others. The doctor may quickly find the essential trait among all the factors by examining the decision tree. They can also produce highly affecting characteristics in many people. The decision tree, based on entropy, clearly indicates the dataset’s significance [21]. K-Nearest Neighbor (K-NN) The KNN algorithm is a technique for supervised classification. It classifies items built upon their similarity to one more. It is a good example of a specific instance learning method. The Euclidean distance is utilized and calculates the distance between an attribute and its neighboring points. It uses a set of named points to determine how to mark another point. The data are grouped based on their similarity, and K-NN may be used to fill in the missing values in the data. After the misplaced values have been filled in, the data set is subjected to various prediction approaches. Applying a different combination of these methods is feasible to make progress accurately [22, 23]. Table 1 describes the techniques used to detect heart disease using machine learning, whereas Table 2 describes the accuracy achieved for various algorithms in a similar domain. Table 1. Machine-Learning techniques applied to heart disease. References
Classification Technique
Advantages
[16, 20]
Naive Bayes
Handles discrete and continuous data
[17, 21]
Decision Tree
Accuracy, Sensitivity, No additional relevant data is used to test Specificity, Precision, Area the algorithm. Overfitting may arise as a under curve result of the tree's recurrent growth.
[22]
KNN
Simple, high accuracy
Limitations Low with limited dataset.
Local information structure sensitivity. Calculated slowly.
156 Computational Intelligence for Data Analysis, Vol. 2
Rehman et al.
Table 2. Classification accuracy. References Year [20, 22] [21] [24, 25]
2021
Algorithm Naive Baye
2020 Decision Tree 2019
KNN
Accuracy
ML Technique
Chronic disease diagnosis between 82% and 92%. Supervised Learning 86.3% (testing phase) 87.3%.
Supervised Learning
97.3%
Supervised Learning
Issues and Challenges Three supervised ML techniques are utilized in the survey for chronic heart diseases. The heart disease dataset was analyzed using these methods. For this method, the Classification Performance should be compared [26]. This research should be expanded to predict heart disease using fewer variables. To get superior outcomes, several researchers and developers have investigated many ML sectors in telemedicine. Unfortunately, such studies do not provide ideal solutions to difficulties relating to healthcare delivery. We discovered that the researchers' worries are spread across various ML aims. We organized these issues into divisions based on the nature of ML, which we classified as classification, estimation, and identification. CHRONIC DISEASE: DETECTION OF DISEASE BREAST CANCER In most cancer-affected females, Breast cancer is the leading reason for losing a life. Mammography is one of the highly reliable breast cancer early detection and diagnostic techniques, and it helps to minimize death. Mammograms come about radio-graphic pictures of the breast to detect early breast cancer signs. A breast cancer diagnosis is tough and time-consuming to do by hand, prompting an automated process. Current healthcare systems have developed to be beneficial, although they are predisposed to making mistakes [27]. As a result, CAD-based medical image classification has evolved as a useful tool for clinicians to categorize medical pictures into distinct categories, allowing for immediate diagnosis and treatment. In this perspective, ML techniques and DL algorithms must be developed to diagnose the condition more reliably at a previous stage, reducing the frequency of re-admissions in clinics and hospitals [28]. As a result, AI approaches can help accelerate the development of new fidelity guidelines in medicine while also lowering healthcare expenses associated with misdiagnosis [29]. Inflammatory breast cancer is a frequent and serious type of breast cancer that spreads and develops quickly. Nipple Paget infection is a type of breast cancer that remains extremely rare. It begins in the breast canals and progresses to the nipple skin, but then the areola. Phyllodes tumors remain a relatively uncommon kind of breast cancer. They make progress in the stroma of the breast
Machine Learning in Detection of Disease
Computational Intelligence for Data Analysis, Vol. 2 157
(connective tissue). The majority are non-cancerous, but some are cancerous. The following techniques are widely practiced in this regard. CAD System Both computer-aided detection and diagnosis are frequently used with CAD. In practice, CAD is called computer-aided detection to identify and categorize suspected suspicious portions of pictures. It is suggested as CAD when it can evaluate and classify benign and cancerous breast cancer. CAD systems must assess diagnostic pictures, which necessitates the creation of complex computer algorithms [30]. Although CAD helps detect early-stage invasive breast cancer, it also increases the likelihood of false-positive findings. Thus, precision and workflow efficiency should be handled while incorporating CAD into clinical practice. Secondly, user training is also necessary to understand the CAD system’s limitations and capabilities to avoid improper exploitation [31]. Deep Learning It is classified as a representational learning technique because Neural Networks can study transitory, hierarchy, and further cognitive characterization of data before categorizing the full image. DL uses numerous nonlinear computation layers to learn attributes directly from data [32]. According to a recent study, a convolutional neural network (CNN) shows excellent performance in tumor detection and treatment. The layers that follow the input data are more general and absorb low-level data. Layers next to the output layer, on the other hand, acquire distinct aspects of the input picture and are thus more specialized [33]. Machine-Learning Techniques ML Technique’s learning is frequently divided into SL and unsupervised learning. In supervised learning, a collection of data examples is utilized for training the system, and each example remains labelled to guarantee that the proper out- put is produced. Uncontrolled learning, on either side, lacked pre-existing knowledge sets or concepts of the intended outcome, meaning that the goal is difficult to achieve. Some of the famous techniques are listed below: 1. Support Vector Machine (SVM): A SVM-based Breast cancer detection is automated using a support vector machine. The suggested model is tested, and the results demonstrate that it performs at an average of 87.12 percent. The study’s authors did not explain how the hyper-parameters while training a support vector machine are determined [34].
158 Computational Intelligence for Data Analysis, Vol. 2
Rehman et al.
2. Random Forest (RF): joining the forest of trees needs a lot of decision-making trees. A randomly sampled size N from the replacements data set or a random selection from the forecasts without substitution is selected using an algorithmic technique. The gathered data is separated into segments. After then, the remaining data is destroyed. These steps are done multiple times based on the number of trees required. Finally, these trees classify and count the observations. The clustering algorithm is then decided on by most of the cases to classify them. Convolutional Neural Network Model (CNN) The input, output, and additional hidden layers in the Convolutional Neural Network architecture are all connected layers known as morphing pooling. AbdelZaher and Eldeib devised a CNN-based breast cancer detection approach that depended on an unmonitored possessing range of deep-faith beliefs and an assumed to vary channel [35, 36]. The suggested arrangements are only intended to give a limited number of the labelled data training set. In contrast, CNN generally uses a substantial portion of data sets for parameter training and modification. Data measuring, task collection, data separation cross, and CNN were all included in the diagnostic gadget. The four forms of medically significant tissue are normal tissue, benign lesions, carcinoma in situ, and aggressive malignancy. The proposed CNN architecture tries to combine data on several scales. A deep learning technique was used to classify photos from BreaKHis [37], a public database of breast cancer pathologic images. They suggested an approach for CNN development based on the abstraction of picture patches and combining these finishing classification patches. However, research is costly due to a lack of data [38]. Issues and Challenges Mammogram categorization is a challenging but important task because it is one of the first estimates to determine risk status. Even though clinics now have more advanced imaging capabilities, researchers should address device-level mammography anomalies by generalizing their models across several datasets. Several of the papers are likewise difficult to understand [45]. In their discussion on the opportunity of CAD in medical science, Yanase and Triantaphyllou [44] addressed certain critical topics. Data gathering and quality estimates, improved segmentation, feature engineering, dealing with massive data, and fostering essential implementation assessment for all generated standards are examples. With 73 percent (389) of the papers selected addressing the diagnosis task, it is critical to detect Breast-Cancer sooner to enhance treatment effectiveness. The bulk of the studies suggested developing a new computer-assisted diagnosis system to assist clinicians in determining the type of tumor with greater precision.
Machine Learning in Detection of Disease
Computational Intelligence for Data Analysis, Vol. 2 159
Only one thing suggested a recent solution again used for managerial jobs, with 17 percent (88) publications interested in screening, 6 percent (26) papers interested in treatment, and 6 percent (26) articles interested in prediction. While being one of the most significant medical fields in Breast Cancer, the treatment task received little attention in the selected publications; for example, it discussed the significance of integrating ML approaches in the test type to create a solid roadmap for the cure of the cure Breast Cancer [46 - 48]. Here, comparisons of various methodologies, classification methods, and feature extraction for disease prediction were reviewed. Big data must be compacted with data loss, which enhances classification effectiveness because selecting characteristics allows us to eliminate unneeded data. However, a good subset guideline is challenging, worsened because it necessitates intricate reliance on a wide variety of parameters. We could integrate guidelines and extract features to improve classifier outcomes in the future. Additionally, you can start experimenting with algorithm potential for most data sets with distinguishing features such as noisy data, sparsity, and missing values to enhance model accuracy. Hence, you can try time to experiment with algorithm possibility for most data sets with specific attributes such as noisy data, sparseness, and misplaced value to improve model accuracy, as illustrated in Table 3. Table 3. Detection of breast disease approaches advantages and disadvantages. References
Approaches
Pros
Cons
[39]
CNNs
Larger receptive field. Accuracy in Image recognition problems.
Arbitrary. Slow training rate at complex tasks
[40]
SVM
Effective in high Dimensional space. Effective in Situations Where the number of measurements exceeds The number of samples.
In circumstances in which the set of features exceeds the number of samples, control is critical.
[41]
KNN
In a simple and effective approach, distinguish variance from the norm in mammography pictures.
Limited prominent feature extraction techniques
[41]
RF
Works well with non-linear data. Lower risk of overfitting.
Slow Training
[42]
Deep Multi-Instance For sparsely labelled data, this is a good option.
Fails with imprecise instance classes.
160 Computational Intelligence for Data Analysis, Vol. 2
Rehman et al.
(Table 3) cont.....
References
Approaches
Pros
Cons
[43]
The neural network feed-forward classifier
High performing of Feature extraction.
For sparsely labelled data, this is a good option.
[44]
CAD
Image saving during the analysis of the images. Decreased false negative rate and improved sensitivity
Diagnostic accuracy was not improved with CAD on any performance metric assessed.
CHRONIC DISEASE: DETECTION OF DISEASE DIABETES Every year, over 1.6 million people die because of diabetes, according to the WHO (World Health Organization). Diabetes remains a disease that arises when the human body’s blood glucose/blood sugar level is extremely high. Insulin is among the body’s most vital hormones. It helps the body convert sugar, carbohydrates, and other foods into the energy it needs to function. Nevertheless, if the body does not generate or use insulin adequately, excess sugar will be excreted through pee. Diabetes is the medical term for this condition. The exact aetiologia of diabetes is unknown, while obesity and a lack of physical activity seem to play a role. High blood sugar is a symptom of diabetes, a chronic condition. It can induce a variety of serious illnesses, such as stroke, kidney failure, and heart attack. In 2014, diabetes afflicted over 422 million individuals around the world. In 2040, the population will be 642 million. The primary goal of this research is to create a machine learning-based system that can predict diabetic individuals. Diabetes develops when the pancreas, an organ in the human body, cannot create enough insulin, and the insulin produced cannot be utilized by the body’s cells. Diabetes is a disease with no known cure; therefore, early detection is essential. DM, ML techniques, and Neural Network models’ methodologies to predict diabetes are proposed in [49]. Table 4 describes the accuracy achieved so far to detect diabetes disease. The Pima Indian diabetes data collection was used in this study. Cross- validation predicts better than percentage split (70:30) in this data set. Using Classification Technique and Percentage Split, J48 achieves an accuracy of 74.8698 percent and 76.9565 percent, respectively. Using a percentage split up test, algorithms determine the maximum accuracy [50, 51]. ML techniques have been developed to be particularly useful for identifying acceptable threshold of dangerous variables and physiological responses impacting diabetes, thanks to the ongoing advancement of AI Technology. Second, both machine learning and medicinal diagnostics have the same goal: to obtain accurate and useful information from a large amount of data to make
Machine Learning in Detection of Disease
Computational Intelligence for Data Analysis, Vol. 2 161
conclusions [52, 53]. Following are the conventional ML Techniques detection of Diabetes disease. Table 4. Detection of Diabetics Disease Approaches and accuracy. References
Approaches
Compared With
Performance
ML Technique
[59]
XG Boost
NB, Ada Boost, RF
93.75%
Supervised Learning
[60]
SVM
RF, NB, DT, KNN
77.73%
Supervised Learning
[61]
Logistic Regression
DT, SVM, KNN
77.90%
Supervised Learning
[62]
Gradient Boosting
LR, RF
84.7%
Supervised Learning
[63]
Naive Bayes
PS
79.56%
Supervised Learning
Logistic Regression (LR) Statistical model for describing the connection among the logic translation of binary predictor variables and individual or more than independent variables by determining the constants of the most excellent fitted linear pattern. This simple prediction model gives’ baseline accuracy values for similarities and nonparametric machine learning [54]. Random Forest Classifier (RFC) An ensemble type that uses a bagging method to create several random DT. Each tree represents a probable outcome in an analysis diagram. For global categorization, the average forecast among trees is used. The disadvantage of big variance for decision trees is reduced because of this. Impurity and gain ratios are used to do decision splits [55]. Gradient Boosted Trees (GBT) A decision tree-based ensemble prediction technique. Indifference to Randomized Forests, this model creates decision trees step-by-step utilizing gradient descent to reduce a loss function. A weighted democratic majority of all the DT is used to make a final forecast. We discuss XGBoost, a gradient boosting method tuned for speed and reliability [56]. Weighted Ensemble Model (WEM) To aggregate the outcomes of all the previous models. Multiple forecasts from different models can be averaged using weights based on the performance of each model. The weighted ensemble, corresponding to the model’s logic, might benefit from the strength of numerous models to give more precise findings [57].
162 Computational Intelligence for Data Analysis, Vol. 2
Rehman et al.
The regression model and SVM algorithms were preferred as standard models for our investigation based on existing research in the domain. To the benefit of nonlinear connections inside the data for calculation and diagnosis, RFC, GBT, and WEM models are built into our study, and classification comparison results are shown in Fig. (5) [58].
Fig. (5). Detection of diabetes disease data accuracy models.
Issues and Challenges SVM Classifier, DT, and Neural Networks are the algorithms employed in this work. Any optimization approaches will be used to improve this work in the future. The author employs a data mining method to examine the accuracy of forecasting diabetic status. This work will be expanded in the future by using more optimization techniques. However, poor glucose control harms people`s opinions of telemedicine`s utility, particularly in men. Beyond the epidemic, telemedicine could be a viable option for improving the efficiency and costeffectiveness of diabetic care. CHRONIC DISEASE: DETECTION OF LIVER DISEASE The liver is the body’s largest organ and is responsible for processing food and absorbing poisonous substances. It is feasible to prevent liver failure by detecting and treating liver problems early on. The first stage of liver illness is caused by inflammation, which may or may not is accompanied by any symptoms. Chronic inflammation causes tissue to replace healthy liver tissue, causing the disease to
Machine Learning in Detection of Disease
Computational Intelligence for Data Analysis, Vol. 2 163
progress to the second phase, fibrosis, which is also mostly asymptomatic. Cirrhosis, the third stage, is caused by severe damage to the liver. Symptoms such as stomach pain, weariness, weakness, and jaundice begin to appear at this stage viruses and alcohol usage cause liver damage, putting a person’s life in jeopardy. Hepatitis, cirrhosis, liver tumors, liver cancer, and various other illnesses affect the liver. Liver disorders and cirrhosis are the leading causes of death [63]. For the analysis and prediction of liver disease, machine learning has had a considerable impact in the biomedical area. ML promises to improve disease diagnosis and prediction, which has piqued interest in the medical area while also increasing the objectivity of the decisionmaking process. Medical issues can be easily solved with machine learning techniques, and diagnostic costs can be decreased. The primary goal of this study is to improve the accuracy of prediction to decrease the cost of diagnostics in the medical field [64]. The SVM classifier and NB Classification algorithms predict liver disease. UCI gave the ILPD data set. This data set has 560 occurrences and ten attributes. A comparison is made based on execution accuracy. The naive Bayes algorithm achieves 61.28 percent accuracy in 1670.00 milliseconds. In 3210.00 milliseconds, SVM achieves 79.66%. The implementation is done in MATLAB. SVM predicts liver illness with the best accuracy when compared to Naive Bayes. Compared to the SVM, the Navies Bayes takes less time to run [65, 66]. Developing a machine learning technique for liver disease diagnosis and prediction, the followings are the steps to take and the proposed model ML classification proposed in Fig. (6). Major steps for the detection of Liver Disease are discussed below. Data Selection and Pre-Processing The data collection process necessitates selecting pertinent information and acquiring useful knowledge using various data mining techniques, using a dataset from the UCI ML for this experiment. Furthermore, the original dataset is gathered. There are 583 liver patients in this dataset, with 75.64 percent of male cases and 24.36 percent of female patients. We chose ten parameters for additional research and 1 factor as a class label from this dataset, with 11 specific parameters [67]. Pre-processing of data that converts raw data to clean data can improve prediction performance for ML. The techniques used are cleaning, normalization, transformation, missing value filling, verifying inconsistent values, eliminating repeating groups, eliminating noisy data, feature combination, feature sampling, dimension reduction, and feature encoding.
164 Computational Intelligence for Data Analysis, Vol. 2
Rehman et al.
Fig. (6). Architecture proposed model ML classification.
Feature Selection Selection is locating input features for a predictive model while excluding noncontributing features. One of the most advanced applications which require optimization methods is the genetic algorithm. This approach of natural selection is like Darwin’s. Initializing, fitness assignment, selection, crossover, and mutation are the steps followed by the genetic algorithm. Classification Algorithm Classification is known as constructing a model of class characteristics from a dataset to allocate a class label to an unknown track record with high accuracy. For classification, ML algorithms utilize training data with important characteristics. There are a variety of machine learning algorithms to choose from, depending on the sort of prediction and diagnosis models to be produced.
Machine Learning in Detection of Disease
Computational Intelligence for Data Analysis, Vol. 2 165
Supervised Learning and Unsupervised Learning These algorithms convert the input to an expected outcome and work following the desired outcome. The learning process is repeated using these methods until the model is generated. DT, Random Forest, KNN, SVM, LR and NB are examples of these techniques (NB). No goal outcome is required in these unsupervised learning techniques, and they are mostly employed to solve clustering issues. The Apriori algorithm, k-means, and other algorithms are examples of these algorithms. Performance Metrics Analysis This stage creates the ML algorithm used to achieve the intended outcomes and hypotheses. Using a cross-validation approach, the efficiency of methods is evaluated to identify the one that provides the maximum accuracy. The final predictive or diagnostic model is produced by employing the tailored dataset containing significant features for training and testing the ML classification model, which provides the final predictive or diagnostic model that makes automatic judgments about the new and unfamiliar dataset. Predicted Results For liver datasets, the performance of several ML techniques has been assessed. Oversampling is accomplished with the help of a combination of classifiers from the ML repository, and changeable content selection is accomplished with the help of a combination of classifiers from the UCI ML repository. To solve the problem, an isolation forest was employed. The performance of most algorithms has improved significantly because of feature extraction and outlier reduction, and the time taken to train and the test has decreased significantly. Issues and Challenges The main goal of this survey is to develop an efficient diagnosis method for patients with chronic liver disease using several supervised learning and unsupervised machine learning models. The absence of big training data sets is frequently cited as a stumbling block. This assumption, however, is only partially valid. Unsupervised approaches are appealing because they allow network preparation and a vast amount of un-labelled data that is accessible. An additional reason to believe that unsupervised methods would then continue to play an important role is the analogue to cognitive behavior, which appears to be much more data effective and occurs in an unsupervised manner to some extent; we can learn to know artefacts than frameworks deprived of knowing this same label [68]. Calculations get complicated when values are unclear, or many outcomes are
166 Computational Intelligence for Data Analysis, Vol. 2
Rehman et al.
related. We may integrate recommendations and feature selection to improve the classifiers’ performance in the future. In addition, novel feature selection methods, such as ant colony, have been developed. It is feasible to test optimization to increase quality. You may try exploring algorithm possibilities for most data sets with different traits like, noisy data, sparsity, missing data—value, and so on to improve model accuracy, as described in Table 5. Table 5. Approaches and limitations for dengue predictions. References Approaches ML- Technique Performance
Limitation
[73]
ANN
Un- Supervised Learning
99.8
Algorithms like decision trees, SVM, RF on the other hand, are quite interpretable.
[40]
SVM
Supervised Learning
90.42%
Number of samples is needed for precise results.
SEASONAL DISEASE: DETECTION OF DENGUE DISEASE Most research focuses on environmental variables such as weather patterns linked to the transmission and prevalence of dengue fever. An increasing body of data has shown a link between these variables and the disease’s prevalence [69]. Dengue fever has been the most frequent viral fever among humans. It is also referred to as a life-threatening illness as well as the number of people infected with dengue fever is rising every day. Many people are at risk of contracting dengue fever. Early detection of dengue fever can save a person’s life by alerting them to take proper precautions and treatment. However, it is impossible to predict whether this will occur. Dengue diagnostic examinations can take up to ten days to detect the virus in a patient’s blood, but the patient’s condition can be improved in that period. This study aims to see if Artificial Neural Networks (ANN) technology can be used to identify the onset of dengue fever in a patient [70, 71]. The outputs of the three parameters employ certain genuine parameters such as average temperatures, average relative humidity, total rainfall, and reported dengue cases. They utilized ANN in this project since it is an excellent technique for learning a problem from examples and does not require any computational models of any situation. This study relies on the identification of confirmed dengue cases [72]. As a result, the examination of the clinical characteristics of 523 dengue cases was reported. In this example, researchers employed unsupervised learning to discover natural groupings linked to specific confirmed dengue infections. Utilizing laboratory, clinical, and demographic information helped determine the relationship between predictive characteristics and the
Machine Learning in Detection of Disease
Computational Intelligence for Data Analysis, Vol. 2 167
likelihood of contracting dengue fever. This study relies on the identification of confirmed dengue cases [73]. Issues and Challenges Dengue fever is a major global emerging sickness that wreaks havoc on endemic countries’ health systems, necessitating the development of a high-performance ensemble-level learning forecasting approach. It is difficult to distinguish dengue from those other prevalent febrile diseases before complications - a simple and low-cost strategy is urgently needed to promote early diagnosis, both to improve care coordination and to facilitate the efficient use of limited resources, as well as for patients who may be at high risk of side effects. The current study on Dengue disease aims to detect the diseases in clinical literature and link them with time frames. The dataset is labelled, and feature extraction and classification methods are utilized to investigate the condition. The dataset will be analyzed to achieve an effective outcome, and a classification approach will be utilized. The current researchers believe that the system is improved by applying more powerful methods in providing output-wise dengue impact analysis. We will continue to increase overall speed and computation using a variety of characteristics in the future. Several dengue fever prediction models have been created. Because the field is still in its inception, additional research is required, but there is still much work. As a result, it is important to develop a new upgraded innovative composite model for predicting dengue outbreaks that incorporates a variety of machine learning algorithms. SEASONAL DISEASE: DETECTION OF COVID-19 DISEASE SARS-CoV-2, a virus that first arose as an outbreak in the Chinese region of Hubei, is the cause of the current COVID-19 pandemic. COVID-19 sufferers’ treatment is still complex and disputed, as one would expect, given the disease’s early appearance. The first indications of COVID-19 are like those of many other respiratory illnesses and inflammatory disorders, including fever, sneezing, rhinitis, a protracted cough, and tiredness with body pains [74]. Recent discoveries indicate that ML can play a significant role in COVID-19 research, prediction, and discrimination. Finally, machine learning may analyze and triage COVID-19 instances in health provider programmers and plans [75]. With 92.9 percent testing accuracy, supervised learning out- performed rival unsupervised learning methods. Continuous supervised learning may be used to improve accuracy [76]. ML and transfer having to learn solutions may learn elements repeatedly utilizing hidden layers, and they are usually trained on supercomputers to save time. When combined with sophisticated extracting features and selecting algorithms, ML-based systems can improve accuracy [31, 32].
168 Computational Intelligence for Data Analysis, Vol. 2
Rehman et al.
Furthermore, if the retrieved subjective feature is picked correctly, one can get greater classification accuracy and define the illness. This study looks at nine different AI models. First, one DL-based CNN, followed by VGG16, DenseNet121, DenseNet169, DenseNet201, and MobileNet, are all deep learningbased transfer learning techniques. The remaining three ML models, ANN, DT, and RF were created to classify CT-segmented lung COVID versus Controls. Nine AI models are compared to increase categorization accuracy. Fig. (7) further illustrates the accuracy achieved using standard deviation.
Fig. (7). A plot of measurement, data, and standard deviations.
For COVID-19, a hybrid technique combining machine and deep learning has been used. They established a decision tree using inherent COVID-19 genomic fingerprints as patterns, a non-alignment method for estimating the COVID-19 virus’s gene sequence. Only raw DNA sequence data is processed in the alignment-free technique, which results in the quick taxonomic categorization of new diseases [77]. The suggested technique was evaluated by a huge dataset with over 5000 distinct viral genomic sequences. These numbers came after a wellestablished database called Virus-Host DB. The findings revealed that the suggested technique is viable for analyzing pathogen genome sequences and providing correct taxonomic classes for previously unknown sequences in realtime [78]. Offered an alternate strategy for avoiding COVID-19 outbreaks in India. The researchers used the COVID-19 Indian dataset. NML and second derivative regression are two well-known methods combined in the suggested method. PDL was used to normalize the data, and NML was used to forecast the outcome. Compared to prior efforts, the experimental findings showed that this approach is superior in classification performance and prediction time [79]. For COVID-19, a hybrid technique combining machine and DL has been used. It was proposed to use an iteratively trimmed DL model. The researchers employed X-ray scans of COVID-19 pulmonary symptoms. In this procedure, the unique
Machine Learning in Detection of Disease
Computational Intelligence for Data Analysis, Vol. 2 169
feature representation of COVID-19 was learned using a bespoke CNN and a trained model using ImageNet. Patients were then classified as COVID-19-viral abnormalities, normal, or bacterial pneumonia cases using the newly acquired knowledge. The suggested model performs well in experiments, with just an efficiency of 99.01 percent and an AUC value of 0.9972 [80]. Similarly, deeplearning (CNN) was used to estimate the difficulty of COVID-19 patients’ lung sickness. Their method was evaluated using a dataset of 131 CXRs collected from 84 COVID-19 patients in US hospitals. The data was split into two halves. About 80% of the data was utilized for training, with the remaining 20% being used for testing. The suggested technique is evaluated using correlation analysis and square error analysis. The authors said their technique should be evaluated with a bigger dataset because the findings were good in a small sample. Furthermore, the scientists stated that their method might be used to determine the severity of lung disorders in COVID-19 patients and examine sickness progression and therapy response [81, 82]. Issues and Challenges To build successful machine learning systems, a large amount of data is required. The use of machine learning in the COVID-19 study is currently riddled with problems. One of the most significant barriers to using DL to diagnose COVID-19 is the lack of standard data. Another problematic issue is the dataset’s sample imbalance. COVID-19 data had fewer X-ray and CT scans than influenza and typical case samples. The most common way of coping with an imbalanced dataset is data augmentation. To construct new lesions using COVID-19 data, this approach employs flipping, rotation, zooming, noise addition, and many others. Another benefit of this strategy is feature extraction to alleviate overfitting issues. The amount of data is a key constraint in this study. Some laboratory results could not be obtained for some individuals; hence data from 600 patients were utilized. However, the estimate ranged from 84 percent to 93 percent in a detectable population range. Furthermore, the data was unbalanced. They corrected this by eliminating certain materials. A larger data collection can improve the performance of these models. CONCLUSION Many applications, such as image identification, DM, Natural language, and disease diagnostics, rely on machine learning. ML has potential solutions in all these areas. Medical data sets, on either hand, are frequently multidimensional. ML approaches have proven ineffective, necessitating Big Data technologies. As a result, DL emerged as a subset of ML techniques that allow us to work with these data sets. We also show some instances of algorithms used in different fields of
170 Computational Intelligence for Data Analysis, Vol. 2
Rehman et al.
medicine and analyze some likely patterns based on the aim sought, the approach employed, and the application domain. Furthermore, we outline the benefits and drawbacks of each approach given to aid future research into which technique is best suited for each real-world circumstance covered by other authors. Finally, this paper describes all methodologies, including supervised and unsupervised learning algorithms. CONSENT FOR PUBLICATON Declared none. CONFLICT OF INTEREST The author declares no conflict of interest, financial or otherwise. ACKNOWLEDGEMENT Declared none. REFERENCES [1]
A. Dinh, S. Miertschin, A. Young, and S.D. Mohanty, "A data-driven approach to predicting diabetes and cardiovascular disease with machine learning", BMC Med. Inform. Decis. Mak., vol. 19, no. 1, pp. 211-226, 2019. [http://dx.doi.org/10.1186/s12911-019-0918-5] [PMID: 31694707]
[2]
M. Jiang, Y. Chen, M. Liu, S.T. Rosenbloom, S. Mani, J.C. Denny, and H. Xu, "A study of machinelearning-based approaches to extract clinical entities and their assertions from discharge summaries", J. Am. Med. Inform. Assoc., vol. 18, no. 5, pp. 601-606, 2011. [http://dx.doi.org/10.1136/amiajnl-2011-000163] [PMID: 21508414]
[3]
E.M.F. El Houby, "A survey on applying machine learning techniques for management of diseases", J. Appl. Biomed., vol. 16, no. 3, pp. 165-174, 2018. [http://dx.doi.org/10.1016/j.jab.2018.01.002]
[4]
N. Razavian and D. Sontag, “Temporal convolutional neural networks for diagnosis from lab tests,” arXiv, 2015.
[5]
A. Perotte, R. Ranganath, J.S. Hirsch, D. Blei, and N. Elhadad, "Risk prediction for chronic kidney disease progression using heterogeneous electronic health record data and time series analysis", J. Am. Med. Inform. Assoc., vol. 22, no. 4, pp. 872-880, 2015. [http://dx.doi.org/10.1093/jamia/ocv024] [PMID: 25896647]
[6]
W. Yu, T. Liu, R. Valdez, M. Gwinn, and M.J. Khoury, "Application of support vector machine modeling for prediction of common diseases: the case of diabetes and pre-diabetes", BMC Med. Inform. Decis. Mak., vol. 10, no. 1, p. 16, 2010. [http://dx.doi.org/10.1186/1472-6947-10-16] [PMID: 20307319]
[7]
N. Razavian, S. Blecker, A.M. Schmidt, A. Smith-McLallen, S. Nigam, and D. Sontag, "PopulationLevel Prediction of Type 2 Diabetes From Claims Data and Analysis of Risk Factors", Big Data, vol. 3, no. 4, pp. 277-287, 2015. [http://dx.doi.org/10.1089/big.2015.0020] [PMID: 27441408]
[8]
I. Devi, G.R. Karpagam, B.K. Vinoth. A survey of machine learning techniques. Interal Journal of Computational Systems Engineering, 2017, vol. 3, no. 4, pp. 203–212.
Machine Learning in Detection of Disease
Computational Intelligence for Data Analysis, Vol. 2 171
[http://dx.doi.org/10.1504/IJCSYSE.2017.089191] [9]
G. Ashish; T. Shigeyoshi Natural Computing Series (Advances in Evolutionary Computing). In: A Survey of Evolutionary Algorithms for Data Mining and Knowledge Discovery, Chapter 33, 99. 819–845, 2003. [http://dx.doi.org/10.1007/978-3-642-18965-4]
[10]
P. Chaudhari and H. Agarwal. ‘Progressive review towards deep learning techniques’, in Satapathy, E.S.C. (Ed.): Proceedings of the International Conference on Data Engineering and Communication Technology, Advances in Intelligent Systems and Computing, Springer Science + Business Media, Singapore, pp.151–158, 2007.
[11]
Z. Zhang, P. Chen, M. McGough, F. Xing, C. Wang, M. Bui, Y. Xie, M. Sapkota, L. Cui, J. Dhillon, N. Ahmad, F.K. Khalil, S.I. Dickinson, X. Shi, F. Liu, H. Su, J. Cai, and L. Yang, "Pathologist-level interpretable whole-slide cancer diagnosis with deep learning", Nat. Mach. Intell., vol. 1, no. 5, pp. 236-245, 2019. [http://dx.doi.org/10.1038/s42256-019-0052-1]
[12]
J.G. Richens, C.M. Lee, S. Johri, "Improving the accuracy of medical diagnosis with causal machine learning", Nat. Commun., Vol. 11, Article number: 3923, 2020. [http://dx.doi.org/10.1038/s41467-020-17419-7]
[13]
M.L. Graber, "The incidence of diagnostic error in medicine", BMJ Qual. Saf., vol. 22, no. Suppl 2, suppl. Suppl. 2, pp. ii21-ii27, 2013. [http://dx.doi.org/10.1136/bmjqs-2012-001615] [PMID: 23771902]
[14]
K.H. Yu, A.L. Beam, and I.S. Kohane, "Artificial intelligence in healthcare", Nat. Biomed. Eng., vol. 2, no. 10, pp. 719-731, 2018. [http://dx.doi.org/10.1038/s41551-018-0305-z] [PMID: 31015651]
[15]
A.F. Otoom, E.E. Abdallah, Y. Kilani, A. Kefaye, and M. Ashour, "Effective Diagnosis and Monitoring of Heart Disease", Int. J. Softw. Eng. Appl., vol. 9, pp. 143-156, 2015.
[16]
Y. Karaca, and C. Cattani, “7. Naive Bayesian classifier,” Comput. Methods Data Anal., pp. 229–250, 2018, [http://dx.doi.org/10.1515/9783110496369-007]
[17]
T.M. Lakshmi, A. Martin, R.M. Begum, and V.P. Venkatesan, "An Analysis on Performance of Decision Tree Algorithms using Student’s Qualitative Data", International Journal of Modern Education and Computer Science, vol. 5, no. 5, pp. 18-27, 2013. [http://dx.doi.org/10.5815/ijmecs.2013.05.03]
[18]
S. Pouriyeh, S. Vahid, G. Sannino, G. De Pietro, H. Arabnia, J. Gutierrez, et al. “ A Comprehensive Investigation and Comparison of Machine Learning Techniques in the Domain of Heart Disease”, 22nd IEEE Symposium on Computers and Communication (ISCC 2017): Workshops - ICTS4eHealth 2017.
[19]
Z. Arabasadi, R. Alizadehsani, M. Roshanzamir, H. Moosaei, and A.A. Yarifard, "Computer aided decision making for heart disease detection using hybrid neural network-Genetic algorithm", Comput. Methods Programs Biomed., vol. 141, pp. 19-26, 2017. [http://dx.doi.org/10.1016/j.cmpb.2017.01.004] [PMID: 28241964]
[20]
R. Mahajan, T. Viangteeravat, and O. Akbilgic, Improved detection of congestive heart failure via probabilistic symbolic pattern recognition and heart rate variability metrics., Int. J. Med. Inform., vol. 108, pp. 55-63, 2017. [http://dx.doi.org/10.1016/j.ijmedinf.2017.09.006] [PMID: 29132632]
[21]
S. Bai, K.F. Hew, and B. Huang, "Does gamification improve student learning outcome? Evidence from a meta-analysis and synthesis of qualitative data in educational contexts", Educ. Res. Rev., vol. 30, p. 100322, 2020. [http://dx.doi.org/10.1016/j.edurev.2020.100322]
172 Computational Intelligence for Data Analysis, Vol. 2
Rehman et al.
[22]
H. A. Esfahani and M. Ghazanfari, "Cardiovascular disease detection using a new ensemble classifier," in Proc. IEEE 4th Int. Conf. Knowl.- Based Eng. Innov. (KBEI), Dec. 2017, pp. 1011–1014. [http://dx.doi.org/10.1109/KBEI.2017.8324946]
[23]
A. Kondababu, V. Siddhartha, B.H.K.B. Kumar, and B. Penumutchi, "WITHDRAWN: A comparative study on machine learning based heart disease prediction", Mater. Today Proc., 2021. [http://dx.doi.org/10.1016/j.matpr.2021.01.475]
[24]
M. Diwakar, A. Tripathi, K. Joshi, M. Minakshi, P. Singh, N. Kumar, "Latest trends on heart disease prediction using machine learning and image fusion. Materials Today: Proceedings", vol. 37, Part 2, pp. 3213-3218, 2021. [http://dx.doi.org/10.1016/j.matpr.2020.09.078]
[25]
Heart Disease Detection Using Machine Learning Majority Voting Ensemble Method. IEEE, 2019.
[26]
H. Fujita, U.R. Acharya, V.K. Sudarshan, D.N. Ghista, S.V. Sree, L.W.J. Eugene, and J.E.W. Koh, "Sudden cardiac death (SCD) prediction based on nonlinear heart rate variability features and SCD index", Appl. Soft Comput., vol. 43, pp. 510-519, 2016. [http://dx.doi.org/10.1016/j.asoc.2016.02.049]
[27]
N. Khuriwal, N. Mishra. "Cancer Diagnosis Using Deep Learning". Proc - IEEE 2018 Int Conf Adv Comput Commun Control Networking, ICACCCN, pp. 98–103, 2018.
[28]
Z. Lai, and H. Deng, "Medical image classification based on deep features extracted by deep model and statistic feature fusion with multilayer perceptron", Comput. Intell. Neurosci., vol. 2018, pp. 1-13, 2018. [http://dx.doi.org/10.1155/2018/2061516] [PMID: 30298088]
[29]
A. Duggento, M. Scimeca, N. Urbano, E. Bonanno, M. Aiello, and C. Cavaliere, "A random initialization deep neural network for discriminating malignant breast cancer lesions", Proc Annu Int Conf IEEE Eng Med Biol Soc EMBS, pp. 912-5, 2019. [http://dx.doi.org/10.1109/EMBC.2019.8856740]
[30]
J. Yanase, and E. Triantaphyllou, "A systematic survey of computer-aided diagnosis in medicine: Past and present developments", Expert Syst. Appl., vol. 138, p. 112821, 2019. [http://dx.doi.org/10.1016/j.eswa.2019.112821]
[31]
K.J. Geras, R.M. Mann, and L. Moy, "Artificial intelligence for mammography and digital breast tomosynthesis: current concepts and future perspectives", Radiology, vol. 293, no. 2, pp. 246-259, 2019. [http://dx.doi.org/10.1148/radiol.2019182627] [PMID: 31549948]
[32]
Z. Hu, J. Tang, Z. Wang, K. Zhang, L. Zhang, and Q. Sun, "Deep learning for image-based cancer detection and diagnosis − A survey", Pattern Recognit., vol. 83, pp. 134-149, 2018. [http://dx.doi.org/10.1016/j.patcog.2018.05.014]
[33]
G. Carneiro, J. Nascimento, and A.P. Bradley, "Automated analysis of unregistered multi-view mammograms with deep learning", IEEE Trans. Med. Imaging, vol. 36, no. 11, pp. 2355-2365, 2017. [http://dx.doi.org/10.1109/TMI.2017.2751523] [PMID: 28920897]
[34]
M. Ahmed, Abdel-Zaher, Ayman M. Eldeib, Breast cancer classification using deep belief networks, expert systems with application. Elsevier, 2016.
[35]
W. Sun, T.L.B. Tseng, J. Zhang, and W. Qian, "Enhancing deep convolutional neural network scheme for breast cancer diagnosis with unlabeled data", Comput. Med. Imaging Graph., vol. 57, pp. 4-9, 2017. [http://dx.doi.org/10.1016/j.compmedimag.2016.07.004] [PMID: 27475279]
[36]
A. Teresa, A. Guilherme, C. Eduardo, R. Jos, A. Paulo, E. Catarina, P. Antnio, and C. Aurlio, "Classification of breast cancer histology images using ConvolutionalNeural Networks", PLoS One, pp. 1-14, 2017.
Machine Learning in Detection of Disease
Computational Intelligence for Data Analysis, Vol. 2 173
[37]
F. A. Spanhol, L. S. Oliveira, C. Petitjean and L. Heutte, "Breast cancer histopathological image classification using Convolutional Neural Networks," International Joint Conference on Neural Networks (IJCNN), Vancouver, BC, Canada, pp. 2560-2567, 2016.
[38]
M. Ram, A. Najafi, and M.T. Shakeri, "Classification and Biomarker Genes Selection for Cancer Gene Expression Data Using Random Forest", Iran. J. Pathol., vol. 12, no. 4, pp. 339-347, 2017. [http://dx.doi.org/10.30699/ijp.2017.27990] [PMID: 29563929]
[39]
B. Savelli, A. Bria, M. Molinara, C. Marrocco, and F. Tortorella, "A multi-context CNN ensemble for small lesion detection", Artif. Intell. Med.. vol. 103, p. 101749, 2020. [http://dx.doi.org/10.1016/j.artmed.2019.101749] [PMID: 32143786]
[40]
T.G. Debelee, A. Gebreselasie, F. Schwenker, M. Amirian, and D. Yohannes, "Classification of mammograms using texture and CNN based extracted features J", Journal of Biomimetics, Biomaterials and Biomedical Engineering, vol. 42, pp. 79-97, 2019. [http://dx.doi.org/10.4028/www.scientific.net/JBBBE.42.79]
[41]
H. Viswanath, L. Guachi-Guachi, and S. P. Thirumuruganandham, “EasyChair Preprint Breast Cancer Detection Using Image Processing Techniques and Classification Algorithms Breast Cancer Detection Using Image Processing Techniques and Classification Algorithms,” 2019.
[42]
M. Yousefi, A. Krzyżak, and C.Y. Suen, "Mass detection in digital breast tomosynthesis data using convolutional neural networks and multiple instance learning", Comput. Biol. Med., vol. 96, pp. 283293, 2018. [http://dx.doi.org/10.1016/j.compbiomed.2018.04.004] [PMID: 29665537]
[43]
M.A. Arpana, "Prathiba Kiran Feature extraction values for digital mammograms Int", J. Soft Comput. Eng., vol. 4, no. 2, 2014.
[44]
J. Yanase, and E. Triantaphyllou, "The seven key challenges for the future of computer-aided diagnosis in medicine", Int. J. Med. Inform., vol. 129, pp. 413-422, 2019. [http://dx.doi.org/10.1016/j.ijmedinf.2019.06.017] [PMID: 31445285]
[45]
J. Melnikow, J.J. Fenton, E.P. Whitlock, D.L. Miglioretti, M.S. Weyrich, J.H. Thompson, and K. Shah, "Supplemental screening for breast cancer in women with dense breasts: a systematic review for the us preventive services task force", Ann. Intern. Med., vol. 164, no. 4, pp. 268-278, 2016. [http://dx.doi.org/10.7326/M15-1789] [PMID: 26757021]
[46]
M. Hosni, I. Abnane, A. Idri, J.M. Carrillo de Gea, and J.L. Fernández Alemán, "Reviewing ensemble classification methods in breast cancer", Comput. Methods Programs Biomed., vol. 177, pp. 89-112, 2019. [http://dx.doi.org/10.1016/j.cmpb.2019.05.019] [PMID: 31319964]
[47]
A. Rampun, H. Wang, B. Scotney, P. Morrow, and R. Zwiggelaar, "School of Computing, Ulster University, Coleraine, Northern Ireland, UK Department of Computer Science, Aberystwyth University, UK", Conf. Image Process pp. 2072-2076, 2018.
[48]
M. Ezzat, and A. Idri, Reviewing Data Analytics Techniques in Breast Cancer Treatment. vol. 2. Springer International Publishing, 2020. [http://dx.doi.org/10.1007/978-3-030-45697-9_7]
[49]
S. Kaveeshwar, and J. Cornwall, "The current state of diabetes mellitus in India", Australas. Med. J., vol. 7, no. 1, pp. 45-48, 2014. [http://dx.doi.org/10.4066/AMJ.2014.1979] [PMID: 24567766]
[50]
A. Iyer, J. S, and R. Sumbaly, "Diagnosis of Diabetes Using Classification Mining Techniques", International Journal of Data Mining & Knowledge Management Process, vol. 5, no. 1, pp. 01-14, 2015. [http://dx.doi.org/10.5121/ijdkp.2015.5101]
[51]
S. Shah, X. Luo, S. Kanakasabai, R. Tuason, and G. Klopper, "Neural networks for mining the associations between diseases and symptoms in clinical notes", Health Inf. Sci. Syst., vol. 7, no. 1, p. 1,
174 Computational Intelligence for Data Analysis, Vol. 2
Rehman et al.
2019. [http://dx.doi.org/10.1007/s13755-018-0062-0] [PMID: 30588291] [52]
T. Chauhan, S. Rawat, S. Malik, P. Singh, "Supervised and Unsupervised Machine Learning based Review on Diabetes Care. 2021 7th International Conference on Advanced Computing and Communication Systems (ICACCS)", 2021.
[53]
Z.C. Lipton, D.C. Kale, and R. Wetzel, "Directly Modeling Missing Data in Sequences with RNNs: Improved Classification of Clinical Time Series", Proceedings of the 1st Machine Learning for Healthcare Conference, PMLR 56, . pp. 253-270, 2016.
[54]
S.K. Sen, and S. Dash, "Application of Meta Learning Algorithms for the Prediction of Diabetes Disease", International Journal of Advance Research in Computer Science and Management Studies, vol. 2, pp. 396-401, 2014.
[55]
A. Antoniadis, S. Lambert-Lacroix, and J.M. Poggi, "Random forests for global sensitivity analysis: A selective review", Reliab. Eng. Syst. Saf.. Vol. 206, Article ID: 107312, 2021. [http://dx.doi.org/10.1016/j.ress.2020.107312]
[56]
Z. Qin, L. Yan, H. Zhuang, Y. Tay, R.K. Pasumarthi,W. Xuanhui, M. Bendersky, M. Najork, "Are Neural Rankers still Outperformed by Gradient Boosted Decision Trees?" 2021.
[57]
V.S. Tseng, T.B. Ho, Z.-H. Zhou, L.P.C. Arbee, H.-Y. Kao. Lecture Notes in Computer Science. In: Advances in Knowledge Discovery and Data Mining. A Relevance Weighted Ensemble Model for Anomaly Detection in Switching Data Streams (Chapter 38), Vol. 8444, pp. 461–473, 2014. [http://dx.doi.org/10.1007/978-3-319-06605-9]
[58]
T. Sharma, and M. Shah, "A comprehensive review of machine learning techniques on diabetes detection", Visual Computing for Industry, Biomedicine, and Art, vol. 4, no. 1, p. 30, 2021. [http://dx.doi.org/10.1186/s42492-021-00097-7] [PMID: 34862560]
[59]
Z. Xu, and Z. Wang, "A risk prediction model for type 2 diabetes based on weighted feature selection of random forest and XGBoost ensemble classifier", pp. 278-283, 2019. [http://dx.doi.org/10.1109/ICACI.2019.8778622]
[60]
N. Sneha, and T. Gangil, "Analysis of diabetes mellitus for early prediction using optimal features selection", J. Big Data, vol. 6, no. 1, p. 13, 2019. [http://dx.doi.org/10.1186/s40537-019-0175-6]
[61]
M. Kriještorac, A. Halilović, and J. Kevric, "The impact of predictor variables for detection of diabetes mellitus type-2 for Pima Indians", Springer International Symposium on Innovative and Interdisciplinary Applications of Advanced Technologies, pp. 388-405, 2019.
[62]
B. Alić, L. Gurbeta, and A. Badnjević, "Machine learning techniques for classification of diabetes and cardiovascular diseases", Proceedings of 2017 6th mediterranean conference on embedded computing (MECO), . pp. 1-4, 2017. [http://dx.doi.org/10.1109/MECO.2017.7977152]
[63]
B. He, K.I. Shu, H. Zhang Machine learning and data mining in diabetes diagnosis and treatment IOP Conf. Ser.: Mater. Sci. Eng., vol. 490, no. 4, Article 042049, 2019.
[64]
B. Peter, The global burden of liver disease: a challenge for methods and for public health. BMC Med., Vol. 12, Article No. 159, 2014. [http://dx.doi.org/10.1186/s12916-014-0159-5]
[65]
L. A. Auxilia, Accuracy Prediction Using Machine Learning Techniques for Indian Patient Liver Disease. 2nd International Conference on Trends in Electronics and Informatics (ICOEI), 2018.
[66]
S. Vijayarani, and S. Dhayanand, "Liver Disease Prediction using SVM and Naive Bayes Algorithms. International Journal of Science", Engineering and Technology Research, vol. 4, pp. 816-820, 2015. [IJSETR].
[67]
UCI
Machine
Learning
Repository.
ILPD
(Indian
Liver
Patient
Dataset)
Data
Set.
Machine Learning in Detection of Disease
Computational Intelligence for Data Analysis, Vol. 2 175
https://archive.ics.uci.edu/ml/datasets/ILPD+(Indian+Liver +Patient+Dataset) [68]
A. Kendall, Y. Gal, What uncertainties do we need in Bayesian deep learning for computer vision? 2017. arXiv: 1703. 04977.
[69]
S. Naish, P. Dale, J.S. Mackenzie, J. McBride, K. Mengersen, and S. Tong, "Climate change and dengue: a critical and systematic review of quantitative modelling approaches", BMC Infect. Dis., vol. 14, no. 1, p. 167, 2014. [http://dx.doi.org/10.1186/1471-2334-14-167] [PMID: 24669859]
[70]
World Health Organization. Dengue and severe dengue. Fact Sheet. http://www.who.int/ mediacentre/factsheets/fs117/en/index/html (2017). Accessed 15 Jan 2017.
[71]
E. Fernández, M. Smieja, S.D. Walter, and M. Loeb, "A predictive model to differentiate dengue from other febrile illness", BMC Infect. Dis., vol. 16, no. 1, p. 694, 2016. [http://dx.doi.org/10.1186/s12879-016-2024-y] [PMID: 27876005]
[72]
G. Macedo Hair, F. Fonseca Nobre, and P. Brasil, "Characterization of clinical patterns of dengue patients using an unsupervised machine learning approach", BMC Infect. Dis., vol. 19, no. 1, p. 649, 2019. [http://dx.doi.org/10.1186/s12879-019-4282-y] [PMID: 31331271]
[73]
A.S. Fathima, and D. Manimeglai, "Predictive Analysis for the Arbovirus-Dengue using SVM Classification", IACSIT Int. J. Eng. Technol., vol. 2, pp. 521-527, 2012.
[74]
J. Yang, X. Chen, X. Deng, Z. Chen, H. Gong, H. Yan, Q. Wu, H. Shi, S. Lai, M. Ajelli, C. Viboud, and P.H. Yu, "Disease burden and clinical severity of the first pandemic wave of COVID-19 in Wuhan, China", Nat. Commun., vol. 11, no. 1, p. 5411, 2020. [http://dx.doi.org/10.1038/s41467-020-19238-2] [PMID: 33110070]
[75]
A. Alimadadi, Artificial intelligence and machine learning to fight COVID-19. American Physiological Society: Bethesda, 2020. [http://dx.doi.org/10.1152/physiolgenomics.00029.2020]
[76]
L.A. Amar, A.A. Taha, and M.Y. Mohamed, "Prediction of the final size for COVID-19 epidemic using machine learning: A case study of Egypt", Infect. Dis. Model., vol. 5, pp. 622-634, 2020. [http://dx.doi.org/10.1016/j.idm.2020.08.008] [PMID: 32864516]
[77]
C. L. Nikias, C. Hsing-Hsing, Higher-order spectrum estimation via noncausal autoregressive modeling and deconvolution, IEEE Trans Acoust Speech Signal Process, vol. 36, no. 12, pp. 1911–1913, 1988.
[78]
A.K. Abasi, A.T. Khader, M.A. Al-Betar, S. Naim, Z.A.A. Alyasseri, and S.N. Makhadmeh, "A novel hybrid multi-verse optimizer with K-means for text documents clustering", Neural Comput. Appl., vol. 32, no. 23, pp. 17703-17729, 2020. [http://dx.doi.org/10.1007/s00521-020-04945-0]
[79]
D.P. Kavadi, R. Patan, M. Ramachandran, and A.H. Gandomi, "Partial derivative Nonlinear Global Pandemic Machine Learning prediction of COVID 19", Chaos Solitons Fractals. p. 110056, 2020. [http://dx.doi.org/10.1016/j.chaos.2020.110056] [PMID: 32834609]
[80]
S. Rajaraman, J. Siegelman, P.O. Alderson, L.S. Folio, L.R. Folio, and S.K. Antani, "Iteratively pruned deep learning ensembles for COVID-19 detection in chest X-rays", IEEE Access, vol. 8, pp. 115041-115050, 2020. [http://dx.doi.org/10.1109/ACCESS.2020.3003810] [PMID: 32742893]
[81]
J. Zhu, B. Shen, A. Abbasi, M. Hoshmand-Kochi, H. Li, and T.Q. Duong, "Deep transfer learning artificial intelligence accurately stages COVID-19 lung disease severity on portable chest radiographs", PLoS One. p. e0236621, 2020. [http://dx.doi.org/10.1371/journal.pone.0236621] [PMID: 32722697]
176 Computational Intelligence for Data Analysis, Vol. 2 [82]
Rehman et al.
M. Agarwal, L. Saba, S.K. Gupta, A. Carriero, Z. Falaschi, A. Paschè, P. Danna, A. El-Baz, S. Naidu, and J.S. Suri, "A Novel Block Imaging Technique Using Nine Artificial Intelligence Models for COVID-19 Disease Classification, Characterization and Severity Measurement in Lung Computed Tomography Scans on an Italian Cohort", J. Med. Syst., vol. 45, no. 3, p. 28, 2021. [http://dx.doi.org/10.1007/s10916-021-01707-w] [PMID: 33496876]
Computational Intelligence for Data Analysis, 2023, Vol. 2, 177-193
177
CHAPTER 11
Breakthrough in Management of Cardiovascular Diseases by Artificial Intelligence in Healthcare Settings Lakshmi Narasimha Gunturu1,*, Raghavendra Naveen Nimbagal3
Girirajasekhar
Dornadula2
and
Scientimed Solutions Private Limited, Mumbai, Maharashtra, India Department of Pharmacy Practice, Annamacharya College of Pharmacy, Rajampeta, India 3 Department of Pharmaceutics, Sri Adichunchanagiri College of Pharmacy, Adichunchanagiri University, Karnataka 571418, India 1 2
Abstract: The cardiovascular system includes the heart and its associated blood vessels. Disorders of this cardiac system are called Cardiovascular disorders (CVD). Management of CVDs is often complex due to challenges like inadequate patient care, readmissions, low cost-effectiveness, and cost reductions in preventions, treatments, and lifestyle modifications. Hence, to overcome these challenges, Artificial Intelligence (AI) is being developed. They addressed emerging problems in clinical and health care settings and had a tremendous impact on the public. Implementation of AI in cardiovascular medicine affects more on new findings. It also provides a high level of supporting evidence that may be useful within the evidence-based research paradigm. A review of available free full-text literature in the PubMed database was carried out to study the influence of AI on health care settings. This work reviews AI-based algorithms used in cardiac practice and the applications of AI in cardiovascular medicine in terms of interpretation of results and medical image analysis.
Keywords: Algorithms, Applications, Artificial Intelligence, Cardiovascular Disorders, Healthcare. INTRODUCTION Artificial intelligence is a technology-based approach established in the mid1950s. It is the trendiest technology in the contemporary world due to the imitating nature of human intelligence. Artificial means discovered by humans and intelligence represents thinking ability. This covers the aspects like machinery/systems/algorithms based on beneficial results in making decisions. Corresponding author Lakshmi Narasimha Gunturu: Scientimed Solutions Private Limited, Mumbai, Maharashtra, India; E-mail: [email protected] *
Mariya Ouaissa, Mariyam Ouaissa, Zakaria Boulouad, Inam Ullah Khan, Sailesh Iyer (Eds.) All rights reserved-© 2023 Bentham Science Publishers
178 Computational Intelligence for Data Analysis, Vol. 2
Gunturu et al.
They work like the biological brains and maximize the success ratio [1]. Artificial intelligence is being used for various purposes in the present world. It can solve complex problems efficiently in multiple industries like Healthcare, finance, education, agriculture, robotics. AI is useful in the healthcare field and used in various fields like education, Astronomy, Agriculture, business, Surveillance, Social media, apps, Transportation, Gaming, Banking, and Electronic prescriptions. Cardiovascular diseases (CVDs) raise death concerns globally, which accounts for 17.9 million lives annually. CVDs constitute the pumping organ, heart, circulatory system like arteries, veins, capillaries, etc. Their damage may lead to fatal conditions, such as Coronary blockade, inflammatory conditions of heart valves leading to cardiomyopathy. Statistics provide data that most CVD deaths are mainly because of heart attacks, strokes, and nearly one-third of deaths in 70 years aged people (https://www.who.int/health-topics/cardiovascular-diseases). These are caused due to various factors like genetic, environmental, and lifestyle changes leading to the alteration of valuable microorganisms in the intestinal system of humans. The annual report of the American Heart Association (AHA) and the National Institute of Health confirmed possible causes for heart problems among the population are mentioned below: 1. Cigar inhalation, 2. Lack of Physical exercise, 3. Inappropriate food habits, 4. Weight gain, 5. Cholesterol, 6. Rise in systolic blood pressure (BP), 7. Elevated blood sugars [2]. There exists a two-way correlation allying technology as well as health. To explore the challenges among advanced and growing countries, they utilize innovation as a prime focus to cope with the problems. Hence technology-based approaches especially AI in health care, slowly drifted the entire medical system. It also transformed the hospitals, lives of the public. For example, they aid in interpreting results, lessening the test, and allowing the public to take their verdicts.
Management of Cardiovascular Diseases
Computational Intelligence for Data Analysis, Vol. 2 179
The use of expert systems within the cardiology department is limited to only some countries for various reasons; however, instead of its limitations, it finds very useful. Using such approaches, we can analyze the disease models efficiently, and also a considerable amount of data can be easily saved in the databases. Advantages with the integration of AI in cardiovascular settings are described in Fig. (1).
Fig. (1). Advantages with the integration of AI in cardiovascular.
However, this version needs several fixed hypotheses that do not depend on factors like observations and other multicollinearity values. The conventional logistic regression method improves the statistical value, but it is sometimes independent towards the target, therefore, obstructing the model’s execution. On the other side, AI algorithms are used in various clinical departments as they provide results exactly by using the information stored in the data sets. The impact of computational intelligence in cardiology clinics changed cardiac clinic experimentation patterns and, thus valuable for clinical practice and general public health, as shown in Fig. (2) [3].
180 Computational Intelligence for Data Analysis, Vol. 2
Gunturu et al.
Fig. (2). Impact of AI on cardiology.
In the contemporary world, AI-based approaches have become popular in preventing and managing cardiovascular disorders. In implementing AI-based solutions to assess the risk factors and detect the early signs of CVD risk, machine learning has gained the special spotlight. Due to this reason, applications of machine learning are more comprehensive in CVD diagnosis. For example, the strengths, challenges, and potential drawbacks faced by machine learning in the condition of cardiovascular risk estimation are clearly explained by Hodos et al. [4]. In another work of El-Saadawy et al., regression classifiers models such as SoftMax, Support Vector Machine (SVM), and Probabilistic Neural Network are utilized to diagnose heart diseases [5]. Maheswari et al. explained the cardiac classifier and models like a decision tree and Naive Bayes, which are helpful in heart disease prediction [6]. Another work of Sasko Ristov et al. [7] explained the management of cardiovascular disease using CVD risk factor analysis that is acclaimed internationally. Jabbar et al. in their work proposed a combination model of K-nearest neighbor (KNN) and Genetic algorithms, which is effective in predicting cardiovascular risk factors and their management [8]. Dominic et al. work explained the heart disease datasets and AI-based machine learning technologies that help determine the level of risk prediction in heart diseases [9]. Das et al. developed a diagnostic method for heart diseases based on neural network integration through SAS software [10].
Management of Cardiovascular Diseases
Computational Intelligence for Data Analysis, Vol. 2 181
This chapter focus on the recent developments in the applications of AI in cardiovascular medicine, AI algorithms used in the medical field towards the disease or risk factors identification, complications minimized by the implementation of AI in cardiology, and their challenges are discussed. MATERIALS AND METHODS a. Creating a Systematic Search Strategy. b. An appropriate question was constructed. c. An appropriate database Pub Med was used to explore the results. A research question is uncertain about a problem; hence it must be carefully analyzed and examined to provide better clarity. A successful work depends upon how a better investigator formulates the research question to solve the existing problem. A research question mainly compares two treatments or diseases, a suitable diagnostic technique, etc. [11]. The PICO (patients, interventions, comparator, and outcomes) is widely considered the strategy to formulate a background question. Following are the list of questions formulated: a. Do AI algorithms really beneficial in the prediction of risk associated with heart diseases? b. Can AI algorithms help to interpret the diagnostic techniques (Electrocardiogram, Magnetic resonance imaging, etc.) better and improve treatment outcomes? c. Finally, does integrating AI techniques in healthcare settings change the shift in cardiovascular risk reduction? Table 1. PICO included in the research question formulation. Parameter
Included
Excluded
Patients (P)
Older age
Younger age (< 35 years).
Intervention (I)
Influence of AI
Other technology applications like fuzzy systems, 5G.
Comparator (C)
Cardiovascular patients
Non-cardiovascular patients
Outcome (O)
AI-based effective control of cardiovascular complications
Death, Late diagnosis of the result, Other co-morbid patients.
Table 1 shows a question-searching strategy based on data obtained from Pub Med database that was searched with the Keywords Healthcare, Cardiology, Artificial Intelligence, and Algorithms till Oct 20, 2020, from the past ten years.
182 Computational Intelligence for Data Analysis, Vol. 2
Gunturu et al.
Title and abstract screening were done manually. Full-text articles of these screened articles were further included for the inclusion and exclusion criteria. We included all the studies of AI, and cardiovascular complications with the age group above 35 years and excluded the younger age population and other technology applications like fuzzy systems, 5G network applications, and other co-morbid patients. We involved 11 studies that evaluated the use of AI to prevent cardiovascular complications. The structural outline of our work is shown in Fig. (3).
Fig. (3). The search strategy included in the study.
ALGORITHMS USED IN CARDIOVASCULAR DISEASES As discussed above, a non-invasive medical decision-supporting system was developed using machine learning techniques to reduce the treatment costs and complexity involved in treatment procedures and provide an accurate diagnosis. Those are considered as Algorithms that are most popularly used in the prediction of heart diseases. Some of the most popular algorithms used in clinical applications of the cardiology field are discussed below: 1. K- nearest neighbour (K-NN) 2. Artificial neural network (ANN) 3. Decision tree (DT) 4. Logistic regression (LR)
Management of Cardiovascular Diseases
Computational Intelligence for Data Analysis, Vol. 2 183
5. AdaBoost (AB) 6. Support vector machine (SVM) K-Nearest Neighbour (KNN) It is a machine-learning algorithm based on a supervised learning technique. This algorithm acts as a repository for entirely all cases and is classified on the roots of the resemblance. It is easy to handle and accurate. In K-NN, K denotes the notation for nearest neighbours. Here the principal value is K, and when it is equal to one is described as the KNN. This algorithm resolves the Classification and regression predictive issues. The study of Jabbar reported that KNN incorporated with Particle Swam Optimization (PSO) technique could effectively predict heart diseases with 100% accuracy. Medical data consists of both relevant and unwanted information; therefore, most of the irrelevant and unwanted information can be discarded using the PSO technique in the KNN model [6]. Artificial Neural Network (ANN) This type of programming and data processing system is similar to living neurons of animal intellect. It has also been referred to as Connectionist systems. It consists of nodes called neurons that mimic nerve cells of the human cerebrum. Every link looks the same in the way of the myoneural junctions of the human cerebrum that process information to the following nerve cells or neurons. In a simple way, the signal is received and transmitted to the next level of nerve cells, making all the neuronal systems activate [12]. In a recent study by Niranjana et al., it was reported that an ANN model with one layer of trained hidden neuronal architecture can be able to identify the degree related to heart block features of heart valves with an accuracy of 95.5% [13]. Decision Tree (DT) DT is maybe simple and provides us excellent results when the information is usually categorical and depends on conditions. It is often explained with an example as if an individual had systolic and diastolic dysfunction due to the chronic activation of hemodynamic and neuron-hormonal compensatory responses. This condition is often further divided into branches (causes) like cardiac hypertrophy, Myocyte loss due to necrosis, and abnormal myocardial energy until they are available to choose the exact reason for the person to develop systolic, and diastolic dysfunction. The conditions are referred to as the interior nodes, and they split to return to a choice which is understood as the leaf.
184 Computational Intelligence for Data Analysis, Vol. 2
Gunturu et al.
Logistic Regression (LR) This algorithm is used when examining data consisting of more than one variable. These variables are responsible for the final result [14]. AdaBoost (AB) Adaboost converts weak learners to strong learners and is one of the boosting algorithms utilized in machine learning algorithms. It starts by identifying the original information from each set and provides equal weight to every observation. In Li et al. study, it was reported that Adaboost could estimate the multilevel risk prediction of cardiovascular diseases by using precision and accuracy values [15]. Support Vector Machine (SVM) It is helpful to differentiate information depending on various factors or causes. It utilizes hyperplane as a sort of conclusion among many options. It also brings out information isolated by many hyperplanes, and this data was divided into portions; in turn, every portion was confined to a single data set. In the study by Ashtiyani et al., it is reported that algorithms developed with SVM are often ready to track the variations in a heartbeat and thus helpful in identifying cardiac arrhythmias [16]. Other algorithms that come into the role are supervised algorithms and unsupervised algorithms. The Supervised algorithm requires predictable variables and labeled outcomes. For example, if it is desired to predict whether a patient is more susceptible to developing angioedema using amlodipine. This analysis should be performed based on the health data sets of patients that showed such a reaction and another group in which this reaction was not observed. Unsupervised learning primary variables are not provided, thus identifying unhidden structural layers. According to radiological findings, the best example of unsupervised algorithms is a characterization of the database of patients with myocardial infarction. Table 2 lists the algorithms and their use in cardiology. Table 2. Machine learning algorithms used in cardiology. Algorithm
Description
Use
K nearest neighbour (KNN)
A machine learning and non-parametric algorithm.
Prediction of heart diseases.
Artificial Neural Networks (ANN)
It is similar to the human cerebrum.
Predict the degree related to angiographic features in coronary heart diseases.
Management of Cardiovascular Diseases
Computational Intelligence for Data Analysis, Vol. 2 185
(Table 2) cont.....
Algorithm Decision tree Adaboost Support vector machine
Description
Use
Here the data is categorized depending on Useful to decide what is the appropriate the various choices available. cause for the diseases. It can able to convert weak classifiers into strong classifiers.
Multilevel risk prediction of cardiovascular heart diseases.
The Support vector machine concludes as Able to detect variations in heart rate different classes. New results are classified and applicable in arrhythmic conditions. using this.
RESULTS AND DISCUSSION We included 11 different studies, as presented in Table 3, that concluded the use of AI in the detection of various cardiovascular complications. These studies proved that the implementation of AI in cardiology would bring rapid transformation in the health and well-being of society. Table 3. Summary of studies that used AI to estimate cardiac complications. Study
Algorithm Used
Complications Resolved
Sengupta et al. [17]
Cognitive-based Learning Algorithm.
This algorithm can effectively distinguish between constructive heart failure and Restrictive heart failure.
Narula et al. [18]
Machine Algorithm based on supervised futures.
It can able to detect the athlete’s heart from cardiomyopathy.
Ouyang et al. [19]
Deep Learning Algorithm.
3D-enabled view for the determination of the electrical activity of the heart.
Avendi et al. [20]
Convolution Neuronal Networks.
This model can detect the Myocardial scars of the ventricle.
Dawes et al. [21]
Machine learning programming.
It enables us to study the systolic patterns of the heart.
Coenen et al. [22]
Machine learning-based Image analysis.
Determine risk assessment in heart blockage.
Wolterink et al. [23]
Convolutional Neuronal Network.
This algorithm can effectively determine the presence of calcium in high amounts in the Coronary arteries.
Gonzales et al. [24]
Convolutional Neuronal networkbased AI algorithm.
Use to calculate the Agatson score.
Mannil et al. [25]
Machine learning-based texture image analysis.
This can able to detect the presence of MI even in the non-contrast medium.
Attia et al. [26]
A 6-layered neuronal network membrane.
It can able to identify left ventricular dysfunction.
Stehlik et al. [27]
Wearable machine learning models.
To predict Heart failure exacerbation.
186 Computational Intelligence for Data Analysis, Vol. 2
Gunturu et al.
Impact of AI on Echocardiography (ECG) An Echocardiogram is the most important in the cardiology field for accurately detecting and treating heart abnormalities; hence, appraisal of heart functioning can be judged. Technology-based AI tools come up with new possibilities to amplify the precision of the likeliness of image elucidation in ECG, especially when there are chances of misdiagnosis to identify a disease. AI-based computational tools had the possibility to explicate unutilized data in a programmed way by the emergence of advanced processes like Speckle tracking imaging. It further benefits in a cutback of the investigation period or duration and also improves reliability [28, 29]. A study by Sengupta et al. came about a cognitive-based learning algorithm coupled with speckle tracking performance that can distinguish constrictive heart failure from restrictive heart failure [17]. Another report by Narula et al. concluded that machine learning based on supervised features could distinguish abnormalities in the functioning of an athlete’s heart with increased precision in contrast to conventional computational methods [18]. Nevertheless, ML models can also detect heart valvular motions to identify and cure ailments, mainly during the pre-operative times [30, 31]. Recent three-dimensional machine learning Echonet dynamic proposed by Ouyang et al. achieved a reliability of nearly 0.92% that can evaluate ECG irregularities with accuracy better than cardiac experts [19]. Role of AI on Magnetic Resonance Imaging (MRI) ML models are more frequently used in the cardiac MRI-like ventricular segmentation area. Using ML models, we can enhance the volumetric and reliability of clinical findings [20, 32]. The deep learning algorithm suggested by Avendi et al. could able to identify the existing myocardium scars in the right portion of ventricles. This detection of sub-acute or chronic myocardial scar is another area where ML models are applied [33]. In another study by Dawes et al., he used the three-dimensional portion of the heart muscle to foresee the traditional perils, and unfavorable consequences like premature demise in patients with pulmonary diseases [21]. Use of AI on Cardiac Computed Tomography (CT) Machine learning features are progressively utilized in CT examination for identification together with evaluation of risk in coronary infarction. It is also used in the identification of calcium presence and flow of blood in the blood
Management of Cardiovascular Diseases
Computational Intelligence for Data Analysis, Vol. 2 187
vessels. Computed tomography type of non-invasive procedure has the potential to detect the narrowing of blood vessels due to varied reasons like accumulation of plaque, inflammatory reactions, thrombus formation, etc. [22]. Consequently, many programming models have evolved to identify non-invasive flow rates to increase CT functioning by estimating the exact cause of the illness [34]. The ability to detect plaque formation using AI-based devices comes with additional benefits in hospital settings and lowering false positives. A study by Wolterink et al. uses machine learning programming with supervised features that can recognize and enumerate defects in coronary arteries [23, 35]. A study published by Gonzales et al. utilized neuronal networks used to figure out the presence of calcium deposits in the heart that causes calcification [24]. Another side use of machine learning-dependent cardiac CT is used to spot infarction even in a non-contrast medium [25]. Impact of AI on Electrocardiography This application of AI points out the irregularities in heart function. They could also involuntarily identify heartbeat variations to minimize the time for image analysis and independent unevenness [36, 37]. A study reported by Lyon et al. related and graded the ECG makeup correlated to arrhythmia. Another study by Attia et al. utilized a six-layered neuronal membrane to detect the existence of ventricular dysfunction [26]. Another study by LINK-HF concluded that AI-based programming devices could precisely estimate the failure of heart aggravation in contrast to invasive tools [27]. Table 4 lists the prediction models used in cardiovascular diseases. Table 4. Prediction models used in the detection of cardiovascular diseases. Year Country
Version of Algorithm
Prognostic Result
Method Utilized
Predicted Factors
1997
Japan
Artificial Neuronal Network
ECG analysis for QS complex
Programmed Alterations in Q electrocardiograms and R waves and the V2 and V7 lead fluctuations.
2005
US
ANN
Authentic protection for identification of paediatric heart sounds.
Phonocardiograms
Reverberation of recordings of heart.
Diagnosis ST-segment elevation myocardial infarction. Heart whispers.
188 Computational Intelligence for Data Analysis, Vol. 2
Gunturu et al.
(Table 4) cont.....
Year Country
Version of Algorithm
Prognostic Result
Method Utilized
Predicted Factors
Diagnosis -
2016
US
A flexible investigative algorithm.
Computerized 3D enumeration of trans thoracic echocardiography obtained atria and ventricle volumes.
Echocardiography; electrocardiogram
Size of heart chambers, namely atria and ventricles.
2018
US
Markov prototype
An unprecedented m-Health facility plan.
Listening to heart sounds through mobile phones.
Cardiac sounds, Heart valves irregular regurgitation, heartbeats mitral valve prolapse
AI also includes a vast group of technology features that includes the following: 1. Cloud technology 2. Telemedicine 3. Mobile health 4. Big data, etc. All these technological features are integrated with the most widely used mobilebased apps. Such apps, coupled with AI-based features, help check the vitals of patients. These apps are primarily two types Patient-related and Physician related. The patients to know their body vitals directly use patient-related apps. These are not supervised by the physician; later are doctor-mediated apps that are useful for the physicians to check the patient's body condition and use as a tool to save the information in the medical literature. These are also useful in calculating the formulas in cardiovascular medicine, predicting the ECG features, monitoring blood glucose levels, the temperature of the body, etc. The best example of this is Heart 360 or cardio brilliant 360. It consists of AI-based integrated features with web-based apps and mobile-based applications connected to smartphones connected with AI-based decision-supporting systems. This AI-based app provides the following features: 1. It helps in maintaining the electronic medical record of patients where it contains the complete patient details, treatment history, and medical and medication records from a single hospital. For the prediction of heart diseases and future purposes, cardiologists would easily access such record-based data. 2. This app can record the basic patient vitals such as Blood pressure, Glucose levels, temperature, etc., which are easily monitored and evaluated by the
Management of Cardiovascular Diseases
Computational Intelligence for Data Analysis, Vol. 2 189
physician in the device that is connected to this app even in the patient’s absence in the clinics or hospitals. Simply this can be easily explained, as remote monitoring of patient vitals is possible with this app. 3. It can also detect abnormal values or vitals that cross the threshold limits during pathological conditions or in some other adverse drug reactions. So that fatal conditions can be averted. 4. It also contains various modules for Cardiogenic Patients modules and Message management modules that aid cardiologists in clinical decision-making and precise estimation of heart diseases. This is used as a potential tool in areas like health care and also a research application tool for estimating the properties of the heart [38]. CHALLENGES Although AI is well established for its beneficial role in cardiovascular medicine, there are still many hurdles to overcome to implement AI-based technological features in clinical settings. Firstly, the main problem is the lack of infrastructure in the hospitals. As we know, most clinics do not have a spacious infrastructure; therefore, the role of AI is underrated. Another critical factor is the lack of welltrained technicians who can handle advanced technology features. In the Indian scenario, the entire patient data is stored in the various database platforms, so even if the AI is established in such clinics, it would take more time to analyze the patient characteristics as data is scattered into the different systems. Another essential thing to consider is Ethical issues. Ethical concerns and their usage is the vital thing in AI. Sometimes, by using AI technology, we can predict the fortunate in cases like determining the sex of the baby and estimating whether a person is prone to cardiovascular diseases. Therefore, this will undoubtedly lead to illegal conduct. Another essential thing is the fear of loss of employment if AI is much occupied in the healthcare settings. However, we can assure you that by implementing AI devices, identifying the diseases or their features would have been easy for physicians for precise and accurate decision-making. However, they cannot replace humans because the data has to be analyzed by the humans at the end, whatever the computerized machine generates it. Therefore, it must be considered a myth about employment loss. With the advent of these technological features in clinical medicine, one must consider safe and reliable instruments for patient care.
190 Computational Intelligence for Data Analysis, Vol. 2
Gunturu et al.
CONCLUSION There is a huge amount of patient data to analyze and interpret in clinical practice. Hence, the best way is the inclusion of AI and its algorithms in clinical practice. This would help physicians make decisions more accurately, which improves patient care. The use of AI has proliferated in recent years. Its applications in various departments like Cardiology, Ophthalmology, and Pathologic findings are evident, thus leading to change in health care. The applications of AI by cardiologists need additional developments because it is still in the budding stage. Apart from its uses and applications, there are also limitations, such as ethical issues, dealing with errors, database management, etc. However, the use of AIbased mathematical models integrated with AI helps physicians in the clinical practice to identify and characterize patient disorders that finally lead to an accurate diagnosis of disease at greater depths. Therefore, the inclusion of AI is not something that should not be feared; instead, it is a change that must be accepted. CONSENT FOR PUBLICATON Declared none. CONFLICT OF INTEREST The author declares no conflict of interest, financial or otherwise. ACKNOWLEDGEMENT Declared none. REFERENCES [1]
B. Mondal, "Artificial intelligence: state of the art,” in Intelligent Systems", Ref. Libr., vol. 172, pp. 389-425, 2019.
[2]
S.S. Virani, A. Alonso, E.J. Benjamin, M.S. Bittencourt, C.W. Callaway, A.P. Carson, A.M. Chamberlain, A.R. Chang, S. Cheng, F.N. Delling, L. Djousse, M.S.V. Elkind, J.F. Ferguson, M. Fornage, S.S. Khan, B.M. Kissela, K.L. Knutson, T.W. Kwan, D.T. Lackland, T.T. Lewis, J.H. Lichtman, C.T. Longenecker, M.S. Loop, P.L. Lutsey, S.S. Martin, K. Matsushita, A.E. Moran, M.E. Mussolino, A.M. Perak, W.D. Rosamond, G.A. Roth, U.K.A. Sampson, G.M. Satou, E.B. Schroeder, S.H. Shah, C.M. Shay, N.L. Spartano, A. Stokes, D.L. Tirschwell, L.B. VanWagner, and C.W. Tsao, "Heart disease and stroke statistics—2020 update: A report from the American Heart Association", Circulation, vol. 141, no. 9, pp. e139-e596, 2020. [http://dx.doi.org/10.1161/CIR.0000000000000757] [PMID: 31992061]
[3]
K.W. Johnson, J. Torres Soto, B.S. Glicksberg, K. Shameer, R. Miotto, M. Ali, E. Ashley, and J.T. Dudley, "Artificial Intelligence in Cardiology", J. Am. Coll. Cardiol., vol. 71, no. 23, pp. 2668-2679, 2018. [http://dx.doi.org/10.1016/j.jacc.2018.03.521] [PMID: 29880128]
[4]
R.A. Hodos, B.A. Kidd, K. Shameer, B.P. Readhead, and J.T. Dudley, In silico methods for drug
Management of Cardiovascular Diseases
Computational Intelligence for Data Analysis, Vol. 2 191
repurposing and pharmacology., Wiley Interdiscip. Rev. Syst. Biol. Med., vol. 8, no. 3, pp. 186-210, 2016. [http://dx.doi.org/10.1002/wsbm.1337] [PMID: 27080087] [5]
H. El-Saadawy, M. Tantawi, H. A. Shedeed, and M. F. Tolba, “Electrocardiogram (ECG) heart disease diagnosis using PNN, SVM and Softmax regression classifiers,” in 2017 IEEE 8th International Conference on Intelligent Computing and Information Systems, pp. 106–110, 2017 [http://dx.doi.org/10.1109/INTELCIS.2017.8260040]
[6]
S. Maheswari, and R. Pitchai, "Heart Disease Prediction System Using Decision Tree and Naive Bayes Algorithm", Curr. Med. Imaging Rev., vol. 15, no. 8, pp. 712-717, 2019. [http://dx.doi.org/10.2174/1573405614666180322141259] [PMID: 32008540]
[7]
S. Ristov and A. Pečkov, “Machine Learning Approach for Early Detection of Cardiovascular Deceases (CVD),” In: ICT Innovations, pp. 41-50, 2010.
[8]
M.A. Jabbar, "Prediction of heart disease using k-nearest neighbor and particle swarm optimization", Biomed. Res. (Aligarh), vol. 28, no. 9, pp. 4154-4158, 2017.
[9]
V. Dominic, D. Gupta, and S. Khare, "An Effective Performance Analysis of Machine Learning Techniques for Cardiovascular Disease", Appl. Med. Informatics Orig. Res., vol. 36, no. 1, pp. 23-32, 2015.
[10]
R. Das, I. Turkoglu, and A. Sengur, "Effective diagnosis of heart disease through neural networks ensembles", Expert Syst. Appl., vol. 36, no. 4, pp. 7675-7680, 2009. [http://dx.doi.org/10.1016/j.eswa.2008.09.013]
[11]
S. Aslam, and P. Emmanuel, "Formulating a researchable question: A critical step for facilitating good clinical research", Indian J. Sex. Transm. Dis., vol. 31, no. 1, pp. 47-50, 2010. [http://dx.doi.org/10.4103/0253-7184.69003] [PMID: 21808439]
[12]
M. Ismail, V. Harsha Vardhan, V. Aditya Mounika, and K. Surya Padmini, "An effective heart disease prediction method using artificial neural network", Int. J. Innov. Technol. Explor. Eng., vol. 8, no. 8, pp. 1529-1532, 2019.
[13]
"ANN Model to Predict Coronary Heart Disease Based on Risk Factors", Bonfring Int. J. Man Mach. Interface, vol. 3, no. 2, pp. 13-18, 2013. [http://dx.doi.org/10.9756/BIJMMI.4473]
[14]
O. Miguel-Hurtado, R. Guest, S.V. Stevenage, G.J. Neil, and S. Black, "Comparing machine learning classifiers and linear/logistic regression to explore the relationship between hand dimensions and demographic characteristics", PLoS One, vol. 11, no. 11, p. e0165521, 2016. [http://dx.doi.org/10.1371/journal.pone.0165521] [PMID: 27806075]
[15]
R. Li, S. Shen, G. Chen, T. Xie, S. Ji, B. Zhou, and Z. Wang, "Multilevel Risk Prediction of Cardiovascular Disease based on Adaboost+RF Ensemble Learning", IOP Conf. Series Mater. Sci. Eng., vol. 533, no. 1, p. 012050, 2019. [http://dx.doi.org/10.1088/1757-899X/533/1/012050]
[16]
M. Ashtiyani, S. Navaei Lavasani, A. Asgharzadeh Alvar, and M.R. Deevband, "Heart rate variability classification using support vector machine and genetic algorithm", J. Biomed. Phys. Eng., vol. 8, no. 4, pp. 423-434, 2018. [http://dx.doi.org/10.31661/jbpe.v0i0.614] [PMID: 30568932]
[17]
P.P. Sengupta, Y.M. Huang, M. Bansal, A. Ashrafi, M. Fisher, K. Shameer, W. Gall, and J.T. Dudley, "Cognitive Machine-Learning Algorithm for Cardiac Imaging", Circ. Cardiovasc. Imaging, vol. 9, no. 6, p. e004330, 2016. [http://dx.doi.org/10.1161/CIRCIMAGING.115.004330] [PMID: 27266599]
[18]
S. Narula, K. Shameer, A.M. Salem Omar, J.T. Dudley, and P.P. Sengupta, "Machine-Learning Algorithms to Automate Morphological and Functional Assessments in 2D Echocardiography", J. Am. Coll. Cardiol., vol. 68, no. 21, pp. 2287-2295, 2016.
192 Computational Intelligence for Data Analysis, Vol. 2
Gunturu et al.
[http://dx.doi.org/10.1016/j.jacc.2016.08.062] [PMID: 27884247] [19]
D. Ouyang, B. He, A. Ghorbani, N. Yuan, J. Ebinger, C.P. Langlotz, P.A. Heidenreich, R.A. Harrington, D.H. Liang, E.A. Ashley, and J.Y. Zou, "Video-based AI for beat-to-beat assessment of cardiac function", Nature, vol. 580, no. 7802, pp. 252-256, 2020. [http://dx.doi.org/10.1038/s41586-020-2145-8] [PMID: 32269341]
[20]
M.R. Avendi, A. Kheradvar, and H. Jafarkhani, "Automatic segmentation of the right ventricle from cardiac MRI using a learning-based approach", Magn. Reson. Med., vol. 78, no. 6, pp. 2439-2448, 2017. [http://dx.doi.org/10.1002/mrm.26631] [PMID: 28205298]
[21]
T.J.W. Dawes, A. de Marvao, W. Shi, T. Fletcher, G.M.J. Watson, J. Wharton, C.J. Rhodes, L.S.G.E. Howard, J.S.R. Gibbs, D. Rueckert, S.A. Cook, M.R. Wilkins, and D.P. O’Regan, "Machine learning of threedimensional right ventricular motion enables outcome prediction in pulmonary hypertension: A cardiac MR imaging study", Radiology, vol. 283, no. 2, pp. 381-390, 2017. [http://dx.doi.org/10.1148/radiol.2016161315] [PMID: 28092203]
[22]
A. Coenen, Y.H. Kim, M. Kruk, C. Tesche, J. De Geer, A. Kurata, M.L. Lubbers, J. Daemen, L. Itu, S. Rapaka, P. Sharma, C. Schwemmer, A. Persson, U.J. Schoepf, C. Kepka, D. Hyun Yang, and K. Nieman, "Diagnostic accuracy of a machine-learning approach to coronary computed tomographic angiography–Based fractional flow reserve result from the MACHINE Consortium", Circ. Cardiovasc. Imaging, vol. 11, no. 6, p. e007217, 2018. [http://dx.doi.org/10.1161/CIRCIMAGING.117.007217] [PMID: 29914866]
[23]
J.M. Wolterink, T. Leiner, B.D. de Vos, R.W. van Hamersvelt, M.A. Viergever, and I. Išgum, "Automatic coronary artery calcium scoring in cardiac CT angiography using paired convolutional neural networks", Med. Image Anal., vol. 34, pp. 123-136, 2016. [http://dx.doi.org/10.1016/j.media.2016.04.004] [PMID: 27138584]
[24]
G. González, G.R. Washko, R.S.J. Estépar, M. Cazorla, and C. Cano Espinosa, "Automated Agatston score computation in non-ECG gated CT scans using deep learning", Proc SPIE Int Soc Opt Eng, 2018p. 10574 [http://dx.doi.org/10.1117/12.2293681]
[25]
M. Mannil, J. von Spiczak, R. Manka, and H. Alkadhi, "Texture Analysis and Machine Learning for Detecting Myocardial Infarction in Noncontrast Low-Dose Computed Tomography", Invest. Radiol., vol. 53, no. 6, pp. 338-343, 2018. [http://dx.doi.org/10.1097/RLI.0000000000000448] [PMID: 29420321]
[26]
Z.I. Attia, S. Kapa, F. Lopez-Jimenez, P.M. McKie, D.J. Ladewig, G. Satam, P.A. Pellikka, M. Enriquez-Sarano, P.A. Noseworthy, T.M. Munger, S.J. Asirvatham, C.G. Scott, R.E. Carter, and P.A. Friedman, "Screening for cardiac contractile dysfunction using an artificial intelligence–enabled electrocardiogram", Nat. Med., vol. 25, no. 1, pp. 70-74, 2019. [http://dx.doi.org/10.1038/s41591-018-0240-2] [PMID: 30617318]
[27]
J. Stehlik, C. Schmalfuss, B. Bozkurt, J. Nativi-Nicolau, P. Wohlfahrt, S. Wegerich, K. Rose, R. Ray, R. Schofield, A. Deswal, J. Sekaric, S. Anand, D. Richards, H. Hanson, M. Pipke, and M. Pham, "Continuous wearable monitoring analytics predict heart failure hospitalization: The link-hf multicenter study", Circ. Heart Fail., vol. 13, no. 3, p. e006513, 2020. [http://dx.doi.org/10.1161/CIRCHEARTFAILURE.119.006513] [PMID: 32093506]
[28]
M. Alsharqi, W.J. Woodward, J.A. Mumith, D.C. Markham, R. Upton, and P. Leeson, "Artificial intelligence and echocardiography", Echo Res. Pract., vol. 5, no. 4, pp. R115-R125, 2018. [http://dx.doi.org/10.1530/ERP-18-0056] [PMID: 30400053]
[29]
K. Seetharam, N. Kagiyama, and P.P. Sengupta, "Application of mobile health, telemedicine and artificial intelligence to echocardiography", Echo Res. Pract., vol. 6, no. 2, pp. R41-R52, 2019. [http://dx.doi.org/10.1530/ERP-18-0081] [PMID: 30844756]
[30]
A. Sengur, "Support vector machine ensembles for intelligent diagnosis of valvular heart disease", J.
Management of Cardiovascular Diseases
Computational Intelligence for Data Analysis, Vol. 2 193
Med. Syst., vol. 36, no. 4, pp. 2649-2655, 2012. [http://dx.doi.org/10.1007/s10916-011-9740-z] [PMID: 21590303] [31]
C. Martin-Isla, V.M. Campello, C. Izquierdo, Z. Raisi-Estabragh, B. Baeßler, S.E. Petersen, and K. Lekadir, "Image-Based Cardiac Diagnosis With Machine Learning: A Review", Front. Cardiovasc. Med., vol. 7, pp. 1-19, 2020. [http://dx.doi.org/10.3389/fcvm.2020.00001] [PMID: 32039241]
[32]
L.K. Tan, R.A. McLaughlin, E. Lim, Y.F. Abdul Aziz, and Y.M. Liew, "Fully automated segmentation of the left ventricle in cine cardiac MRI using neural network regression", J. Magn. Reson. Imaging, vol. 48, no. 1, pp. 140-152, 2018. [http://dx.doi.org/10.1002/jmri.25932] [PMID: 29316024]
[33]
B. Baessler, M. Mannil, S. Oebel, D. Maintz, H. Alkadhi, and R. Manka, "Subacute and chronic left ventricular myocardial scar: Accuracy of texture analysis on nonenhanced cine MR images", Radiology, vol. 286, no. 1, pp. 103-112, 2018. [http://dx.doi.org/10.1148/radiol.2017170213] [PMID: 28836886]
[34]
B.L. Nørgaard, J. Leipsic, S. Gaur, S. Seneviratne, B.S. Ko, H. Ito, J.M. Jensen, L. Mauri, B. De Bruyne, H. Bezerra, K. Osawa, M. Marwan, C. Naber, A. Erglis, S.J. Park, E.H. Christiansen, A. Kaltoft, J.F. Lassen, H.E. Bøtker, and S. Achenbach, "Diagnostic performance of noninvasive fractional flow reserve derived from coronary computed tomography angiography in suspected coronary artery disease: the NXT trial (Analysis of Coronary Blood Flow Using CT Angiography: Next Steps)", J. Am. Coll. Cardiol., vol. 63, no. 12, pp. 1145-1155, 2014. [http://dx.doi.org/10.1016/j.jacc.2013.11.043] [PMID: 24486266]
[35]
I. Išgum, M. Prokop, M. Niemeijer, M.A. Viergever, and B. van Ginneken, "Automatic coronary calcium scoring in low-dose chest computed tomography", IEEE Trans. Med. Imaging, vol. 31, no. 12, pp. 2322-2334, 2012. [http://dx.doi.org/10.1109/TMI.2012.2216889] [PMID: 22961297]
[36]
A. Mincholé, J. Camps, A. Lyon, and B. Rodríguez, "Machine learning in the electrocardiogram", J. Electrocardiol., vol. 57, pp. S61-S64, 2019. [http://dx.doi.org/10.1016/j.jelectrocard.2019.08.008] [PMID: 31521378]
[37]
A. Lyon, A. Mincholé, J.P. Martínez, P. Laguna, and B. Rodriguez, "Computational techniques for ECG analysis and interpretation in light of their contribution to medical advances", J. R. Soc. Interface, vol. 15, no. 138, p. 20170821, 2018. [http://dx.doi.org/10.1098/rsif.2017.0821] [PMID: 29321268]
[38]
E. Sourla, S. Sioutas, V. Syrimpeis, A. Tsakalidis, and G. Tzimas, "CardioSmart365: Artificial Intelligence in the Service of Cardiologic Patients", Adv. Artif. Intell., vol. 2012, pp. 1-12, 2012. [http://dx.doi.org/10.1155/2012/585072]
194
Computational Intelligence for Data Analysis, 2023, Vol. 2, 194-209
CHAPTER 12
Smart Cane: Obstacle Recognition for Visually Impaired People Based on Convolutional Neural Network Adnan Hussain1, Bilal Ahmad1 and Muhammad Imad2,* 1 2
Islamia College University Peshawar, Peshawar, Pakistan Abasyn University, Peshawar, Pakistan Abstract: According to the World Health Organization (WHO), there are millions of visually impaired people in the world who face a lot of difficulties in moving independently. 1.3 billion people are living with some visual impairment problem, while 36 million people are completely visually impaired. We proposed a system for visually impaired people to recognize and detect objects based on a convolutional neural network. The proposed method is implemented on Raspberry Pi. The ultrasonic sensors detect obstacles and potholes by using a camera in any direction and generate an audio message for feedback. The experimental results show that the Convolutional Neural Network yielded impressive results of 99.56% accuracy.
Keywords: Convolutional Neural Network, Raspberry Pi, Ultrasonic Sensors, Voice Message. INTRODUCTION According to the World Health Organization, at least 2.2 billion people have vision impairment problems, of which 1 billion people have been prevented or yet to be left addressed. The world faces extensive challenges related to eye care, including treatment, shortage of trained eye care service providers, and poor integration of eye care in the health system [1]. The report was released on world sight day to warn the population of aging, an increasing number of vision impairments, and blindness problems from different eye diseases, such as Presbyopia, affecting 1.8 billion individuals, which can occur at an early age. Myopia affects 2.6 billion, with 312 billion under the age of 19. Other diseases are cataract, which affects 65.2 million, glaucoma affects 6.9 million, corneal opacities affect 4.2 million, diabetic retinopathy affects 3 million, trachoma affects 2 million people, which is the cause of vision impairments [2]. *
Corresponding author Muhammad Imad: Abasyn University, Peshawar, Pakistan; E-mail: [email protected] Mariya Ouaissa, Mariyam Ouaissa, Zakaria Boulouad, Inam Ullah Khan, Sailesh Iyer (Eds.) All rights reserved-© 2023 Bentham Science Publishers
Neural Network
Computational Intelligence for Data Analysis, Vol. 2 195
A visually impaired person faces many difficulties that take a sighted person's aid to find their way. In an unfamiliar environment, blind persons can't find their path themselves. Generally, visually impaired persons use a white cane or walking cane. Electronic Aid technology like Ultrasonic sensors can be used to help visually impaired persons. In the ultrasonic system, energy waves are emitted, reflecting from the obstacles on any side (left, right, front) to help the visually impaired person detect the obstacle within the defined range. The distance between visually impaired persons and the objects is calculated according to the starting and ending pulse from the ultrasonic sensor. We have used a buzzer sensor to inform the blind user about the obstacles. Suppose the obstacles are too close to the blind user. In that case, the proposed system will generate a voice message, and also, a buzzer will be active to inform the visually impaired person about the obstacles. The main objectives of the smart cane system are to develop a model of cane for the blind, which is widely used in mobility. The second objective is to design a proposed model that consumes less power, is weightless, and has precise, accurate object/obstacle detection and recognition performance. This proposed design can provide full support against obstacle avoidance with a voice message. The proposed module will be useful for blind people. It is easier for them to find daily activities without using the standard mobility aid and available for individuals with this disability since mobility aids are less costly, light in size, and can be taken easily anywhere. The Nivedita presented an electronic aid device consisting of a Raspberry pi device, an ultrasonic sensor, a web came microphone, and an LDR sensor. The ultrasonic sensor with a camera has been used to detect obstacles. The LDR sensor has been used to detect the brightness of the environment to check whether there is dark or bright automatically. The proposed system detects the object around the visually impaired person and sends feedback by voice using a microphone [3]. The Mary presents a Raspberry pi-based system for blind people with unique features, such as tracking the object around, sending feedback through voice, and providing information about the environment. The most important feature of the paper is to track the location of the blind person and notify the caretaker to ensure safety [4]. The rest of the paper comprises a literature review in Section 2, details of the proposed methodology (convolutional neural network) are given in Section 3, experimental result analysis, and discussion in Section 4. Finally, the paper is concluded in Section 5.
196 Computational Intelligence for Data Analysis, Vol. 2
Hussain et al.
LITERATURE STUDY The Tapu has been implemented as a smartphone-based obstacle to detect and classify to aid visually impaired people to walk freely and safely in an indoor and outdoor environments. The Lucas Kanade algorithm has been used to extract the feature from images or frames [5]. The camera and the background motion have been estimated through homographic transforms. The obstacles are detected as critical or normal, based on a specific distance, and then the obstacle has been detected and recognized [6]. The model has been proposed, which they used for obstacle detection techniques. Two methods have been used: adaptive colour segmentation and stereo-based colour homography. The pixels are classified into a frame or image in colour segmentation to find the obstacle or free space. The present system has been built on a training algorithm [7]. The proposed design model consists of a 2D laser scanner, foot-mounted pedometer, and three-axis gyroscopes to aid visually impaired people in an indoor environment. They presented 2-layered estimators. In the first estimated layer, the blind cane location has been tracked in the last layer. The second estimated layer finds the person's location keeping a blind cane is tracked [8]. Rodríguez has been presented as a method for obstacle prevention devices to help visually impaired persons. They have implemented an incremental map of the environment with the help of optical SLAM techniques to provide spatial direction and location of the visually impaired people at the same. The proposed design also provided audio feedback to inform the blind person about obstacles [9]. The electronic aid device consisted of three ultrasonic sensors and a microcontroller to detect the object range. The audio and vibration system also has been used to warn visually impaired people to avoid obstacles [10]. Pradeep proposed a device that involves an RGB-D camera for environment perception. The proposed system imposes three things; self-localization, obstacle detection, and object recognition. In self-localization, depth has been perceived based on the tracking technology of colour information. Obstacle detection and recognition provide meaningful information to visually impaired people and help recognize the obstacles such as stairs, walls, vehicles, and doors [11]. Leung has proposed a head-mounted, stereo-vision-based direction system for visually aiding impaired people. They have used visible odometry and feature-
Neural Network
Computational Intelligence for Data Analysis, Vol. 2 197
based SLAM in the proposed design system to make a 3D map for obstacle detection [12]. Shahdib designed a model that involves the head of mounted stereo vision to search ground planes and break up six degrees of freedom (6DOF) into ground plane motion and planer motion which assist visually impaired people. Due to the investigation of the variation array, they evaluate the ground plane and normal to the ground plane using optical/visible data or with the IMU reading [13]. Maidenbaum implements a system based on ultrasonic sensors and a camera for obstacle detection and recognition. While the camera has been used to recognize and measure the size of obstacles [14]. Patil has been designing a system based on two infrared emitter (IR) sensors to emit and receive beams. The sensor has been used to detect waist-level obstacles within a five-meter range and ground-level obstacles at 45 degrees to provide direction to visually impaired people [15]. MATERIALS AND METHODS Dataset Description The dataset consists of different obstacles, such as Vehicles (1,510 images), Doors (1,349 images), Pillars (1,559 images) and Stairs (1,300 images), through various operations and variations to make it in perfect shape. The dataset has been created from the short videos through a Raspberry pi camera, and each frame is extracted from the videos with a size of 640,480 in width and height. The primary rule of deep learning is to divide the data into two parts, training and testing phases, while 70% of the dataset is used for training and 30% for testing purposes (Fig. 1). Methods This paper aims to detect and recognize the obstacles (Vehicles, Doors, Pillars, and Stairs) for visually impaired people based on the deep convolutional network. The number of steps involved in our proposed system explained in this section is presented in Fig. (2).
198 Computational Intelligence for Data Analysis, Vol. 2
Fig. (1). Example of dataset.
Fig. (2). Proposed system framework.
Hussain et al.
Neural Network
Computational Intelligence for Data Analysis, Vol. 2 199
Ultrasonic Sensors As shown in Fig. (3), the four ultrasonic sensors are connected to Raspberry Pi. The three ultrasonic sensors are used to detect obstacles from three directions (front, left, and right), while one sensor is used to detect potholes and obstacles.
Fig. (3). Ultrasonic sensor.
Visual Sensor The visual sensor has a resolution of 5 megapixels. The sensors are connected to Raspberry Pi to recognize the obstacles from the front side (Fig. 4).
Fig. (4). Visual sensor.
200 Computational Intelligence for Data Analysis, Vol. 2
Hussain et al.
Buzzer Sensor The buzzer sensor alerts the blind person about the obstacles by generating a voice through vibration (Fig. 5).
Fig. (5). Buzzer sensor.
Jumper Wires The female-to-male jumper wires perfectly connect the GPIO pins on the Raspberry Pi directly to the breadboard. The reusable, soft, reliable, and strong jumper wires are easy to detect and so easy to connect and disconnect with the breadboard. These jumper wires are suitable and appropriate for prototyping. Each jumper wire Package involves 40-wires, and each wire is 300 mm long (Fig. 6).
Fig. (6). Jumper wire.
Breadboard A breadboard is one of the most important parts of building circuits. A breadboard provides an extension of a single wire. There are a number of holes in the plastic box, arranged in a particular style. A breadboard design contains two types of a section called strips. Two sections are the bus strip and socket strip (Fig. 7).
Neural Network
Computational Intelligence for Data Analysis, Vol. 2 201
Fig. (7). Breadboard.
Bus Strip Bus strips are commonly used to supply power to the circuit. It is composed of two columns, one for ground and the next for power voltage. Socket Strip Socket strips are usually used to handle most of the components in a circuit. Typically, socket strips include two segments, and each section consists of 64 columns and five rows. Commonly each column is connected electrically from the inside. Power Bank Raspberry Pi has always required a specific amount of power as compared to a fully-fledged desktop PC1. Therefore, we use a battery to supply power to the device. A power bank can also be used to supply power to the Raspberry Pi3 (Fig. 8).
Fig. (8). Power bank.
Earphone/Speaker The earphone/Speaker is usually used for audio feedback. In this proposed design, earphones warn visually impaired people about obstacles and tell the direction and
202 Computational Intelligence for Data Analysis, Vol. 2
Hussain et al.
distance from the obstacles. It is better than a buzzer since it provides more accurate results and perception, thereby helping the person react more easily (Fig. 9).
Fig. (9). Earphones.
Traditional Cane Generally, the traditional walking cane provides physical support and perceptionaiding feedback to the user. This cane also helps prevent the user from falling and from a collision with obstacles. If any obstacles are detected, the cane should contact the objects/obstacles before the visually impaired people can decide to alter their direction (Fig. 10).
Fig. (10). Traditional cane.
Smart/Modern Cane The smart walking cane is an updated form of a long cane; it consists of four ultrasonic sensors and one Raspberry Pi camera. The ultrasonic sensor is used to detect the obstacles and also to measure the distance between the smart cane users and the obstacles. The visual sensor (camera) is used to recognize the obstacles. The proposed design is also equipped with a buzzer sensor and audio feedback system [16].
Neural Network
Computational Intelligence for Data Analysis, Vol. 2 203
Proposed Device Architecture The device architecture consists of Raspberry Pi, Ultrasonic Sensors, Visual Sensor, Breadboard, Jumper Wires, Buzzer, Earphone and a power bank. A breadboard is one of the most important parts of establishing a circuit. Breadboard works as a bridge between sensors and Raspberry Pi. The jumper wires are used to connect the sensor indirectly to Raspberry Pi. A power bank device has mounted on the top of the Raspberry Pi to provide a specific amount of power compared to a fully-fledged desktop PC. Raspberry Pi consists of forty GPIO pins, a Memory Card, Camera Interface (CSI), and an HDMI port. The sensors are connected to the Raspberry Pi with the help of jumper wires. The Raspberry Pi and power bank are mounted on the wooden cane to supply power to the Raspberry Pi. The four ultrasonic sensors are connected to Raspberry Pi, which requires 5 voltages of power. Three out of four ultrasonic sensors are used to detect obstacles from three directions (front, left, and right), while the fourth one is used to detect potholes. The visual sensor has a resolution of 5 megapixels and is directly connected to the Raspberry Pi via a camera port to recognize the obstacles from the front side. The buzzer sensor is connected to the Raspberry Pi to alert the blind person about the obstacles, which require three voltages of power. The Earphone is used for audio feedback and transmits a voice message to warn the visually impaired person about the presence of obstacles and provides direction and distance calculation from the obstacles (Fig. 11) [17].
Fig. (11). Smart cane for the proposed system.
204 Computational Intelligence for Data Analysis, Vol. 2
Hussain et al.
Deep Convolutional Neural Network Deep learning has many uses in many areas for classification, segmentation and detection of model data. Deep learning is a sub-branch of the machine learning field, inspired by the structure of the brain [18]. The deep Convolutional Neural Network (CNN) based AlexNet architecture is used to detect and recognize four obstacles: Vehicles, doors, Pillars, and Stairs for visually impaired persons in real-time. The pre-trained model Alex-Net is used to extract the deep feature to classify complex images that cannot be classified with simple handcrafted features. The AlexNet architecture consists of eight layers, five are the convolutional layers and three are fully connected layers. The architecture consists of five convolutional layers and three fully connected layers. The first layer of the Alex-Net model defines the dimensions of the input image as 227x227x3. The intermediate five layers are the sequence convolution layer, and 3 are the dense layers. The Alex-Net model has been trained on the dataset consisting of four different classes (Fig. 12) [19]. 4096
227x227x256
input
55x55x96
conv1 96@11 x11 Stride = 4 Pad = 0
27x27x96
pool1 3x3 Stride = 2
27x27x256
conv2 256@5 x5 Stride = 1 Pad = 2
13x13x256
pool2 3x3 Stride = 2
13x13x384
conv3 384@3 x3 Stride = 1 Pad = 1
13x13x384
conv4 384@3 x3 Stride = 1 Pad = 1
13x13x256
conv5 256@3 x3 Stride = 1 Pad = 1
4096 1000
6x6x256
pool3 3x3 Stride = 2
fc8
fc6
fc7
Fig. (12). Alex-net architecture.
EXPERIMENTAL RESULTS ANALYSIS The study represents the recognition of 4 different obstacles (Vehicles, Doors, Pillars and Stairs) for visually impaired people. Four different performance metrics were used to evaluate the proposed method: Accuracy, Recall, Precision and F1-score, which is given in Eqs. (1-4). Accuracy = (TN + TP) / (TN + TP + FN + FP)
(1)
Recall = TP / (TP + FN)
(2)
Precision = TP / (TP + FP)
(3)
Neural Network
Computational Intelligence for Data Analysis, Vol. 2 205
(4)
F1-Score = 2 * ((Precision * Recall)/ (Precision + Recall))
TP, FP, TN and FN gave in Eqs. (1-4) represent the value of true positive, false positive, true negative, false negative, respectively. The accuracy, precision, recall and F1-score for the proposed system were calculated based on the different measurement factors. Table 1 shows that the proposed method achieved the best performance with an accuracy of 99.63%, a precision of 99.59%, a recall of 99.31%, f1-score is 99.56%. Table 1. Performance results obtained from pre-trained CNN models (Alex-Net). Categories
Accuracy
Precision
Recall
F1-Score
Doors
99.61%
99.04%
100%
99.51%
Pillars
99.70%
99.35%
100%
99.67%
Stairs
99.23%
100%
98.23%
99.10%
Vehicles
100%
100%
100%
100%
Total
99.63%
99.59%
99.31%
99.56%
The proposed model is trained on a pre-trained Caffe Alex-Net model with the best validation accuracy. The validation accuracy of the model during training is 99.5139% and has a loss validation of 0.0201%, as shown in the below graph (Fig. 13). 100
30
1.4
90
accuracy (val) 1.2
99.5139
80
0.0201327
loss (val)
70
Loss
60 0.8
50 40
0.6
30
0.4
20 0.2 0
10
0
5
0
5
10
10
loss (train)
Fig. (13). Loss and cross-validation accuracy.
15 Epoch
15
accuracy (val)
20
25
30
20
25
30
loss (val)
0
Accuracy (%)
1
206 Computational Intelligence for Data Analysis, Vol. 2
Hussain et al.
The confusion matrix contains the actual positive and false-positive values for each category presented in test data for the classification task. It identifies how many test samples are correctly classified or misclassified. The confusion matrix involves the precision and recall values of each class for classification and the accuracy of the proposed technique (Fig. 14) [20].
0 0.0%
0 0.0%
0 0.0%
100% 0.0%
pillars
8 0.5%
462 26.7%
0 0.0%
0 0.0%
98.3% 1.7%
stairs
10 0.6%
6 0.3%
389 22.5%
0 0.0%
96.0% 4.0%
vehicles
0 0.0%
0 0.0%
0 0.0%
453 26.2%
100% 0.0%
95.7% 4.3%
98.7% 1.3%
100% 0.0%
100% 0.0%
98.6% 1.4%
ve
hi
cl
rs st ai
do or
es
403 23.3%
pi lla rs
doors
s
Output Class
Confusion Matrix
Target Class Fig. (14). Confusion matrix.
The experimental results of the proposed techniques for visually impaired people can be illustrated in Table 1. The four obstacles (Vehicles, Doors, Pillars and Stairs) are recognized from the pre-trained Alex-Net architecture. The proposed technique provides an impressive result, as presented in Fig. (15).
Neural Network
Computational Intelligence for Data Analysis, Vol. 2 207
Fig. (15). Examples of obstacle recognition (vehicles, doors, pillars, stairs).
CONCLUSION This study uses four ultrasonic sensors, a visual camera, and a buzzer for the smart cane system. The proposed system will help the blind in navigation within indoor and outdoor environments without the aid of the guidance of a sighted person. Three sensors are used to detect objects in three directions (front, left, right), and one is used to detect potholes. The visual sensor is used for the recognition of obstacles in the way of smart cane users. Overall, the existing design system provides advantages instead of traditional cane. The current proposed model is also to be placed on the self-made wooden cane. Improvements could be made to ensure the system is more efficient and effective than the currently proposed design. In the future, we aim to install GPS, which helps the visually impaired person in an outdoor location that helps their relatives to find them easily and provide a guideline.
208 Computational Intelligence for Data Analysis, Vol. 2
Hussain et al.
CONSENT FOR PUBLICATON Declared none. CONFLICT OF INTEREST The author declares no conflict of interest, financial or otherwise. ACKNOWLEDGEMENT Declared none. REFERENCES [1]
"World report on vision", Who.int, 2020. [Online]. https://www.who. int/publications-detail/worl-report-on-vision
[2]
M. Imad, S.I. Ullah, A. Salam, W.U. Khan, F. Ullah, and M.A. Hassan, "Automatic Detection of Bullet in Human Body Based on X-ray Images Using Machine Learning", Int. J. Comput. Sci. Inf. Secur., 2020.
[3]
M.M. D’souza, A. Agrawal, and V. Tina, "V. HR, and T. Navya, “Autonomous Walking with Guiding Stick for The Blind Using Echolocation and Image Processing", Methodology (Gött.), vol. 7, no. 05, 2019.
[4]
A. Nivedita, M. Sindhuja, G. Asha, R. Subasree, and S. Monisha, "Smart cane navigation for visually impaired", In: International Journal of Innovative Technology and Exploring Engineering (IJITEE). pp. 2278-3075, 2019.
[5]
R. Tapu, B. Mocanu, A. Bursuc, and T. Zaharia, "A smartphone-based obstacle detection and classification system for assisting visually impaired people", 2013 IEEE International Conference on Computer Vision Workshops, pp. 444-451. [http://dx.doi.org/10.1109/ICCVW.2013.65]
[6]
P. H. Batavia and S. Singh, "Obstacle detection using adaptive colour segmentation and color stereo homography," In Proceedings 2001 ICRA. IEEE International Conference on Robotics and Automation (Cat. No. 01CH37164), vol. 1: pp. 705-710, 2001.
[7]
B. D. Lucas and T. Kanade, "An iterative image registration technique with an application to stereo vision," 1981.
[8]
J.A. Hesch, and S.I. Roumeliotis, "Design and analysis of a portable indoor localization aid for the visually impaired", Int. J. Robot. Res., vol. 29, no. 11, pp. 1400-1415, 2010. [http://dx.doi.org/10.1177/0278364910373160]
[9]
A. Rodríguez, L. M. Bergasa, P. F. Alcantarilla, J. Yebes, and A. Cela, "Obstacle avoidance system for assisting visually impaired people," in Proceedings of the IEEE Intelligent Vehicles Symposium Workshops, Madrid, Spain, vol. 35, p. 16, 2012.
[10]
M. Vlaminck, Q.L. Hiep, H. Vu, P. Veelaert, and W. Philips, "Indoor assistance for visually impaired people using an RGB-D camera", 2016 IEEE Southwest Symposium on Image Analysis and Interpretation (SSIAI). pp. 161-164, 2016. [http://dx.doi.org/10.1109/SSIAI.2016.7459200]
[11]
V. Pradeep, G. Medioni, and J. Weiland, "Robot vision for the visually impaired", 2010 IEEE Computer Society Conference on Computer Vision and Pattern Recognition-Workshops, . pp. 15-22, 2010.
[12]
T-S. Leung, and G. Medioni, "Visual navigation aid for the blind in dynamic environments",
Neural Network
Computational Intelligence for Data Analysis, Vol. 2 209
Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops. pp. 565-572, 2014. [http://dx.doi.org/10.1109/CVPRW.2014.89] [13]
F. Shahdib, M.W.U. Bhuiyan, M.K. Hasan, and H. Mahmud, "Obstacle detection and object size measurement for autonomous mobile robot using a sensor", Int. J. Comput. Appl., vol. 66, no. 9, 2013.
[14]
S. Maidenbaum, S. Hanassy, S. Abboud, G. Buchs, D.R. Chebat, S. Levy-Tzedek, and A. Amedi, "The “EyeCane”, a new electronic travel aid for the blind: Technology, behavior & swift learning", Restor. Neurol. Neurosci., vol. 32, no. 6, pp. 813-824, 2014. [http://dx.doi.org/10.3233/RNN-130351] [PMID: 25201814]
[15]
M. Imad, A. Hussain, M. Hassan, Z. Butt and N. Sahar, "IoT Based Machine Learning and Deep Learning Platform for COVID-19 Prevention and Control: A Systematic Review", AI and IoT for Sustainable Development in Emerging Countries, pp. 523-536, 2022. [http://dx.doi.org/10.1007/978-3-030-90618-4_26]
[16]
A. Hussain, M. Imad, A. Khan and B. Ullah, "Multi-class Classification for the Identification of COVID-19 in X-ray Images Using Customized Efficient Neural Network", AI and IoT for Sustainable Development in Emerging Countries, pp. 473-486, 2022. [http://dx.doi.org/10.1007/978-3-030-90618-4_23]
[17]
Faiza, S. I. ullah, A. Salam, F. Ullah, M. Imad and M.A. Hassan, "Diagnosing of Dermoscopic Images using Machine Learning approaches for Melanoma Detection", 2020 IEEE 23rd International Multitopic Conference (INMIC), 2020. [http://dx.doi.org/10.1109/INMIC50486.2020.9318114]
[18]
S. Nazish, A. Ullah, W.U. Salam, and M. Imad, "COVID-19 Lung Image Classification Based on Logistic Regression and Support Vector Machine", Artificial Intelligence Systems and the Internet of Things in the Digital Era, pp. 13-23, 2021. [http://dx.doi.org/10.1007/978-3-030-77246-8_2]
[19]
M. imad, N. Khan, F. Ullah, M. Abul Hassan, A. Hussain, and Faiza, “COVID-19 Classification based on Chest X-ray Images Using Machine Learning Techniques ”, JCSTS, vol. 2, no. 2, pp. 01–11, Oct. 2020.
[20]
"M. Imad, F. Ullah, M. Abul Hassan, and Naimullah, “Pakistani Currency Recognition to Assist Blind Person Based on Convolutional Neural Network”", JCSTS, vol. 2, no. 2, pp. 12-19, 2020.
210
Computational Intelligence for Data Analysis, 2023, Vol. 2, 210-228
CHAPTER 13
A Survey on Brain-Computer Interface and Related Applications Krishna Pai1,*, Rakhee Kallimani1, Sridhar Iyer1, B. Uma Maheswari2, Rajashri Khanai1 and Dattaprasad Torse2 Department of Electronics and Communication Engineering, KLE Dr. M.S. Sheshgiri College of Engineering and Technology, Udyambag, Belagavi, KA, India- 590008 2 Department of Computer Science and Engineering, Amrita School of Engineering, Bengaluru, Amrita Vishwa Vidyapeetham, KA, India- 560035 1
Abstract: Brain Computer Interface (BCI) systems are able to communicate directly between the brain and computer using neural activity measurements without the involvement of muscle movements. For BCI systems to be widely used by people with severe disabilities, long-term studies of their real-world use are needed, along with effective and feasible dissemination models. In addition, the robustness of the BCI systems' performance should be improved, so they reach the same level of robustness as natural muscle-based health monitoring. In this chapter, we review the recent BCIrelated studies, followed by the most relevant applications. We also present the key issues and challenges which exist in regard to the BCI systems and also provide future directions.
Keywords: Artificial Intelligence, BCI, EEG, Machine Learning, MEG. INTRODUCTION Brain Computer Interface (BCI) uses the brain's power to compute to make use of relatively new technology. According to Berger [1], research has been trying to decode the brain's signals since the first discovery of electroencephalography (EEG) a century ago. Until recently, the development of BCIs was thought of as science fiction. BCI system collaborates the brain with an external device that uses signals from the brain for performing external activities, such as moving a wheelchair, robotic arm, or a computer cursor. * Corresponding author Krishna Pai: Department of Electronics and Communication Engineering, KLE Dr. M.S. Sheshgiri College of Engineering and Technology, Udyambag, Belagavi, KA, India- 590008; E-mail: [email protected]
Mariya Ouaissa, Mariyam Ouaissa, Zakaria Boulouad, Inam Ullah Khan, Sailesh Iyer (Eds.) All rights reserved-© 2023 Bentham Science Publishers
Brain-Computer Interface
Computational Intelligence for Data Analysis, Vol. 2 211
There are four main components of a BCI model, namely, the sensing device, the amplifier, the filter, and the control system. According to Herwing et al. [2] and Jurcak et al. [3], the sensing device comprises a cap consisting of the electrodes which are placed according to the international 10–20 standards. Furthermore, according to Zhang et al. [4], an amplifier could be one of several biological amplifiers available on the market, and the research on BCI is geared toward developing a filter and control system that can be applied to brain signals. When a person thinks of performing any task, such as moving a cursor, then in such a case, signals will be generated in the brain, which is transferred from the brain to the finger on the computer’s mouse via the body’s neuromuscular system. As the follow-up step, the finger will move the cursor. In contrast, in BCI, such signals are transferred to an external device where they will be decoded for moving the cursor. As another example, research on BCI also aims to help such people who suffer from issues related to damaged hearing and sight and damaged movement. An estimated 1.5 billion people suffer from neurological and neuroanatomical diseases and injuries worldwide, resulting in movement impairments, which make it difficult to communicate, reach, and grasp independently. A cortical prosthetic system consists of an end effector, which receives a command for a particular action via a BCI that records the cortical activity of individuals who have suffered neurological injuries such as spinal cord injuries, amyotrophic lateral sclerosis, and strokes. In addition, a BCI decodes information pertaining to the intended function. There is a wide range of end effectors in use now, ranging from virtual typing communication systems to robotic arms and hands, as well as functional electrical stimulation for the reanimation of limbs. BCI can be invasive in varying degrees, have varying spatial and temporal resolutions, and record a wide range of signals. In BCI applications, such as the low-throughput communication spelling systems, EEG, MEG, and fMRI can be used as non-invasive brain imaging technologies, according to Speier et al. [5]. Daly et al. [6] enumerated several problems associated with these noninvasive BCI approaches, such as the fact that they are often slow (e.g., fMRI), have a low spatial resolution and are susceptible to being corrupted by external artefacts. Thus, such options are not suitable for complex real-time applications like highperformance communications, tracking multidimensional robotic limbs, or reanimation of paralyzed limbs with coordinated grasps and reaches. An invasive BCI, on the other hand, is able to command higher dimensional systems naturally and restore more complex functions, as a result of its higher resolution and wider transmission bandwidth.
212 Computational Intelligence for Data Analysis, Vol. 2
Pai et al.
According to Rosenow et al. [7], brain implants are among the most promising and popular technologies for assisting patients with motor paralysis (such as paraplegia or quadriplegia) caused by strokes, spinal cord injuries, cerebral palsy, and amyotrophic lateral sclerosis (ALS). Similarly, eye tracking can be used to control external devices by paralyzed people, but this tech has numerous drawbacks, as it relies on cameras or electrodes on the face to record eye movements or electrical signals, such as electrooculography (EOG). As a result of BCI, neural commands are delivered to external devices by translating human brain activity into external actions, according to McIntyre et al. [8], Chen et al. [9], and Donoghue [10]. While BCI is most often used to help disabled individuals with motor system disorders, it is also very helpful to those with healthy motor systems as well as the elderly. The development of intelligent, adaptive, and rehabilitative BCI applications for adults and geriatric patients will enhance their relationships with their families, improve their cognitive and motor skills, and help with household tasks. BCIs are generally regarded as mindreading technologies, but this does not hold true in most cases. As opposed to mind readers, BCIs provide the user with control by using brain signals rather than muscle movements, so they don't extract information from unknowing or unwilling subjects. A BCI and a user are thus working together through training sessions which involve the user generating brain signals that inform the BCI of the intended action, and the BCI converting the signals into instructions that the output device is supposed to carry out. As per the aforementioned, the research community faces numerous challenges in the implementation of BCI devices. Specifically, it is required that the electrodes and the surgical methods used in the BCI process are minimally invasive, which has resulted in much research focus on the non-invasive methods of braincomputer interfacing. In this chapter, we survey the recent research on BCI and its related applications. The related works are detailed in Section 2. Section 3 discusses the most relevant applications of BCI. In Section 4, we highlight the main challenges in regard to BCI and propose the relevant directions. Finally, Section 5 concludes the chapter. RELATED WORKS Any human being usually produces a wide range of signals at any point in time, from the eyes, ears, nose, and other sensory organs in the body. These signals travel to the brain via the nervous system. The cerebral lobes play an important role for humans or animals when understanding perception, thoughts, language, and memory, and thus EEG sensors, NIRS detectors, etc., are used to acquire neural signals. With the help of these signals, the brain activity of a human or an
Brain-Computer Interface
Computational Intelligence for Data Analysis, Vol. 2 213
animal is understood via brain activity measurement algorithms. Fig. (1) demonstrates the detailed BCI interface from which it can be seen that during the process of signal acquisition, pre-processing of the signal is carried out, which includes filtering, sampling and artefact removal, as described by Cao [11]. Using these pre-processed data, the feature extraction process is carried out, following which classification algorithms or Convolutional Neural Network (CNN) classifiers can be used to understand the neural controls which are required for various applications such as medical gaming education, etc. Further, these neural controls are provided as sensory feedback back to the brain in view of understanding whether the activity which was carried out by the neural control is as desired or not. In this regard, the survey in this chapter is mainly focused on the various types of classification algorithms / CNN classifiers which can/have been implemented for the processing of neural signals, and, therefore aid in obtaining the neural controls.
Fig. (1). Block diagram of BCI interface.
The EEG data which is obtained from the brain in the form of neural signals, have multiple channels. Singh et al. [12] have used single-channel EEG data for developing a prototype using the Internet of things (IoT) and BCI technology. The prototype is developed using MATLAB software, where the classifiers are trained using the Weighted K-Nearest Neighbor Algorithm (Wk-NN). An Arduino microcontroller is used as the hardware platform. The authors have demonstrated that, towards the end of prototyping, a low-cost and highly accurate system can be guaranteed which can control the environment. The authors are also able to fetch and provide data to the cloud through If This Then That (IFTTT). The classifiers used in this article are the cognitive state classifier and event-related potential ERP classifier using the Wk-NN algorithm. An accuracy of 95% with 3100 observations per second and an accuracy of 98.3% with 1800 observations per second is achieved by opting for the cognitive state classifier and ERP classifier, respectively.
214 Computational Intelligence for Data Analysis, Vol. 2
Pai et al.
In general, the classifier algorithms depend on the features which are abstracted from the data that has been collected from the neural signals. There are multiple feature extraction techniques available in the literature, and one such feature extraction technique is the flexible analytic wallet transform (FAWT) method. In [13], Chaudhary et al. used the FAWT technique to divide the EEG signals into their sub-bands thereby extracting the movement-based features. Subspace KNN is one of the best classification methods which is known to have reached an accuracy level of 99.33%. The authors have also used other classification methods, such as Support vector machines (SVM), decision trees, Linear Discriminant Analysis (LDA) and standard KNN, which have obtained a large spectrum of results. Specifically, an accuracy of 95.72%, 92.8%, 91.79% and 81.1% is obtained by SVM, standard KNN, decision tree and LDA, respectively. The authors have also considered accuracy, specificity, kappa value, f1-score, and sensitivity as the performance parameters. Jin et al. [14] have shown that extreme learning machine (ELM) is more efficient than SVM, considering the benchmark performance. They also state that the performance of ELM is highly proportional to the number of hidden nodes, which are also known as the network structure of ELM. Sparse Bayesian ELM-based algorithm (SBELM) is shown to exhibit high characterization accuracy on EEG signals. During the comparative study, it was observed that accuracy of 76.3%, 77.1%, 77.8% and 78.5% is obtained via SVM, ELM, BELM and SBELM, respectively. Lastly, the authors have stated that the proposed model can be further improved by adding more distinctive and high-level attributes. An accuracy level of 95% is obtained using the linear regression method on the equation-based electroencephalogram model by Yi et al. [15]. The basic machine learning model can be further improved by applying various methods such as systolic matrix multiplication, inversion, and vector multiplication. The proposed algorithm consists of a normal equation that undergoes many mathematical computations to obtain the output. The development environment in the study is built based on docker technology. According to Zheng et al. [16], when deep learning and ELM are combined, in addition to an improvement in a Long Short Term Memory network (LSTMs) and bagging algorithm, a classification model is formed, which is known as LSTMS-B. This new classification model consists of Swish activation. In the study, an intelligent visual classification is observed with an accuracy of 97.13% with 40 classes. Additional image categories are required for a sophisticated technique to distinguish the EEG signals. Chen et al. [17] have used the temporal-spatial CNN model, which consists of two separate sets of layers, namely, the classification layer and the feature extraction layer. It is observed that a separate set of data is required for training the feature
Brain-Computer Interface
Computational Intelligence for Data Analysis, Vol. 2 215
extraction layers and the classification layers separately. The model provides an overall classification accuracy of 65.7% when spatial filtering is used for feature extraction. The performance observed in this study is found to be comparatively better compared to existing classic models. Deep learning models for the BCI interface demand a huge amount of data for highly accurate classification results. However, in a general scenario, there is huge data scarcity, so the developed model cannot operate at its full potential. To solve this issue, Zhang et al. [18] have proposed a method to generate artificial brain signals which can act as supplementary data along with the actual data. The authors also propose the Deep Convolutional Generative Adversarial Networks (cDCGAN) method for data augmentation, which is evaluated on the Convolutional neural network (CNN) model for corresponding classification performance. The CNN classification model is shown to acquire accuracy of 82.86%, 82.86% and 82.14% by employing the actual EEG data, artificially generated EEG data and mixed EEG data, respectively. The authors [19] have proved the superiority of deep learning-based classification techniques over the existing traditional classification techniques. Five class steady-state visual evoked potential datasets (SSVEP) are used in this study, and a detailed comparison is provided between the CNN and the recurrent neural networks (RNN) using the LSTM architecture. The authors have proved that CNN demonstrates the highest accuracy of classification, corresponding to 69.3%, whereas the traditional classification algorithm, i.e., SVM with Gaussian kernel, achieves a classification accuracy of 66.9%. Rasheed [20] provides a comprehensive review of the role of ML in BCI; the authors have provided a detailed ML method focusing on mental state detection, state categorization and emotion classification. EEG signal classification, eventrelated potential signal classification, motor imagery categorization and limb movement classification. The methods such as Common Spatial Pattern (CSP), Principal Component Analysis (PCA), Independent Component Analysis (ICA), Autoregressive (AR) Method, Wavelet Packet Decomposition (WPD) for feature extraction and selection are explored. Detailed Classification methods such as KNearest Neighbor (KNN), Linear Discriminant Analysis (LDA), Naive Bayes, Extreme Learning Machine (ELM), Support Vector Machine (SVM), and Neural Networks (NNs) are reviewed. The Neural Networks are studied in detail for Multilayer Perceptron (MLP), Artificial Neural Networks (ANN), and Convolutional Neural Networks (CNN). Al-Nafjan et al. [21] performed a comprehensive review of development in emotion detection and classification. The authors have provided a deep insight in reviewing the published articles focusing significantly on the recognition of
216 Computational Intelligence for Data Analysis, Vol. 2
Pai et al.
emotional state based on a specific study performed on the participants/patients at different emotional states using EEG BCI. Also, the researchers provide a section on trends and challenges in the mentioned field and emphasize the growth of the technology in the next decade. Birbaumer et al. [22] provide a complete review of the clinical application of BCI targeting paralyzed patients. The study is focused on locked-in syndrome as well as on Completely locked-in syndrome. The work also demonstrates the application of the Brain-machine interface on chronic stroke patients. The results have shown a positive response to the combination of BCI with psychiatric and clinical psychological issues. The authors also mention the research scope in view of improving the complex behavioural disorders. Min et al. [23] mention different types of neuroimaging BCI methods and provide the merits and demerits of each method. The authors in the article mention the trend of development of the brain-to-brain interface, where individuals are linked with the computer as a mediator. EEG being the popular neuroimaging method, the authors describe the other neuroimaging methods, such as MEG, which is associated with neuro activity based on magnetic fields. fMRI detects the changes in local cerebral blood oxygen level-dependent signal contrast. NIRS is a method employing a near-infrared spectrum and can penetrate the skull and investigate cerebral metabolism. It is the recent development technique in assessing cortical regional activities. The most recent development is the ultrasound Doppler imaging technique fTCD. The limitation of this method is in terms of penetration. Abbasi et al. [24] implemented EEG data parsed with Discrete Wavelet Transform (DWT), and each multilayer perceptron neural network characteristic is statistically analyzed. The proposed approach presents 98.33% accuracy in comparison to other models. The model is proposed to aid in the detection, diagnosis, and classification of epileptic seizures. Wang et al. [25] aimed to design a BCI system to extract features and classify the EEG signals accurately by employing Deep Learning Methods. The work is demonstrated on Convolutional Neural networks and Long-term Short-term memory networks. The study is focused on the data obtained from 2 women and 3 men with subjects as healthy, mental and right-handed students. Torse et al. [26] presented a detailed survey on automated seizure detection. Thereby laying a foundation to solve the manual detection method to diagnose epilepsy via manual operations. This study was conducted around various classifiers such as KNN, LS-SVM, Multilayer Perceptron Neural Network (MLPNN), Naïve Bayes (NB) and Random Forest (RF). The highest classification accuracy was achieved at 97.3% due to an automated system consisting of varied
Brain-Computer Interface
Computational Intelligence for Data Analysis, Vol. 2 217
values of Q that decomposes EEG signals which in turn computes Korsakov and Shannon entropies. Hence, detecting the seizures using a classifier named random forest. The other classification accuracies achieved were 86.1%, 95.6%, 92.5% and 84.4% for KNN, LS-SVM, Multilayer Perceptron Neural Network (MLPNN) and Naïve Bayes (NB), respectively. A highly accurate model for the categorisation of normal or seizure conditions for chronic brain disorders is explored by Torse et al. [27]. The models are based on tuneable-Q wavelet transform (TQWT) and ensemble empirical mode decomposition (EEMD) algorithms. The entropy and TQWT parameters generated an optimum value for high classification performance. The joint method consisting of EEMD-TQWT + RF algorithms demonstrates the highest classification accuracy of 96.2%. In Table 1, we summarize the classifier algorithms implemented by various studies, including the accuracy which has been achieved. Table 1. Details of the algorithms used by various studies and the achieved accuracy. References
Classifier Algorithm Used
Accuracy
[12]
Event-Related Potential (ERP) using Weighted k-Nearest Neighbor (Wk-NN) algorithm
98.3%
[12]
Cognitive State Classifier using Weighted k-Nearest Neighbor (Wk-NN) algorithm
95%
[13]
Subspace KNN
99.33%
[13]
Support vector machines (SVM)
95.72%
[13]
Standard KNN
92.8%
[13]
Decision Tree
91.79%
[13]
Linear Discriminant Analysis (LDA)
81.1%
[14]
Support vector machines (SVM)
76.3%
[14]
Extreme learning machine (ELM)
77.1%
[14]
Bayesian ELM (BELM)
77.8%
[14]
Sparse Bayesian ELM (SBELM)
78.5%
[15]
Linear Regression
95%
[16]
Deep learning + ELM + LSTM + bagging algorithm (LSTMS-B)
97.13%
[17]
Temporal-spatial CNN
65.7%
[18]
Convolutional neural network (CNN)
82.86%
[18]
CNN + Deep Convolutional Generative Adversarial Networks (cDCGAN)
82.86%
[18]
CNN + Mixed Data
82.14%
[19]
CNN/recurrent neural networks (RNN) using LSTM architecture
69.3%
218 Computational Intelligence for Data Analysis, Vol. 2
Pai et al.
(Table 1) cont.....
References
Classifier Algorithm Used
Accuracy
[19]
SVM with Gaussian kernel
66.9%.
[26]
KNN
86.1%
[26]
LS-SVM
95.6%
[26]
Multilayer Perceptron Neural Network (MLPNN)
92.5%
[26]
Naïve Bayes (NB)
84.4%
[26]
Random Forest (RF)
97.3%
[27]
tunable-Q wavelet transform (TQWT) + ensemble empirical mode decomposition (EEMD) + RF
96.2%
APPLICATIONS OF BCI In this section, we detail the state-of-the-art applications of BCI. Several studies are underway to enhance current BCI systems by combining multimodal signal acquisition methods. Several studies have demonstrated that fMRI with simultaneous EEG can yield complementary features through the use of the EEG's effective spatial resolution and the EEG's effective temporal resolution, according to Debener et al. [28]. MEG, as mentioned by Kauhanen et al. [29], can also work in conjunction with EEG, since it provides information regarding radially/tangentially polarized sources in cortical-subcortical networks, and can complement the EEG by adding complementary information. Few studies, such as Min et al. [30], and Piastra et al. [31], contend that EEG and MEG can detect subcortical activities; however, skepticism remains regarding their capability to detect brain activities that originate from subcortical areas. It has become increasingly common to combine different signal acquisition methods for improving BCI efficiency in recent years. In order to translate any brain signal into a command that can be used by a computer or other external device, signal processing combined with ML techniques plays a crucial role. Time-domain representations of the brain signals include the Fourier transform (FT) and autoregressive models, whereas timefrequency representations include short-time FTs and wavelet transforms, as described by Bashashati et al. [32]. When spatial filtering is considered, as described by Saha et al. [33], many inverse models enable the differentiation and projection of actual sources on three-dimensional cortical-subcortical networks. A variety of linear and nonlinear classification algorithms, such as kernel-based support vector machines and linear discriminant analysis can be used to translate the extracted features, as described by Lotte et al. [34]. A deeper understanding of BCI based on deep learning paradigms is receiving increasing attention from researchers thanks to the remarkable advancements in computational
Brain-Computer Interface
Computational Intelligence for Data Analysis, Vol. 2 219
infrastructure over the past decade, as mentioned by Nagel et al. [35]. This is because such systems can evaluate large datasets. The inverse problem, as described by Saha et al. [36], must be solved in order to model the cortical sources. Additionally, new methods for localizing sources have emerged in recent years, including wavelet-based maximum entropy over the average that represents EEG/MEG signals as time-frequency contents and then transforms them into spatial representations. Sensors with customized designs are developed for the acquisition of brain signals. Deisseroth et al. [37] have described many forms of neuro sensors, including electrochemical, optical, chemical, and biological. Nanowire Field Effect Transistors and other p/n junction devices, as described by Oxley et al. [38], have demonstrated the feasibility of neuro-sensing techniques in intracellular recordings, even in deep brain regions, thanks to major advancements in nanotechnology. According to Willett et al. [39], Zhang et al. [40], and Herff et al. [41], BCI systems have also shown the ability to read thoughts in regard to multiple movements. Researchers have used BCI to test the thinking of pigs while running on the treadmill. They were also able to predict what the pig's brain would do next in addition to interpreting the pig's thoughts. Moreover, researchers are carrying out research on the use of brain-computer interfaces to improve communication among people who are paralyzed physically and unable to speak or move. In these situations, BCI enables the person to communicate verbally and in writing. BCI system for brain-to-text transcription also aids paralyzed people to imagine writing the letters, and these letters appear on the screen. In the experiments, the paralyzed patients are asked to think at a rate of 90 characters per minute, which are then decoded and presented on the screen in less than one second. This demonstrates results that are 80% faster than the typical typing speed on a smartphone screen for a person of the same age range. Even after being affected for years with paralysis, the motor cortex is still powerful enough to read by a BCI well enough for typing speed and precision. Rabbani et al. [42], and Rezazadeh Sereshkeh et al. [43] have developed BCI to generate synthesized speech. To obtain the results, electrodes were placed on the surface of the brain to measure the signal and calculate movements of the vocal tract. The captured vocal tract movements are then converted to sounds, and the original voice is then converted into synthesized speech. According to Ptito et al. [44], and Farnum et al. [45], BCIs have also been demonstrated to be used for assisting and restoring the vision of blind people. A camera, positioned on glass, captures a video image, processes it, and then activates a chip installed in a retina to stimulate the eye. The ultimate goal is to be able to use the implant to transform camera inputs into brain activity. Thus, it presents a novel method of treating blindness that targets the brain directly rather than the eyes. According to Hwang
220 Computational Intelligence for Data Analysis, Vol. 2
Pai et al.
et al. [46], and Lucia et al. [47], patients with hearing loss have several options for addressing their hearing loss, including BCIs. It works similarly to the microphone, which picks up sounds and conversations from various people. Implants are positioned behind the ear or beneath the skin, which then transfers the impulses to electrodes which are implanted in the cochlea via a sound processor. Changes occur in the brain and the body as age increases, and elderly people require assistance. Nevertheless, they are generally not provided the assistance they require owing to the expense of the required care and treatment. BCIs have also been found to be helpful in assisting elderly people, as described by Belkacem et al. [48]. In this case, BCI technology can help the elderly in multiple ways such as, improving their communication, controlling domestic appliances, and strengthening their cognitive abilities. Currently, dementia is very common among adults, as mentioned by Fukushima et al. [49], and Rutkowski et al. [50]. BCIs are used to detect early Alzheimer's and are also used to classify and to know the type of Dementia. It is very important to detect early Alzheimer’s type of dementia (AD) as it can be treated using medicine itself. Motion sickness and drowsiness decrease the performance of drivers, and on occasion also result in an accident, as described by Ferreira et al. [51], Lin et al. [52], Wei et al. [53], Pritchett et al. [54], and Li et al. [55]. Motion sickness can be predicted by measuring EEG signals received from different areas of the brain region. EEGbased BCI systems have shown immense potential in a variety of applications including post-stroke treatment, as described by Mrachacz-Kersting et al. [56], and Ruiz et al. [57], illness diagnosis (Abdulkader et al. [58]), emotion identification (Mishra et al. [59]), and gaming (Lim et al. [60]), which have shown great promise for EEG-based BCI systems. BCI systems also aid in predicting brain tumours, as described by Song et al. [61], epilepsy (seizure disorder), as mentioned by Hosseini et al. [62], and encephalitis (brain swelling). Further, it is possible to ensure early detection of Dyslexia using BCI, which in turn helps in treating the children and improving their selfconfidence, as described by Fadzal et al. [63]. The combination of rtfMRI-EEG BCI systems is used for finding depression, as mentioned by Minkowski et al. [64], and emotion recognition, as described by Moschona et al. [65] of patients can also be identified using the fMRI BCI system. The BrainArena, as described by Bonnet et al. [66], connects two brains to a football video game using BCIs and can score by thinking left and right movements. Playing Brainball games twice a week has been demonstrated to improve the students’ learning skills. Lastly, people can measure their stress level by playing the Brainball game in which the player with less stress only can move the balls, as described by Mridha et al. [67].
Brain-Computer Interface
Computational Intelligence for Data Analysis, Vol. 2 221
ISSUES AND CHALLENGES, AND FUTURE DIRECTIONS BCI technology has gained tremendous attention from the research community, including medical and non-medical scientists. The major issues faced within the research domain can be categorized, according to Mudgal et al. [68], into three sectors, namely, neuro-psycho-physiological, technical, and ethical. Neuro-Psycho-Physiological Issues The performance of the brain is affected by anatomy problems, including the complexity owing to the genetic issues and structure, diversity of the brain, and psychological problems such as anxiety, fatigue and emotional state, stress, and memory which vary from one person to another. These issues have been demonstrated to predominantly affect the performance of BCI. Technical Issues The major challenge faced by the BCI system is to select the appropriate components or technology related to the application as it is related to the signal. Selection of the method to acquire the brain signal and then to process it presents a major challenge. Another challenge is educating the operator of the BCI system. Ethical Issues These are related to the safety of the physical and mental, and emotional state of the user. The user data is highly confidential and needs to be maintained by the system. User consent is another prominent issue related to the BCI system. In regard to the technical issues, Mudgal et al. [68] and Vasiljevic et al. [69] detail the significant challenges in developing the BCI. Specifically: 1. The Non-Linearity characteristics of the brain signal, with the non-stationarity behaviour of the signal, presents a key challenge. In addition, noise also aids its vital contribution to the challenges of the BCI system. 2. Another technical challenge is the brain signal transfer rate. Currently, the BCIs are at an extremely slow transfer rate, and this is a major research topic, especially for BCI based on visual stimulus. 3. The selection of appropriate decoding techniques, processing and classification algorithms is a challenge to control the BCI system. 4. Another critical issue is the lack of balancing between the training required for
222 Computational Intelligence for Data Analysis, Vol. 2
Pai et al.
the accurate function of the system and the technical complexity of decoding the activities of the brain. 5. There is a need to focus on the BCI system and provide a systematic approach for a particular performance metric as there are varied performance metrics, and it is a major challenge to select a particular metric for a specific application. In addition, the areas which need to be further explored include the long-term effects which are not known, technology effects on the life quality of the multiple subjects and their relatives/families, the side effects which are related to health, such as quality of sleep, functioning of the normal brain and memory, and the non-convertible alterations which are made to the brain. Further, there also occur multiple legal and social issues which need to be settled, namely, the accountability and responsibility which is required to be taken in regard to the influence of the BCIs, inaccuracy in the translation of the cognitive intentions, the possible changes in the personality, no being able to distinguish between humans and the machine-controlled actions, misuse of techniques during the interrogation by authorities, the capability and privacy of the mind-reading, and the mind control and emotion control related issues. In addition, a major area of concern is the legal responsibility which needs to be finalized for scenarios where accidents occur. The additional challenge that has emerged is in regard to the response of the body to the invasive BCIs, which requires the use of implanted micro-electrodes array which come under direct contact with specific neurons within the brain. These electrodes are recognised as foreign bodies which trigger the natural immune system, and these neurons are surrounded by fibrous capsules of the tissue in turn minimizing the signal recording ability of the electrodes, ultimately resulting in the BCIs’ unusability. Further, minimizing the power consumption for decreasing the battery size and prolonging the lifespan is another key challenge since there exists a trade-off between power consumption and efficient bio-security. In this regard, for ensuring the bio-security, signals are required to be encrypted, which increases the power consumption. Further, the major problem in the implementation of the BCI technology is a lack of efficient sensor modality which provides safe, accurate, and robust access to the brain signals. Also, the development of such sensors, with additional channels for improving the accuracy and reducing the corresponding power usage is a major challenge. The ethical, legal, and social implications of the BCIs may also slow down, stop, or divert this technology into a completely different path compared to the initial aim. Lastly, over the last decade, technology related to genetic sequencing technology and modern tools to map have aided the increase
Brain-Computer Interface
Computational Intelligence for Data Analysis, Vol. 2 223
in understanding the neuronal firing patterns and the manner in which they lead to different actions. The brain-interfacing devices are now becoming more sensitive, smaller, smarter, and portable over time, and future technologies must address the issues which are related to ease of use, performance robustness, and cost reduction. CONCLUSION The BCI community is conducting much research to provide standardized platforms and to assist the complex and non-linear dynamics that BCI systems encounter. In this chapter, we have reviewed the most recent studies related to BCI systems. We have also presented and detailed the state-of-the-art BCI systems. Lastly, we have listed the key issues and challenges in regard to the BCI systems followed by some future directions that can enhance the research to be conducted on the BCI systems. CONSENT FOR PUBLICATON Declared none. CONFLICT OF INTEREST The author declares no conflict of interest, financial or otherwise. ACKNOWLEDGEMENT Declared none. REFERENCES [1]
H. Berger, "Über das Elektrenkephalogramm des Menschen", Arch. Psychiatr. Nervenkr., vol. 108, no. 3, pp. 407-431, 1938. [http://dx.doi.org/10.1007/BF01824101]
[2]
U. Herwig, P. Satrapi, and C. Schönfeldt-Lecuona, "Using the international 10-20 EEG system for positioning of transcranial magnetic stimulation", Brain Topogr., vol. 16, no. 2, pp. 95-99, 2003. [http://dx.doi.org/10.1023/B:BRAT.0000006333.93597.9d] [PMID: 14977202]
[3]
V. Jurcak, D. Tsuzuki, and I. Dan, "10/20, 10/10, and 10/5 systems revisited: Their validity as relative head-surface-based positioning systems", Neuroimage, vol. 34, no. 4, pp. 1600-1611, 2007. [http://dx.doi.org/10.1016/j.neuroimage.2006.09.024] [PMID: 17207640]
[4]
B. Zhang, J. Wang, and T. Fuhlbrigge, "A review of the commercial brain-computer interface technology from perspective of industrial robotics", 2010 IEEE International Conference on Automation and Logistics. pp. 379-384, 2010. [http://dx.doi.org/10.1109/ICAL.2010.5585311]
[5]
W. Speier, C. Arnold, and N. Pouratian, "Evaluating true BCI communication rate through mutual information and language models", PLoS One, vol. 8, no. 10, p. e78432, 2013. [http://dx.doi.org/10.1371/journal.pone.0078432] [PMID: 24167623]
224 Computational Intelligence for Data Analysis, Vol. 2
Pai et al.
[6]
J.J. Daly, and J.R. Wolpaw, "Brain–computer interfaces in neurological rehabilitation", Lancet Neurol., vol. 7, no. 11, pp. 1032-1043, 2008. [http://dx.doi.org/10.1016/S1474-4422(08)70223-0] [PMID: 18835541]
[7]
J.M. Rosenow, Anatomy of the Nervous System.Neuromodulation. Elsevier, 2018, pp. 25-39. [http://dx.doi.org/10.1016/B978-0-12-805353-9.00003-6]
[8]
C.C. McIntyre, Patient-Specific Modeling of Deep Brain Stimulation. In: Neuromodulation. Elsevier, 2018, pp. 129-135. [http://dx.doi.org/10.1016/B978-0-12-805353-9.00012-7]
[9]
C. Chen, X. Xiao, A.N. Belkacem, L. Lu, X. Wang, W. Yi, P. Li, C. Wang, S. Sha, X. Zhao, and D. Ming, "Efficacy Evaluation of Neurofeedback-Based Anxiety Relief", Front. Neurosci., vol. 15, p. 758068, 2021. [http://dx.doi.org/10.3389/fnins.2021.758068] [PMID: 34776855]
[10]
J.P. Donoghue, Brain–Computer Interfaces.Neuromodulation. Elsevier, 2018, pp. 341-356. [http://dx.doi.org/10.1016/B978-0-12-805353-9.00025-5]
[11]
Z. Cao, "A review of artificial intelligence for EEG‐based brain−computer interfaces and applications", Brain Science Advances, vol. 6, no. 3, pp. 162-170, 2020. [http://dx.doi.org/10.26599/BSA.2020.9050017]
[12]
M.K. Singh, I. Saini, and N. Sood, "A multi-functional BCI system for exigency assistance and environment control based on ML and IoT", Int. J. Comput. Appl. Technol., vol. 63, no. 1/2, pp. 64-82, 2020. [http://dx.doi.org/10.1504/IJCAT.2020.107912]
[13]
S. Chaudhary, S. Taran, V. Bajaj, and S. Siuly, "A flexible analytic wavelet transform based approach for motor-imagery tasks classification in BCI applications", Comput. Methods Programs Biomed., vol. 187, p. 105325, 2020. [http://dx.doi.org/10.1016/j.cmpb.2020.105325] [PMID: 31964514]
[14]
Z. Jin, G. Zhou, D. Gao, and Y. Zhang, "EEG classification using sparse Bayesian extreme learning machine for brain–computer interface", Neural Comput. Appl., vol. 32, no. 11, pp. 6601-6609, 2020. [http://dx.doi.org/10.1007/s00521-018-3735-3]
[15]
H. Yi, "Efficient machine learning algorithm for electroencephalogram modeling in brain–computer interfaces", Neural Comput. Appl., vol. 3, 2020. [http://dx.doi.org/10.1007/s00521-020-04861-3]
[16]
X. Zheng, W. Chen, Y. You, Y. Jiang, M. Li, and T. Zhang, "Ensemble deep learning for automated visual classification using EEG signals", Pattern Recognit., vol. 102, p. 107147, 2020. [http://dx.doi.org/10.1016/j.patcog.2019.107147]
[17]
J. Chen, Z. Yu, Z. Gu, and Y. Li, "Deep temporal-spatial feature learning for motor imagery-based brain-computer interfaces", IEEE Trans. Neural Syst. Rehabil. Eng., vol. 28, no. 11, pp. 2356-2366, 2020. [http://dx.doi.org/10.1109/TNSRE.2020.3023417] [PMID: 32956061]
[18]
Q. Zhang, and Y. Liu, "Improving brain computer interface performance by data augmentation with conditional Deep Convolutional Generative Adversarial Networks" Jun. 2018. http://arxiv.org/abs/1806.07108
[19]
J. Thomas, T. Maszczyk, N. Sinha, T. Kluge, and J. Dauwels, "Deep learning-based classification for brain-computer interfaces", 2017 IEEE International Conference on Systems, Man, and Cybernetics (SMC), pp. 234-239, 2017. [http://dx.doi.org/10.1109/SMC.2017.8122608]
[20]
S. Rasheed, "A Review of the Role of Machine Learning Techniques towards Brain–Computer Interface Applications", Machine Learning and Knowledge Extraction, vol. 3, no. 4, pp. 835-862, 2021.
Brain-Computer Interface
Computational Intelligence for Data Analysis, Vol. 2 225
[http://dx.doi.org/10.3390/make3040042] [21]
A. Al-Nafjan, M. Hosny, Y. Al-Ohali, and A. Al-Wabil, "Review and Classification of Emotion Recognition Based on EEG Brain-Computer Interface System Research: A Systematic Review", Appl. Sci. (Basel), vol. 7, no. 12, p. 1239, 2017. [http://dx.doi.org/10.3390/app7121239]
[22]
N. Birbaumer and U. Chaudhary, “Learning from brain control: clinical application of brain–computer interfaces,” e-Neuroforum, vol. 6, no. 4, pp. 87–95, 2015. [http://dx.doi.org/10.1007/s13295-015-0015-x]
[23]
B.K. Min, M.J. Marzelli, and S.S. Yoo, "Neuroimaging-based approaches in the brain–computer interface", Trends Biotechnol., vol. 28, no. 11, pp. 552-560, 2010. [http://dx.doi.org/10.1016/j.tibtech.2010.08.002] [PMID: 20810180]
[24]
R. Abbasi, and M. Esmaeilpour, "Selecting Statistical Characteristics of Brain Signals to Detect Epileptic Seizures using Discrete Wavelet Transform and Perceptron Neural Network", International Journal of Interactive Multimedia and Artificial Intelligence, vol. 4, no. 5, p. 33, 2017. [http://dx.doi.org/10.9781/ijimai.2017.456]
[25]
H. Wang, Q. Su, Z. Yan, F. Lu, Q. Zhao, Z. Liu, and F. Zhou, "Rehabilitation Treatment of Motor Dysfunction Patients Based on Deep Learning Brain–Computer Interface Technology", Front. Neurosci., vol. 14, p. 595084, 2020. [http://dx.doi.org/10.3389/fnins.2020.595084] [PMID: 33192282]
[26]
D. Torse, V. Desai, and R. Khanai, “A Review on Seizure Detection Systems with Emphasis on Multidomain Feature Extraction and Classification using Machine Learning,” BRAIN. Broad Res. Artif. Intell. Neurosci., vol. 8, no. 4, pp. 109–129, 2017, [Online]. Available: https://lumenpublishing.com/journals/index.php/brain/article/view/2107
[27]
D. Torse, V. Desai, and R. Khanai, "An optimized design of seizure detection system using joint feature extraction of multichannel EEG signals", J. Biomed. Res., vol. 34, no. 3, pp. 191-204, 2020. [http://dx.doi.org/10.7555/JBR.33.20190019] [PMID: 32561699]
[28]
S. Debener, M. Ullsperger, M. Siegel, and A.K. Engel, "Single-trial EEG–fMRI reveals the dynamics of cognitive function", Trends Cogn. Sci., vol. 10, no. 12, pp. 558-563, 2006. [http://dx.doi.org/10.1016/j.tics.2006.09.010] [PMID: 17074530]
[29]
L. Kauhanen, T. Nykopp, J. Lehtonen, P. Jylänki, J. Heikkonen, P. Rantanen, H. Alaranta, and M. Sams, "EEG and MEG brain-computer interface for tetraplegic patients", IEEE Trans. Neural Syst. Rehabil. Eng., vol. 14, no. 2, pp. 190-193, 2006. [http://dx.doi.org/10.1109/TNSRE.2006.875546] [PMID: 16792291]
[30]
B.K. Min, M.S. Hämäläinen, and D. Pantazis, "New Cognitive Neurotechnology Facilitates Studies of Cortical–Subcortical Interactions", Trends Biotechnol., vol. 38, no. 9, pp. 952-962, 2020. [http://dx.doi.org/10.1016/j.tibtech.2020.03.003] [PMID: 32278504]
[31]
M.C. Piastra, A. Nüßing, J. Vorwerk, M. Clerc, C. Engwer, and C.H. Wolters, "A comprehensive study on electroencephalography and magnetoencephalography sensitivity to cortical and subcortical sources", Hum. Brain Mapp., vol. 42, no. 4, pp. 978-992, 2021. [http://dx.doi.org/10.1002/hbm.25272] [PMID: 33156569]
[32]
A. Bashashati, M. Fatourechi, R.K. Ward, and G.E. Birch, "A survey of signal processing algorithms in brain–computer interfaces based on electrical brain signals", J. Neural Eng., vol. 4, no. 2, pp. R32R57, 2007. [http://dx.doi.org/10.1088/1741-2560/4/2/R03] [PMID: 17409474]
[33]
S. Saha, M.S. Hossain, K. Ahmed, R. Mostafa, L. Hadjileontiadis, A. Khandoker, and M. Baumert, "Wavelet Entropy-Based Inter-subject Associative Cortical Source Localization for Sensorimotor BCI", Front. Neuroinform., vol. 13, p. 47, 2019. [http://dx.doi.org/10.3389/fninf.2019.00047] [PMID: 31396068]
226 Computational Intelligence for Data Analysis, Vol. 2
Pai et al.
[34]
F. Lotte, L. Bougrain, A. Cichocki, M. Clerc, M. Congedo, A. Rakotomamonjy, and F. Yger, "A review of classification algorithms for EEG-based brain–computer interfaces: a 10 year update", J. Neural Eng., vol. 15, no. 3, p. 031005, 2018. [http://dx.doi.org/10.1088/1741-2552/aab2f2] [PMID: 29488902]
[35]
S. Nagel, and M. Spüler, "World’s fastest brain-computer interface: Combining EEG2Code with deep learning", PLoS One, vol. 14, no. 9, p. e0221909, 2019. [http://dx.doi.org/10.1371/journal.pone.0221909] [PMID: 31490999]
[36]
S. Saha, K.A. Mamun, K. Ahmed, R. Mostafa, G.R. Naik, S. Darvishi, "Progress in brain computer interface: challenges and opportunities". Front Syst. Neurosci., vol. 15, 2021. [http://dx.doi.org/10.3389/fnsys.2021.578875] [PMID: 33716680]
[37]
K. Deisseroth, and M.J. Schnitzer, "Engineering approaches to illuminating brain structure and dynamics", Neuron, vol. 80, no. 3, pp. 568-577, 2013. [http://dx.doi.org/10.1016/j.neuron.2013.10.032] [PMID: 24183010]
[38]
T.J. Oxley, P.E. Yoo, G.S. Rind, S.M. Ronayne, C.M.S. Lee, C. Bird, V. Hampshire, R.P. Sharma, A. Morokoff, D.L. Williams, C. MacIsaac, M.E. Howard, L. Irving, I. Vrljic, C. Williams, S.E. John, F. Weissenborn, M. Dazenko, A.H. Balabanski, D. Friedenberg, A.N. Burkitt, Y.T. Wong, K.J. Drummond, P. Desmond, D. Weber, T. Denison, L.R. Hochberg, S. Mathers, T.J. O’Brien, C.N. May, J. Mocco, D.B. Grayden, B.C.V. Campbell, P. Mitchell, and N.L. Opie, "Motor neuroprosthesis implanted with neurointerventional surgery improves capacity for activities of daily living tasks in severe paralysis: first in-human experience", J. Neurointerv. Surg., vol. 13, no. 2, pp. 102-108, 2021. [http://dx.doi.org/10.1136/neurintsurg-2020-016862] [PMID: 33115813]
[39]
F.R. Willett, D.T. Avansino, L.R. Hochberg, J.M. Henderson, and K.V. Shenoy, High-performance brain-to-text communication via handwriting., Nature, vol. 593, no. 7858, pp. 249-254, 2021. [http://dx.doi.org/10.1038/s41586-021-03506-2] [PMID: 33981047]
[40]
X. Zhang, L. Yao, Q.Z. Sheng, S.S. Kanhere, T. Gu, and D. Zhang, "Converting Your Thoughts to Texts: Enabling Brain Typing via Deep Feature Learning of EEG Signals", 2018 IEEE International Conference on Pervasive Computing and Communications (PerCom), pp. 1-10, 2018. [http://dx.doi.org/10.1109/PERCOM.2018.8444575]
[41]
C. Herff, D. Heger, A. de Pesters, D. Telaar, P. Brunner, G. Schalk, and T. Schultz, "Brain-to-text: decoding spoken phrases from phone representations in the brain", Front. Neurosci., vol. 9, p. 217, 2015. [http://dx.doi.org/10.3389/fnins.2015.00217] [PMID: 26124702]
[42]
Q. Rabbani, G. Milsap, and N.E. Crone, "The Potential for a Speech Brain–Computer Interface Using Chronic Electrocorticography", Neurotherapeutics, vol. 16, no. 1, pp. 144-165, 2019. [http://dx.doi.org/10.1007/s13311-018-00692-2] [PMID: 30617653]
[43]
A. Rezazadeh Sereshkeh, R. Yousefi, A.T. Wong, F. Rudzicz, and T. Chau, "Development of a ternary hybrid fNIRS-EEG brain–computer interface based on imagined speech", Brain Comput. Interfaces (Abingdon), vol. 6, no. 4, pp. 128-140, 2019. [http://dx.doi.org/10.1080/2326263X.2019.1698928]
[44]
M. Ptito, M. Bleau, I. Djerourou, S. Paré, F.C. Schneider, and D.R. Chebat, "Brain-Machine Interfaces to Assist the Blind", Front. Hum. Neurosci., vol. 15, p. 638887, 2021. [http://dx.doi.org/10.3389/fnhum.2021.638887] [PMID: 33633557]
[45]
A. Farnum, and G. Pelled, "New Vision for Visual Prostheses", Front. Neurosci., vol. 14, p. 36, 2020. [http://dx.doi.org/10.3389/fnins.2020.00036] [PMID: 32132890]
[46]
H.-J. Hwang, “Brain-Computer Interface Based on Ear-EEG,” in 2021 9th International Winter Conference on Brain-Computer Interface (BCI), pp. 1–3, 2021. [http://dx.doi.org/10.1109/BCI51272.2021.9385299]
[47]
M.J. Lucia, P. Revuelta, A. Garcia, B. Ruiz, R. Vergaz, V. Cerdan, and T. Ortiz, "Vibrotactile
Brain-Computer Interface
Computational Intelligence for Data Analysis, Vol. 2 227
Captioning of Musical Effects in Audio-Visual Media as an Alternative for Deaf and Hard of Hearing People: An EEG Study", IEEE Access, vol. 8, pp. 190873-190881, 2020. [http://dx.doi.org/10.1109/ACCESS.2020.3032229] [48]
A.N. Belkacem, N. Jamil, J.A. Palmer, S. Ouhbi, and C. Chen, "Brain Computer Interfaces for Improving the Quality of Life of Older Adults and Elderly Patients", Front. Neurosci., vol. 14, p. 692, 2020. [http://dx.doi.org/10.3389/fnins.2020.00692] [PMID: 32694979]
[49]
A. Fukushima, R. Morooka, H. Tanaka, H. Kentaro, A. Tugawa, and H. Hanyu, "Classification of dementia type using the brain-computer interface", Artif. Life Robot., vol. 26, no. 2, pp. 216-221, 2021. [http://dx.doi.org/10.1007/s10015-020-00673-9]
[50]
T. M. Rutkowski, M. Koculak, M. S. Abe, and M. Otake-Matsuura, “Brain Correlates of Task–Load and Dementia Elucidation with Tensor Machine Learning Using Oddball BCI Paradigm,” in ICASSP 2019 - 2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pp. 8578–8582, 2019. [http://dx.doi.org/10.1109/ICASSP.2019.8682387]
[51]
L. Ferreira, M. Spinola, J. Camara, S. Bermudez i Badia, and S. Cavaco, “Feasibility of Pitch and Rhythm Musical Distortions as Cueing Method for People with Dementia in AR Cognitive Stimulation Tasks,” in 2021 IEEE 9th International Conference on Serious Games and Applications for Health(SeGAH), pp. 1–8, 2021 [http://dx.doi.org/10.1109/SEGAH52098.2021.9551866]
[52]
Chin-Teng Lin, Shu-Fang Tsai, and Li-Wei Ko, "EEG-based learning system for online motion sickness level estimation in a dynamic vehicle environment", IEEE Trans. Neural Netw. Learn. Syst., vol. 24, no. 10, pp. 1689-1700, 2013. [http://dx.doi.org/10.1109/TNNLS.2013.2275003] [PMID: 24808604]
[53]
C-S. Wei, S-W. Chuang, W-R. Wang, L-W. Ko, T-P. Jung, and C-T. Lin, "Implementation of a motion sickness evaluation system based on EEG spectrum analysis", IEEE International Symposium of Circuits and Systems (ISCAS), pp. 1081-1084, 2011. [http://dx.doi.org/10.1109/ISCAS.2011.5937757]
[54]
S. Pritchett, E. Zilberg, Z. M. Xu, M. Karrar, D. Burton, and S. Lal, “Comparing accuracy of two algorithms for detecting driver drowsiness Single source (EEG) and hybrid (EEG and body movement),” IB2COM 2011 - 6th International Conference on Broadband Communications and Biomedical Applications, Program. pp. 179–184, 2011. [http://dx.doi.org/10.1109/IB2Com.2011.6217916]
[55]
G. Li, and W.Y. Chung, "Combined EEG-Gyroscope-tDCS Brain Machine Interface System for Early Management of Driver Drowsiness", IEEE Trans. Hum. Mach. Syst., vol. 48, no. 1, pp. 50-62, 2018. [http://dx.doi.org/10.1109/THMS.2017.2759808]
[56]
N. Mrachacz-Kersting, and D. Farina, "Modulation of Cortical Excitability with BCI for Stroke Rehabilitation", [http://dx.doi.org/10.1109/IWW-BCI.2019.8737264]
[57]
S. Ruiz, K. Buyukturkoglu, M. Rana, N. Birbaumer, and R. Sitaram, "Real-time fMRI brain computer interfaces: Self-regulation of single brain regions to networks", Biol. Psychol., vol. 95, pp. 4-20, 2014. [http://dx.doi.org/10.1016/j.biopsycho.2013.04.010] [PMID: 23643926]
[58]
S.N. Abdulkader, A. Atia, and M.S.M. Mostafa, "Brain computer interfacing: Applications and challenges", Egyptian Informatics Journal, vol. 16, no. 2, pp. 213-230, 2015. [http://dx.doi.org/10.1016/j.eij.2015.06.002]
[59]
A. Mishra, A. Singh, and A. Ujlayan, "An RFID-Based BCI System for Emotion Detection Using EEG patterns", 2021 IEEE International Conference on RFID Technology and Applications (RFIDTA), pp. 5-8, 2021. [http://dx.doi.org/10.1109/RFID-TA53372.2021.9617423]
228 Computational Intelligence for Data Analysis, Vol. 2
Pai et al.
[60]
H. Lim and J. Ku, “High engagement in BCI action observation game by relevant character’s movement,” in 2019 7th International Winter Conference on Brain-Computer Interface (BCI), pp. 1–3, 2019. [http://dx.doi.org/10.1109/IWW-BCI.2019.8737252]
[61]
Z. Song, T. Fang, J. Ma and Y. Zhang, “Evaluation and Diagnosis of Brain Diseases based on Noninvasive BCI,” in 2021 9th International Winter Conference on Brain-Computer Interface (BCI), pp. 1–6, 2021. [http://dx.doi.org/10.1109/BCI51272.2021.9385291]
[62]
M.P. Hosseini, D. Pompili, K. Elisevich, and H. Soltanian-Zadeh, Optimized Deep Learning for EEG Big Data and Seizure Prediction BCI via Internet of Things., IEEE Trans. Big Data, vol. 3, no. 4, pp. 392-404, 2017. [http://dx.doi.org/10.1109/TBDATA.2017.2769670]
[63]
C.W.N.F.C.W. Fadzal, W. Mansor, and L.Y. Khuan, “Review of brain computer interface application in diagnosing dyslexia,” in 2011 IEEE Control and System Graduate Research Colloquium., 2011, pp. 124-128. [http://dx.doi.org/10.1109/ICSGRC.2011.5991843]
[64]
L. Minkowski, K. V. Mai, and D. Gurve, “Feature Extraction to Identify Depression and Anxiety Based on EEG,” in 2021 43rd Annual International Conference of the IEEE Engineering in Medicine & Biology Society (EMBC), pp. 6322–6325 2021. [http://dx.doi.org/10.1109/EMBC46164.2021.9630821]
[65]
D.S. Moschona, "An Affective Service based on Multi-Modal Emotion Recognition, using EEG enabled Emotion Tracking and Speech Emotion Recognition", 2020 IEEE International Conference on Consumer Electronics - Asia (ICCE-Asia), pp. 1-3, 2020. [http://dx.doi.org/10.1109/ICCE-Asia49877.2020.9277291]
[66]
L. Bonnet, F. Lotte, and A. Lecuyer, "Two Brains, One Game: Design and Evaluation of a Multiuser BCI Video Game Based on Motor Imagery", IEEE Trans. Comput. Intell. AI Games, vol. 5, no. 2, pp. 185-198, 2013. [http://dx.doi.org/10.1109/TCIAIG.2012.2237173]
[67]
M.F. Mridha, S.C. Das, M.M. Kabir, A.A. Lima, M.R. Islam, and Y. Watanobe, "Brain-Computer Interface: Advancement and Challenges", Sensors (Basel), vol. 21, no. 17, p. 5746, 2021. [http://dx.doi.org/10.3390/s21175746] [PMID: 34502636]
[68]
S.K. Mudgal, S.K. Sharma, J. Chaturvedi, and A. Sharma, "Brain computer interface advancement in neurosciences: Applications and issues", Interdiscip. Neurosurg., vol. 20, p. 100694, 2020. [http://dx.doi.org/10.1016/j.inat.2020.100694]
[69]
G.A.M. Vasiljevic, and L.C. de Miranda, "Brain–Computer Interface Games Based on ConsumerGrade EEG Devices: A Systematic Literature Review", Int. J. Hum. Comput. Interact., vol. 36, no. 2, pp. 105-142, 2020. [http://dx.doi.org/10.1080/10447318.2019.1612213]
Computational Intelligence for Data Analysis, 2023, Vol. 2, 229-247
229
CHAPTER 14
Data Augmentation with Image Fusion Techniques for Brain Tumor Classification using Deep Learning Tarik Hajji1,*, Ibtissam Elhassani1, Tawfik Masrour1, Imane Tailouloute1 and Mouad Dourhmi1 Artificial Intelligence for Engineering Sciences Team (IASI), Laboratory of Mathematical Modeling, Simulation and Smart Systems (L2M3S), ENSAM, Moulay Ismail University, 50500 Meknes, Morocco 1
Abstract: Brain tumor (BT) is a serious cancerous disease caused by an uncontrollable and abnormal distribution of cells. Recent advances in deep learning (DL) have helped the healthcare industry in medical imaging for the diagnosis of many diseases. One of the major problems encountered in the automatic classification of BT when using machine learning (ML) techniques is the availability and quality of the learning from data; these are often inaccessible, very confidential, and of poor quality. On the other hand, there are more than 120 types of BT [1] that we must recognize. In this paper, we present an approach for the automatic classification of medical images (MI) of BT using image fusion (IF) with an auto-coding technique for data augmentation (DA) and DL. The objective is to design and develop a diagnostic support system to assist the practitioner in analyzing never-seen BT images. To address this problem, we propose two contributions to perform data augmentation at two different levels: before and during the learning process. Starting from a small dataset, we conduct the first phase of classical DA, followed by the second one based on the image fusion technique. Our approach allowed us to increase the accuracy to a very acceptable level compared to other methods in the literature for ten tumor classes.
Keywords: Artificial Intelligence, Autoencoder, Brain Tumor, CNN, Data Augmentation, Deep Learning, Image Fusion, Machine Learning, Medicine 4.0, Visual Recognition. INTRODUCTION The emerging medicine, Medicine 4.0 [2], has dared to take the step of digitization; three-dimensional imaging, the connectionist power of artificial * Corresponding author Tarik Hajji: Artificial Intelligence for Engineering Sciences Team (IASI), Laboratory of Mathematical Modeling, Simulation and Smart Systems (L2M3S), ENSAM, Moulay Ismail University, 50500 Meknes, Morocco; E-mail: [email protected]
Mariya Ouaissa, Mariyam Ouaissa, Zakaria Boulouad, Inam Ullah Khan, Sailesh Iyer (Eds.) All rights reserved-© 2023 Bentham Science Publishers
230 Computational Intelligence for Data Analysis, Vol. 2
Hajji et al.
intelligence (AI) algorithms, surgical planning with specific software using AI, and the creation of personalized or adaptable implants to face the most complex situations. There are few industries where visual recognition (VR) [3] does not open up huge opportunities. VR technologies are capable of assimilating the textual content of photographed documents, and this ability could have a major impact on employment. The key will probably be to consider these technological solutions not in opposition, but as a complement to human work. This was proven by a Harvard study in the health sector. It found that DL alone effectively analyzes magnetic resonance imaging (MRI) in 92% of cases. For their part, doctors achieve a rate of 96%. But helped by AI, their rate exceeds 99.5% [4]. The Brain can be infected by a lot of tumors, principally if they are malignant or [5] persistent. According to the American association of neurological surgeons, there are 11 BT main classes: gliomas, meningioma, primitive neuroectodermal tumors, pituitary tumors, pineal tumors, choroid plexus tumors, dysembryoplastic neuroepithelial tumor, tumors of nerves sheaths, cysts, skull base, and primary central nervous system lymphoma [5, 6]. Each main class is subdivided into several subclasses to have in total more than 120 types of BT according to the World Health Organization. DL is an application of ML, a field of AI, which represents a set of algorithms that automatically learn to recognize and classify data such as images. DL is inspired by the structure of the human brain and is heavily used in computer vision problems. The challenge of the use of DL in computer vision is how to find the best model (conventional neural network architecture and parameters). This operation requires a data set of labeled images and a specific learning algorithm to optimize the DL model parameters. To solve the BT classification problem, we propose in this paper to use a combination of techniques to achieve very high recognition rates. These techniques are based on convolutional neural networks (CNN), data augmentation, image fusion and autoencoders. We have proven that the right combination of these techniques can overcome the problem of lack of data. The rest of this paper is organized as follows: state of the art to present the different concepts used in this paper, such as DL, CNN, and the different techniques of DA and IF. The section of related works for the presentation of some scientific results concerning the subject of BT classification. The methodology section to the general presentation of our approach is subdivided into three large parts, one for the standard classification, one for the use of the DA techniques, and one for the use of the IF technique based on autoencoders. The results and discussion section for the presentation of the results obtained for each
Image Fusion Techniques for Brain
Computational Intelligence for Data Analysis, Vol. 2 231
proposed approach and a possible comparison with the state-of-the-art methods, and at the end, a general conclusion is proposed. BACKGROUND The history of medicine remains inseparable from innovation, driven by generations of designers, engineers, doctors, clinicians, and then developers, startups, data scientists, and other researchers [7]. AI-based accurate diagnosis (diabetic retinopathy fundus diagnosis, the first direct AI diagnostic algorithm to be recognized by the Food and Drug Administration, in April 2018) [8], automatic and instantaneous skin cancer recognition and classification [9], automatic electrocardiogram reading [10], arrhythmia detection [11] and personalized chemotherapy [12]. Deep Learning DL is one of the main ML technologies subfields of Artificial Intelligence [13]. The ML concept dates back to the middle of the 20th century. In the 1950s, Alan Turing, the British mathematician, imagined a machine able to learn, a “Learning Machine” [14]. Over the following decades, various ML techniques were generated to create algorithms to learn and improve themselves. Among these techniques are artificial neural networks (ANN) [15]. These algorithms are the basis for Deep Learning, but also for technologies such as image recognition or robotic vision. ANN is inspired by the neurons of the human brain [16]. They are made up of several artificial neurons connected to each other (Fig. 1). ML
Input Layer
Output Layer
DL
Hidden Layer
Fig. (1). DL is implemented by a deeper ANN architecture with multiple hidden layers.
Data Augmentation CV processing models use a DA approach to cope with data sparingly and insufficient multiplicity [17]. DA algorithms can increase the accuracy of DL models and perform better in terms of training loss accuracy than a DL model without DA for image [18]. There are a lot of DA methods for images to create diversity in the learning data set of the model. It is easy to find many examples of
232 Computational Intelligence for Data Analysis, Vol. 2
Hajji et al.
coding for these transformations, like adding noise, cropping, flipping, rotation, scaling, translation, brightness, contrast, color augmentation, saturation, adversarial training, neural style transfer, and generative adversarial networksbased augmentation (Fig. 2) [19].
Fig. (2). Example of three simple data augmentation operations (rotation, shearing and scaling).
Image Fusion Image Fusion encompasses all data analysis strategies, displayed in Fig. (3), aimed at combining information from multiple images obtained with the same or different spectroscopic platforms [20]. IF combines two or more registered images of the same object into a single image that is more easily interpreted than any of the originals [21]. The motivation for Medical Image fusion from different modalities is to obtain a high-quality image by intelligently combining the collected essential information from multi-modal input images [22]. IF techniques shown in figure can be classified as spatial and frequency domains.
Image Fusion Techniques for Brain
Computational Intelligence for Data Analysis, Vol. 2 233 Image Fusion
Frequency Domain
Spatial Domain
Simple Average Minimum Maxmum Max-Min Simple Block Replace Weighted Averaging HIS Brovey PCA Guided Filtering
Deep Learning Based Function
Laplacian Pyramid Decomposition Based Image Fusion
Discrete Transform Based Image Fusion Discrete Cosine Transform Wavelet Transform Kekres Wavelet Transform Kekres Hybrid Wavelet Transform Stationary Wavelet Transform Combination of Curvelet and Stationary Wavelet Transform
Fig. (3). Different image fusion techniques [20].
Related Work Several contributions have been proposed to address the topic of tumor classification using several classical and advanced techniques [23 - 31]. It should be noted that the most modern techniques use DL to treat this subject. For example, the authors of [23] proposed to use CNNs for BT classification via the transfer learning (TL) technique. The anticipated method reports a mean classification accuracy of 98%, outperforming all related work. The manuscript focuses on a 3-class BT to adopt the concept of deep TL using GoogLeNet [32] to extract features from MRI. A decision support system for multimodal BT using DL is proposed by Sharif et al. [24]. For the experimental process, the authors used two datasets, BRATS2018, and BRATS2019, and achieved an accuracy of more than 95%. Mohsen et al. [25] proposed a CNN to classify the BT into three classes (glioma, meningioma, and pituitary) with an accuracy of 98.93% and a sensitivity of 98.18% for cropped lesions, while the results for uncropped lesions are 99%
234 Computational Intelligence for Data Analysis, Vol. 2
Hajji et al.
accuracy and 98.52% sensitivity and the results for segmented lesion images are 97.62% for accuracy and 97.40% sensitivity. Saleh et al. [26] used CNN combined with discrete wavelet transform (DWT), and principal component analysis, to classify 66 brain MRIs into 4 classes: normal, glioblastoma, sarcoma, and metastatic bronchogenic carcinoma tumors, as well as the performance evaluation, was relatively good for all performance measures. To bring more scientific value to state of the art, we present an approach that uses data augmentation with fusion techniques for brain tumor classification using deep learning and autoencoder. METHODOLOGY Using a relatively small dataset that we constructed in our L2M3S lab consisting of 24*10 = 240 images representing 10 classes of BT, we began our three-step approach shown in Fig. (4). Collecting Data
Data cleaning
Using Image fusion techniques
Using images augmentation techniques
CNN model
CNN model
Average
Haar wavelet
Daubechies wavelet
CNN model
CNN model
CNN model
Comparing the results
Fig. (4). The proposed methodology consists of three steps: classification without data augmentation, classification with classical data augmentation techniques and classification with data augmentation using fusion techniques.
Image Fusion Techniques for Brain
Computational Intelligence for Data Analysis, Vol. 2 235
Dataset The dataset in Fig. (5) consists of two types of images (test and training), each type consisting of 10 classes (aneurysm, multiple sclerosis, hydrocephalus, stroke, infections, cysts, swelling, hemorrhage, bleeding, inflammation). For the training images, we have 24 images for each class, and for the test images, we have 8 images for each class.
MRI_ANEURYSM S_1
TAR1MRI_ANEUR YSMS_1
TAR2MRI_ANEUR YSMS_1
TAR3MRI_ANEUR YSMS_1
TAR4MRI_ANEUR YSMS_1
TAR5MRI_ANEUR YSMS_1
TAR6MRI_ANEUR YSMS_1
TAR7MRI_ANEUR YSMS_1
TAR8MRI_ANEUR YSMS_1
TAR9MRI_ANEUR YSMS_1
TAR10MRI_ANEU RYSMS_1
TAR11MRI_ANEU RYSMS_1
TAR111tarikMRI_ ANEURYSMS_1
TAR112tarikMRI_ ANEURYSMS_1
Fig. (5). Extracted from the dataset used.
Deep Learning Approach with Classical Data Augmentation We used google colab to take advantage of the computational power of google; we started by loading our dataset and preprocessing the images. After that, we find out the average dimensions of images, as shown in Fig. (6).
0 64.4 10 20
64.2
30 64.0
40 50
63.8 60 0
10
20
30
40
50
60 63.6
63.6
Fig. (6). The average dimensions of brain-tumor.
63.8
64.0
64.2
64.4
236 Computational Intelligence for Data Analysis, Vol. 2
Hajji et al.
Data Pre-Processing for the Model There is too much data for us to charge all of them into memory. We are able to use some Keras functions (we used ImageDataGenerator to do this automatically) to automatically process the data, generate a batch data stream from a directory and also manipulate the images. It is usually a good idea to manipulate the images with some simple parameters (Table 1), like rotation, resizing and scaling, so that the model becomes more robust to the different images that our dataset does not have, as the image in Fig. (7). Table 1. Configuration of the data augmentation generator. Parameter
Value
Description
Rotation_range
20
Rotate the image 20 degrees
Width_shift_range
0.10
Shift the pic width by a max of 5%
Height_shift_range
0.10
Shift the pic height by a max of 5%
Rescale
1/255
Rescale the image by normalizing it
Shear_range
0.1
Shear means cutting away part of the image (max 10%)
Zoom_range
0.1
Zoom in by 10% max
Horizontal_flip
True
Allo horizontal flipping
Fill_mode
nearest
Fill in missing pixels with the nearest filled value
0
0
10
10
20
20
Image Data Generator
30
30
40
40
50
50
60
0
10
20
30
40
50
60
60
0
10
20
30
40
50
60
Fig. (7). Example of automatic generation of an image.
Generation of many Manipulated Images from a Directory To use flow_from_directory, we organized the images into subdirectories. This is an absolute requirement, otherwise, the method will not work. The directories should only contain images of one class, like in Fig. (8), so one folder per image class.
Image Fusion Techniques for Brain
Computational Intelligence for Data Analysis, Vol. 2 237
Tests MRI_ANEURYSMS_1 MRI_ ANEURYSMS_2 MRI_ANEURYSMS_3 MRI_ANEURYSMS_4 MRI_ANEURYSMS_5 MRI_ANEURYSMS_6 MRI_ANEURYSMS_7 MRI_ANEURYSMS_8 MRI_ANEURYSMS_9 MRI_ANEURYSMS_10
Trains MRI_ANEURYSMS_1 MRI_ANEURYSMS_2 MRI_ANEURYSMS_3 MRI_ANEURYSMS_4 MRI_ANEURYSMS_5 MRI_ANEURYSMS_6 MRI_ANEURYSMS_7 MRI_ANEURYSMS_8 MRI_ANEURYSMS_9 MRI_ANEURYSMS_10
Fig. (8). Organization of the data set into a training database and a test database.
Design of the Model Architecture For visual learning and image recognition, the CNN model, like in Fig. (9), is the most common and widely used machine learning algorithm.
Conv_1 Convolution (5 x 5) kernel valid padding
Max-Pooling (2 x 2)
Conv_2 Convolution (5 x 5) kernel valid padding
fc_3 Fully-Connected Neural Network ReLU activation Max-Pooling (2 x 2)
fc_4 Fully-Connected Neural Network (with dropout)
0 1
Fla tte n
ed
INPUT (28 x 28 x 1)
n1 channels (24 x 24 x n1)
n1 channels (12 x 12 x n1)
n2 channels (8 x 8 x n2)
2 9
n2 channels (4 x 4 x n2)
OUTPUT n3 units
Fig. (9). General architecture of a convolution neural network.
Convolution Layer The convolution layer is the key component of CNN, and is always at least their first layer. Its objective is to identify the presence of a set of features in the images received as input. To do this, we perform convolution filtering: the principle is to “drag” a window representing the feature on the image, and to compute the conv-
238 Computational Intelligence for Data Analysis, Vol. 2
Hajji et al.
olution product between the feature and each portion of the scanned image. A feature is then seen as a filter: the two terms are equivalent in this context. Pooling Layer This is a subsampling operation typically applied after a convolutional layer. In particular, the most popular types of pooling are max and average pooling, where the maximum and average values are taken, respectively. Max-Pooling is a sample-based discretization process. Its objective is to subsample an input representation (image, hidden layer output matrix, etc.) by reducing its dimension. Moreover, its interest is that it reduces the computational cost by reducing the number of parameters to be learned and provides invariance by small translations (if a small translation does not modify the maximum of the scanned region, the maximum of each region will remain the same and thus the new matrix created will remain identical). Flatten Layer The Flatten layer provides a connection between the convolution layers and the base layers of Deep Learning. It allows diffusing the data through the layers by reducing its dimension. The data finally reaches a prediction layer, such as the Dense layer, which provides the label detected by the Deep Learning model. Dense Layer A Dense layer which is a classical neural layer, completely connected with the previous and the next layer. A dense output layer with one neuron per class, and a “softmax” activation function that gives an output probability for each neuron, with the output neuron with the highest probability then deciding that its associated class is the predicted class. Fig. (10) shows the architecture of the model used. Learning and Same Parameters In classification tasks, it would be useful to have all outputs fall between 0 and 1. These values can then present probability assignments for each class. There are 2 main types of multi-class situations: non-exclusive and mutually exclusive Classes. We used a configuration for the case of a 10-class exclusive classification problem whose code is presented in Table 2.
Image Fusion Techniques for Brain
Computational Intelligence for Data Analysis, Vol. 2 239
Fig. (10). The architecture used consists of three convolution layers followed by a classifier with two hidden layers, Total params: 4,625,674. Table 2. Categorization of 10 classes to represent the BTs studied. Brain Tumor
Class Categorization
Primary Tumors of the Brain (Gliomas)
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1
Meningioma
0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0
Primitive NeuroEctodermal Tumors (PNET)
0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0
240 Computational Intelligence for Data Analysis, Vol. 2
Hajji et al.
(Table 2) cont.....
Brain Tumor
Class Categorization
Pituitary Tumors
0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0
Pineal Tumors
0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0
Choroid Plexus Tumors
0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0
Dysembroplastic Neuroepithelial Tumor (DNT)
0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0
Tumors of Nerves and/or Nerve Sheaths
0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0
Cysts
0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0
Skull Base/Chondroma
0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0
Data Augmentation: A Comparative Study In order to train our model, we need huge amounts of data. Indeed, the quantity and especially the quality of our dataset will have a major role in the development of a good model. The principle of data augmentation is based on the principle of artificially augmenting our data, by applying transformations. We will be able to increase the diversity and, thus, the learning field of our model, which will be able to adapt better to predict new data. The principal parameters of this method are shown in Table 3.
Image Fusion Techniques for Brain
Computational Intelligence for Data Analysis, Vol. 2 241
Table 3. The proposed data augmentation parameters as an alternative to the default values. Parameter
Default Value
Proposed Value
Rotation_range
20
15
Width_shift_range
0.10
0.05
Height_shift_range
0.10
0.05
Rescale
1/255
1/255
Shear_range
0.1
0.75
Zoom_range
0.1
Zoom in by 10% max
Horizontal_flip
True
True, Vertical_flip = True
Fill_mode
nearest
Brightness_range=[0.1, 1.5]
Data Augmentation with Image Fusion The quality of the medical images affects the diagnostic result; Different sensors produce different characteristics or marks. The IF allows us to recover these different characteristics and to group them in a single image that presents a lot of useful information. It is very difficult to describe the morphological structure of an organ with only one type of image. To establish an efficient solution to extract the important information from the images and then inject them into a single image as shown in Fig. (11), first, an auto encoder is used for code generation, continuing the important information of each images, then NSST is used to decompose the source image into LFS and HFS, then fusion is performed based on fuzzy logic system (FLS) and new summodified-laplacian (NSML) fusion rules respectively, and finally the membership functions (MFs) of The FLSs are optimized by particle swarm optimization (PSO) to get the optimal fusion result. Auto-Encoder Architecture The auto-encoder (Fig. 12) is a specific architecture of a neural network; it is composed of two sub-networks, the first one for encoding the images and the second one for decoding.
242 Computational Intelligence for Data Analysis, Vol. 2 Endoder
Image A
Hajji et al.
LSF and HSF
LSF Fusion
LSF HSF Fusion HSF
LSF Decoder Image B
HSF Image reconstruction
Fig. (11). Data augmentation by image fusion using autoencoders.
Encoder
Input
output
Decoder
Fig. (12). Operating principle of the autoencoders.
RESULTS AND DISCUSSION CNN Result without Data Augmentation We used the accuracy and loss functions to evaluate the proposed model and obtained a rate of 92.85%. Fig. (13) shows the training and validation accuracy and loss.
Image Fusion Techniques for Brain
Computational Intelligence for Data Analysis, Vol. 2 243
Training and Validation Accuracy
Training and Validation Loss Training Loss Validation Loss
1.0
2.0 0.8
1.5 0.6
1.0 0.4
0.5 0.2
Training Accuracy Validation Accuracy
0
2
4
6
8
0
2
4
6
8
Fig. (13). Training and validation accuracy and loss of CNN without data augmentation.
CNN Result with Data Augmentation Automatic Generator The use of proposed data augmentation parameters allowed us to expect a rate of 94.99, with an increase of 2.24%. Fig. (14) shows the training and validation accuracy and loss. Training and Validation Accuracy
Training and Validation Loss Training Loss Validation Loss
1.0 2.5 0.8 2.0 0.6 1.5 0.4
1.0
0.2
0.5 Training Accuracy Validation Accuracy
0.0 0
25
50
75
100 125 150
175
0.0
200
0
25
50
75
100 125 150
175
200
Fig. (14). Training and validation accuracy and loss of CNN using data augmentation proposed parameters.
CNN Result-Based DA using IF with BWT The use of Daubechies wavelet transforms like an image fusion for data augmentation allowed us to expect a rate of 97.33%. Fig. (15) shows the training and validation accuracy and loss.
244 Computational Intelligence for Data Analysis, Vol. 2
Hajji et al. Training and Validation Loss
Training and Validation Accuracy
Training Loss Validation Loss
1.0 2.0 0.8 1.5 0.6 1.0 0.4 0.5 0.2 Training Accuracy Validation Accuracy 0
20
40
60
80
0.0 0
100
20
40
60
80
100
Fig. (15). Result-based data augmentation using image fusion with Daubechies wavelet transform.
CNN Result-Based DA using IF with Auto-Encoder Proposed Approach The use of the autoencoder as an image fusion process to achieve the data increase to obtain a perfect rate of 100%. Fig. (16) shows the training and validation accuracy and loss. Table 4 provides a summary of the most important results. Training and Validation Accuracy
Training and Validation Loss Training Loss Validation Loss
1.0 2.0 0.8 1.5 0.6 1.0 0.4 0.5 0.2 Training Accuracy Validation Accuracy 0
20
40
60
80
0.0 0
100
20
40
60
80
100
Fig. (16). The proposed approach results in CNN and DA using IF with auto-encoder.
Table 4. Summary of the most important results. -
Accuracy
Validation Accuracy
Loss Validation Loss Score
CNN model
0.889
1
0.0417
0.0342
0.928
Data augmentation
0.9429
1
0.0243
0.0092
0.9499
Daubechies wavelet
0.97
0.99
0.0168
0.0027
0.9733
Auto encoder
1
1
36.10
19.10
1
-5
-5
Image Fusion Techniques for Brain
Computational Intelligence for Data Analysis, Vol. 2 245
CONCLUSION There are more than 500 classes of Brain Tumors to detect, and it is very difficult to find enough data for each type. To manage this problem, we propose to use several techniques combined together for DA. First, we built a local database composed of 10 different classes, and we made a comparative study to choose the most suitable CNN architecture for our case. Using our CNN architecture without any DA techniques yielded a 92.88% rate. Afterward, we made another comparative study to choose the most efficient parameterization for the automatic generator of the data increase, and we were able to make consistent progress of about 94.99%. Then we moved on to using IF as a technique to do DA. The use of the DWT method gives a rate of 97.33%. Our proposed approach is based on autoencoders as an image fusion technique for data augmentation. It gives very satisfactory results and an excellent rate. Our perspective is to make a recognition system for other types of brain tumors using other data augmentation techniques such as generative adversarial networks GAN. CONSENT FOR PUBLICATON Declared none. CONFLICT OF INTEREST The author declares no conflict of interest, financial or otherwise. ACKNOWLEDGEMENT Declared none. REFERENCES [1]
The Johns Hopkins University, Hospital and Health System, “Brain Tumor Types“. [Accessed: 24rd Mars. 2022]: https://www.hopkinsmedicine.org/health/conditions-and-diseases/brain-tumor/brin-tumor-type
[2]
B. Wolf, C. Scholze. Medicine 4.0. Curr. Dir. Biomed. Eng., vol. 3, pp. 183–186, 2017.
[3]
T. Serre, G. Kreiman, M. Kouh, C. Cadieu, U. Knoblich, T. Poggio. "A quantitative theory of immediate visual recognition". Prog. Brain Res., vol. 165, pp. 33-56, 2007.
[4]
D. Nicola. AI equal with human experts in medical diagnosis, study finds. The Guardian. vol. 24, 2019.
[5]
H. Tarik, "Towards an Improved CNN architecture for brain tumor classification", International Conference Europe Middle East & North Africa Information Systems and Technologies to Support Learning, pp. 224-234, 2019.
[6]
Y. Douzi, T. Hajji, M. Benabdellah, A. Azizi, T. Masrour, "Classification and Watermarking of Brain Tumor Using Artificial and Convolutional Neural Networks. In the International Conference on Artificial Intelligence & Industrial Applications". Springer, Cham. pp. 61-77, 2020.
246 Computational Intelligence for Data Analysis, Vol. 2
Hajji et al.
[7]
F. Ursin, C. Timmermann, M. Orzechowski and F. Steger, “Diagnosing Diabetic Retinopathy With Artificial Intelligence: What Information Should Be Included to Ensure Ethical Informed Consent?.” Front. Med. vol. 8, 2021. [http://dx.doi.org/10.3389/fmed.2021.695217]
[8]
H. Moutei, A. Bennis, F. Chraibi, M. Abdellaoui, I. Benatiya. L’intelligence artificielle en ophtalmologie. Journal de la Société Marocaine d’Ophtalmologie, vol. 30, no 2, pp. 8-14, 2021.
[9]
N.C.F. Codella, D. Gutman, M.E. Celebi, B. Helba, M.A. Marchetti, S.W. Dusza, A. Kalloo, K. Liopyris, N. Mishra, H. Kittler, and A. Halpern, Skin lesion analysis toward melanoma detection: A challenge at the 2017 international symposium on biomedical imaging (isbi), hosted by the international skin imaging collaboration (isic). In : 2018 IEEE 15th international symposium on biomedical imaging (ISBI 2018), pp. 168-172, 2018.
[10]
A.H. Ribeiro, M.H. Ribeiro, G.M.M. Paixão, D.M. Oliveira, P.R. Gomes, J.A. Canazart, M.P.S. Ferreira, C.R. Andersson, P.W. Macfarlane, W. Meira Jr., T.B. Schön and A.L.P. Ribeiro, Automatic diagnosis of the 12-lead ECG using a deep neural network. Nat. Commun., vol. 11, no 1, p. 1-9, 2020 [PMID: 31911652]
[11]
C.P. Karunadas, and C. Mathew, "Comparison of arrhythmia detection by conventional Holter and a novel ambulatory ECG system using patch and Android App, over 24 h period", Indian Pacing Electrophysiol. J., vol. 20, no. 2, pp. 49-53, 2020. [http://dx.doi.org/10.1016/j.ipej.2019.12.013] [PMID: 31866554]
[12]
N. André, M. Carré, and E. Pasquier, "Metronomics: towards personalized chemotherapy?", Nat. Rev. Clin. Oncol., vol. 11, no. 7, pp. 413-431, 2014. [http://dx.doi.org/10.1038/nrclinonc.2014.89] [PMID: 24913374]
[13]
Y. Douzi, N. Kannouf, T. Hajji, T. Boukhana, M. Benabdellah, and A. Azizi, "Recognition textures of the tumors of the medical pictures by neural networks", J. Eng. Appl. Sci. (Asian Res. Publ. Netw.), vol. 13, pp. 4020-4024, 2018.
[14]
A.L. Fradkov, "Early history of machine learning", IFAC-PapersOnLine, vol. 53, no. 2, pp. 13851390, 2020. [http://dx.doi.org/10.1016/j.ifacol.2020.12.1888]
[15]
T. Hajji, A. Alami Hassani, and M. Ouazzani Jamil, "Incidents prediction in road junctions using artificial neural networks", IOP Conf. Series Mater. Sci. Eng., vol. 353, no. 1, p. 012017, 2018. [http://dx.doi.org/10.1088/1757-899X/353/1/012017]
[16]
H. Tarik, and J.E. Miloud, "Digital Watermarking and Signing by Artificial Neural Networks", American Journal of Intelligent Systems, vol. 4, pp. 21-31, 2014.
[17]
J. Wei, Z. Kai, W. Nan, Y. Miao,"MeshCut data augmentation for deep learning in computer vision." PLoS One, vol. 15, no. 12, p. e0243613, 2020. [http://dx.doi.org/10.1371/journal.pone.0243613]
[18]
J. Nalepa, M. Marcinkiewicz, and M. Kawulok, "Data augmentation for brain-tumor segmentation: a review", Front. Comput. Neurosci., vol. 13, p. 83, 2019. [http://dx.doi.org/10.3389/fncom.2019.00083] [PMID: 31920608]
[19]
"Connor et KHOSHGOFTAAR, Taghi M. A survey on image data augmentation for deep learning", J. Big Data, vol. 6, no. 1, pp. 1-48, 2019.
[20]
H. Kaur, D. Koundal, and V. Kadyan, "Harpreet, KOUNDAL, Deepika, et KADYAN, Virender. Image fusion techniques: a survey", Arch. Comput. Methods Eng., vol. 28, no. 7, pp. 4425-4447, 2021. [http://dx.doi.org/10.1007/s11831-021-09540-7] [PMID: 33519179]
[21]
H. Ghassemian, "A review of remote sensing image fusion methods", Inf. Fusion, vol. 32, pp. 75-89, 2016. [http://dx.doi.org/10.1016/j.inffus.2016.03.003]
Image Fusion Techniques for Brain
Computational Intelligence for Data Analysis, Vol. 2 247
[22]
"Sneha et GUPTA, Deep. Detail enhanced feature-level medical image fusion in decorrelating decomposition domain", IEEE Trans. Instrum. Meas., vol. 70, pp. 1-9, 2020.
[23]
S. Deepak, and P.M. Ameer, Brain tumor classification using deep CNN features via transfer learning., Comput. Biol. Med., vol. 111, p. 103345, 2019. [http://dx.doi.org/10.1016/j.compbiomed.2019.103345] [PMID: 31279167]
[24]
M. I. Sharif, M. A. Khan, M. Alhussein, K. Aurangzeb, and M. Raza, "A decision support system for multimodal brain tumor classification using deep learning". Complex Int. Sys., pp. 1-14, 2021.
[25]
H. Mohsen, E.S.A. El-Dahshan, E.S.M. El-Horbaty, and A.B.M. Salem, "Classification using deep learning neural networks for brain tumors", Future Comput. Inform. J., vol. 3, no. 1, pp. 68-71, 2018. [http://dx.doi.org/10.1016/j.fcij.2017.12.001]
[26]
A. Saleh, R. Sukaik, and S.S. Abu-Naser, "Brain tumor classification using deep learning", In 2020 International Conference on Assistive and Rehabilitation Technologies (iCareTech), pp. 131-136, 2020. [http://dx.doi.org/10.1109/iCareTech49914.2020.00032]
[27]
M. Sajjad, S. Khan, K. Muhammad, W. Wu, A. Ullah, and S.W. Baik, "Multi-grade brain tumor classification using deep CNN with extensive data augmentation", J. Comput. Sci., vol. 30, pp. 174182, 2019. [http://dx.doi.org/10.1016/j.jocs.2018.12.003]
[28]
J.S. Paul, A.J. Plassard, B.A. Landman, and D. Fabbri, Deep learning for brain tumor classification. In Medical Imaging 2017: Biomedical Applications in Molecular, Structural, and Functional Imaging, Vol. 10137, pp. 253-268, 2017.
[29]
A.M. Sarhan, "Brain tumor classification in magnetic resonance images using deep learning and wavelet transform", J. Biomed. Sci. Eng., vol. 13, no. 6, pp. 102-112, 2020. [http://dx.doi.org/10.4236/jbise.2020.136010]
[30]
J. Seetha, and S.S. Raja, "Brain tumor classification using convolutional neural networks", Biomed. Pharmacol. J., vol. 11, no. 3, pp. 1457-1461, 2018. [http://dx.doi.org/10.13005/bpj/1511]
[31]
V.G.P. Rathi, and S. Palani, "Brain tumor detection and classification using deep learning classifier on MRI images", Res. J. Appl. Sci. Eng. Technol., vol. 10, no. 2, pp. 177-187, 2015.
[32]
R.U. Khan, X. Zhang, and R. Kumar, "Analysis of ResNet and GoogleNet models for malware detection", Journal of Computer Virology and Hacking Techniques, vol. 15, no. 1, pp. 29-37, 2019. [http://dx.doi.org/10.1007/s11416-018-0324-z]
248
Computational Intelligence for Data Analysis, 2023, Vol. 2, 248-263
CHAPTER 15
Convergence Towards Blockchain-Based Patient Health Record and Sharing System: Emerging Issues and Challenges Mahendra Kumar Shrivas1,*, Ashok Bhansali1, Hoshang Kolivand2 and Kamal Kant Hiran3 Department of Computer Science Engineering, O. P. Jindal University, Raigarh, India Faculty of Engineering and Technology, Liverpool John Moores University, Liverpool, England 3 Faculty of IT and Design, Aalborg University, Copenhagen, Denmark 1 2
Abstract: The traditional technologies and digital systems of managing and maintaining data are inherently prone to manipulation at various levels. Ensuring the anonymity of the patient's identity, the safety of the medical records, and preventing the patient data from accidental and intended manipulations have been the industry's biggest challenges for decades. Failing to control the integrity of the Patient Health Records (PHRs) and Medical Health Records (MHRs) in the Healthcare Data Management System (HDMS)/ Healthcare Information System (HIS) may create challenges in identifying, diagnosing, and treating the disease and puts the patient at a greater risk. The frequency of healthcare data breaches, the magnitude of compromised records, and the financial impact rapidly increase with time. This chapter systematically and critically reviews the issues and challenges faced by various healthcare stakeholders in PHRs/MHRs-based HDMS/HIS systems. Blockchain powered patient health record and sharing schemes can be used to ensure the integrity and safety of healthcare data and share data among various healthcare ecosystem stakeholders using smart contracts to promote transparency, tamper-proofing, and consented access to data in distributed multi-stakeholder environment. This chapter highlights the need for post-quantum cryptography and recommendations for future improvements in blockchain-based patient health records and sharing system.
Keywords: Blockchain, Distributed Ledger, Healthcare Data Management System, Healthcare Information System, Medical Health Records, Patient Health Record, Post-Quantum Cryptography, Smart Contract. Corresponding author Mahendra Kumar Shrivas: Department of Computer Science Engineering, O. P. Jindal University, Raigarh, India; Email: [email protected] *
Mariya Ouaissa, Mariyam Ouaissa, Zakaria Boulouad, Inam Ullah Khan, Sailesh Iyer (Eds.) All rights reserved-© 2023 Bentham Science Publishers
Blockchain-Based Patient Health Record
Computational Intelligence for Data Analysis, Vol. 2 249
INTRODUCTION The demand for quality healthcare has grown significantly not only in urban as well as rural areas. Modern healthcare industries are driven by privatization, liberalization, and globalization. Most of the new hospitals in developing economies are driven by public-private partnerships. Today's healthcare industry revolves around data, and the data's integrity, privacy, and truthfulness are at the center of the industry's evolution, operation, and further growth. The adoption of technology in healthcare has been increasing significantly, and the healthcare industry is becoming more paperless to reduce costs and enhance their service quality and efficient management of healthcare-related data. The healthcare data management system started with paperless departmental planning records and medical items inventory, followed by patient billing, laboratory testing, diagnostic report, and pharmacy stock and billing management [1]. The most important task is in HDMS/HIS collection of PHR/MHR data responsible, authentic, reliable, and secure. In Section 1, the authors tried to represent the healthcare ecosystem, followed by the methodology adopted in Section 2. The chapter discusses HDMS/HIS and its evolutions followed by the issues and challenges these systems face in Section 3. Section 4 explains the fundamental concepts of Blockchain, which can be best fit to solve the problems in data collection and sharing of PHR/MHR records to various healthcare stakeholders in a secure, transparent, and real-time manner in a distributed and multi-stakeholder environment without the need for any trusted third party in section 4. Section 5 discusses various Blockchain-based models and highlights the issues and challenges of multiple models. The authors presented the overall points, challenges, and recommendations for future improvements in Blockchainbased patient health record and sharing systems. METHODOLOGY In this chapter, the authors adopted a systematic review-based qualitative research methodology. The authors followed phenomenology and explanatory research to study conceptual terms of healthcare, PHR/MHR, Blockchain, and DMS/HISrelated phenomena from the available literature. We used analysis and synthesis techniques to study various healthcare information systems and narrowed our focus area to Patient Health Records (PHRs) and Sharing Systems in HDMS/HIS, then followed successive refinement and implementation methodology to lay down the foundation of future enhancements of the Blockchain-based patient health record and sharing systems by highlighting the issues and challenges in the past and current such systems.
250 Computational Intelligence for Data Analysis, Vol. 2
Shrivas et al.
THE HEALTHCARE DATA MANAGEMENT SYSTEM (HDMS) OR HEALTHCARE INFORMATION SYSTEM (HIS) The Healthcare Data Management System (HDMS) or Healthcare Information System (HIS) is useful in systematically managing records, ease of access and usage. It provides fast and easy information retrieval, fiscal control and financial planning, inventory management, efficient process management, human resource management, and patient clinical and diagnosis history management to enhance the quality of service and achieve healthcare governance excellence [2]. Evolution of the Health Data Management System (HDMS)/ Health Information System (HIS) The HDMS/HIS has seen a significant transformation over the last few decades from a paper-based record maintenance approach to computerized maintenance of records. The web-based systems have given liberty to access various features of HDMS/HIS using multiple offices over the Internet. While Cloud Computing has offered the Infrastructure as a Service (IaaS) model to reduce the ownership of IT infrastructures, and Software as a Service (SaaS) has given a standard version of HDMS/HIS at a low cost [3, 4]. Using Internet of Things (IoT) devices, real-time health data is collected, and patient health is monitored. In contrast, complex clinical trials, the effectiveness of medicines, diseases spread over a geographical area, mutation patterns, etc., can be analyzed and forecasted using big data analytics. The Blockchain addresses the issues of trust, security, data tempering, secure communication, transparency, privacy issues, well informed and authorized use of medical/health data using consensus. The evolution of the Health Data Management System (HDMS)/ Health Information System (HIS), along with timelines, is given in Table 1 [5]. The vast processing power and resources are needed to perform big data analytics. Supercomputers and highly reliable Cloud Computers, and it is expected that in the future, HDMS/HIS systems, will be using Blockchain on distributed Quantum networks and powerful Quantum Computers as nodes in Blockchain networks. Table 1. Technical Evolution of the HDMS/HIS. S. No.
HDMS/HIS Approach
Timelines
1
Paper Based
1793
2
Standalone Computer Based
1960
3
Client-Server Web Based
1990
4
Cloud Computing Based
2012
5
Internet of Things (IoT) Based
2012
Blockchain-Based Patient Health Record
Computational Intelligence for Data Analysis, Vol. 2 251
(Table 1) cont.....
S. No.
HDMS/HIS Approach
Timelines
6
Big Data Based
2015
7
Blockchain-Based
2017
8
Blockchain Quantum Computing Based
In Future
Current Status, Issues, and Challenges As far as patient medical care, prescription and administration of drugs are concerned, most hospitals still use paper-based file and folder systems. In the case of transfer patients, doctors/hospitals exchange information about the patient's medical conditions. They also need detailed information about the drugs given to the patient during the treatment. The doctors and hospitals share this information in paper or images via email or file-sharing messengers. Huge Data Volume and Velocity and Paper-Based Record Keeping The Out-Patient Department (OPD) generates massive records set daily. Recording these patient medical data becomes time-consuming, and the same is true in the In-Patient Department (IPD) case due to the vast volume and velocity of data generated. Patient medical record management is still paper-based, and this is a major issue in medical data management. Most hospitals do not maintain OPD patient medical records, and the medical file/folder and the history files are given back to the patient. They retain medical records of IPD patients only in paper form and rarely enter those records into the system by themselves because it takes extra-human time with cost. Only a few patient data are recorded in case of claim settlement or reimbursement by insurance companies. Some patient data are also recorded from paper-based records into the healthcare record management system as required by accreditation bodies or other international institutions for further medical research and clinical trials. Interoperability and Data Sharing In most developing countries, various hospitals and clinics use different software systems based on availability and budget. Most hospitals stop using healthcare management systems over time due to the lack of local technical support from vendors, high costs and low-quality software systems. Adopting a multi-vendor health data management system is another challenge, as interoperability and data sharing among various systems require additional data formatting efforts. Country-wide or Worldwide standardization of process flow
252 Computational Intelligence for Data Analysis, Vol. 2
Shrivas et al.
and medical data set is needed. There are standards for some diseases, but local software vendors do not implement them, or record entry is not done correctly. Due to this, substantial clinical records are set lying down in paper format, which is exposed to fire, takes more space, and has other issues. The Healthcare providers, Life Science Researchers, Pharmaceutical companies, Patients, Doctors, and other stakeholders do not communicate in terms of realtime data sharing in the Healthcare Ecosystem. Interoperability among HDMS/HIS is a significant issue as they do not interact with each other or interact with a minimal scope at a low level with time-consuming technical support. For the past few decades, Healthcare ecosystems have faced various issues due to this and the biggest hurdle in innovation and research as far as patient health and public health are concerned. Data Governance, Manipulation, Privacy and Security Threats Lack of good hospital governance and administration also contributes to the nonavailability of clinical data in electronic forms regardless of the availability of good and standard healthcare record management systems in hospitals. Some hospitals or healthcare professionals are involved in malpractices to hide procedures/human mistakes or to support high-profile criminals. Patient medical records are sometimes counterfeited, misplaced, or destroyed to escape from law enforcement agencies. Generally, HDMS/HIS store all the data in a central database, either in the Cloud or on-premises, which is within reach of attackers (insiders/outsiders). Once the attacker gains access to the database, they can easily steal/modify healthcare data [6, 7]. Privacy is a significant issue in HDMS/HIS. As data belongs to the patient, the patient should give their consent to use it for any purpose. The PHR data is susceptible in nature, and data leaks could be dangerous and life-threatening for any patient. Looking at the various security breaches and data hacks in multiple countries and hospitals, it can be concluded that current HDMSs/HISs are not secure and have become obsolete [8]. BLOCKCHAIN FUNDAMENTALS, CONCEPTS, AND FEATURES Satoshi Nakamoto proposed and implemented the Blockchain in 2008-09 to implement a crypto-currency platform called Bitcoin. The Blockchain is the most reliable and robust technology of the current time to implement a transparent multi-party system in a centralized, distributed, or hybrid environment. The
Blockchain-Based Patient Health Record
Computational Intelligence for Data Analysis, Vol. 2 253
forged entry in Blockchain is almost impossible because the block ownership can be traced and verified. Without the consensus of all participants, entry into the Blockchain is not allowed. The recorded transactional blocks in the Blockchain cannot be removed or altered without the consensus of all participants or 51% of the majority in general [9]. The Block structure can be created as per need and specification on different Blockchain Platforms as per the need. A transaction is digitally signed using a private key. Nodes verify the digital signature using the public key. If the digital signature is confirmed against the private key using the public key, then a transaction is known to have originated from a genuine wallet address [10]. A transaction is digitally signed using a private key. Nodes verify the digital signature using a public key. If the digital signature is confirmed against a private key using a public key, then a transaction is known to have originated from a genuine wallet address [11]. And then, nodes try to verify the ownership of the digital asset to avoid double-spending, then nodes check the Block's authenticity by generating a block hash value using an approved hash function and verifying it against the hash value stored inside the Block. If the value matches, it is a genuine block; otherwise, it is fake and dropped. Upon successful ownership and block hash verification, the verifying node broadcasts proof of this Block to nodes in the Blockchain network, and after a fixed number of confirmations, the verified Block is added to the local copy Blockchain by all nodes using consensus protocol. This process of generating a new block is called mining, and the miner node gets rewards in terms of new Crypto-Coin or transaction fees [12]. Blockchain Categorization The Blockchain can be categorized into various types based on data accessibility (Public, Private, Consortium, and Hybrid), authorization model (Permissioned, Permissionless, and Hybrid), and Smart Contract support (Stateless, Stateful) [10]. Evolution of Blockchain Technology Blockchain 1.0 has seen the usage of crypto/digital currency. While using Blockchain 2.0, we can implement smart contracts in multi-party and distributed environments [13]. Blockchain 3.0 focuses on applications and use-cases of Blockchain technologies to solve problems in various areas like good governance, voting, literacy, health, science, art, culture, etc. [14]. While Blockchain 4.0 is dedicated to solving real and complex problems of Industry 4.0 and making Blockchain usable for solving real-life problems of governments and enterprises.
254 Computational Intelligence for Data Analysis, Vol. 2
Shrivas et al.
Blockchain 5.0 focuses on Non-Fungible Tokens and Metaverse, while Blockchain 6.0 will be primarily based on Quantum Blockchain, where real Blockchain distributed applications (DApps) would be running on Quantum Computers powered with post-quantum public key and hashing algorithms. Transaction in Blockchain Network In Fig. (1) [11] high-level view of transactions in the Blockchain network is depicted. Someone initiates a transaction through their wallet application by sending the transaction to the nearest node. The node broadcasts it to a decentralized P2P Blockchain network. After successful verification, the miner generates one new Block, validated and approved by participating nodes through a consensus process. Upon approval, the newly generated Block updates the previous block hash value and is then added to the distributed Blockchain ledger. Someone initiates transactions
Verify Transactions
$ $
00101 11001
Transaction
$ The requested transaction is broadcast to a decentralized P2P network consisting several nodes
Miner
A new block created
Block New block added to the existing chain
Consensus
The block is validated and approved by reaching consensus by all the nodes in the network
Fig. (1). High-level view of transaction in blockchain network.
In the whole transaction process, reliability and security are maintained in the following manner: 1. Submission of transaction or origination of data-source from trusted and authentic users. A private key is issued to the user to sign messages digitally. Nodes verify the digital signature of the messages against the public key to ensure a trusted party initiates the transaction. 2. Inter-Node communication is done through remote procedure calls in peer-topeer distributed Blockchain networks. The communication channel is protected using a secure socket layer to ensure end-to-end encryption.
Blockchain-Based Patient Health Record
Computational Intelligence for Data Analysis, Vol. 2 255
3. Each block is timestamped and stores current and previous block hash values in the distributed Blockchain. Any changes to any block would result in a different block hash value that can be tracked. To modify one Block, one has to change all the blocks that come after the changed blocks and make the same changes in the Blockchain of all participating nodes on Blockchain networks. It is almost impossible to do so in secret without reaching a consensus, and the cost of doing so would be very costly as it requests vast computing resources. HEALTHCARE AND BLOCKCHAIN The complete Healthcare ecosystems face severe challenges due to security, privacy, interoperability, data manipulation, etc. Due to the inability to securely share data and isolated management of the Electronic Medical Record (EMR) and the Patient Health Record (PHR), healthcare professionals do not track the complete history of medical records, which leads to improper treatment. These challenges are not only blocking research and innovation in the area of health science but also a severe threat to the health of the patients and the general public. Validation of Clinical trials and the effectiveness of the new drugs or vaccines cannot be predicted due to the lack of sufficient data in the long term. Pharmaceutical companies have been facing a severe issue of counterfeit drugs in their supply chain for a very long time [15]. Blockchain technology offers immense distributed processing capabilities and storage resources, providing scalable, secure, transparent solutions without compromising the privacy and security of healthcare data. Blockchain promotes transparency among participant stakeholders and promotes interoperability among diverse healthcare systems. Blockchain can be implemented to reduce the complexities of healthcare data management and data sharing among various healthcare stakeholders while taking consent from patients and respecting their privacy. Blockchain technology offers much potential for integration, innovation, interoperability, and sustainability in healthcare. Blockchain-Based Systems Models for the HDMS/HIS Pandi-Perumal et al., in their research, proposed a Blockchain model to record, retrieve and share PHR to administer sleeping medicines [16]. They focused on the direct delivery of sleep medicines from pharmaceutical companies to doctors, nurses, or patients based on electronic prescriptions shared by doctors on the Blockchain platform. To achieve this, they proposed the following modules in their Blockchain-based System, as depicted in Fig. (2) [16].
256 Computational Intelligence for Data Analysis, Vol. 2
Shrivas et al.
Step 2:The patients data is collected, encrypted and given an ID that is stored on the patients blockchain.This data is sent to the cloud storage.
Step 1: Sleep data generated
from patients wearables, polysomnography, sleep laboratory reports, doctors notes and medicine prescriptions
data
Encrypted storage
encryption
Data Encryption
Decryption
Step 4:The patients data is decrypted and displayed on the doctors/somnologists device or application for further managmenmt
Step 3:The patients data is
requested and the ID on the patients blockchain is used to retrieve the encrypted data from cloud storage.
Hyperledger
Fig. (2). The Blockchain-based approach to administering sleep medicine.
1. Data Collection: The patient's sleep behaviour data is collected from their wearable IoT devices. If an improper sleeping pattern is detected in Polysomnography, the certified laboratory report is collected per the doctor's recommendations. The doctors prescribed the medicine to the patient. The doctor's notes and the prescription are recorded in the Blockchain. 2. Secure Data Storage: After data collection from various stakeholders, this data block is allocated relevant, unique ID and encrypted using security keys. After encryption, the ID and data block hash are recorded into the Blockchain, and the encrypted data block is stored in the Cloud. 3. Data Request: Only ID is shared with other healthcare providers, pharmaceutical companies, and healthcare professionals. This ID can retrieve data block hash from the Blockchain, and then encrypted data is downloaded from the cloud storage. 4. Verification, Permission, and Decryption: To download data from the Cloud storage, patient permission is mandatory, and after downloading encrypted data, a block hash is generated and verified against the data block hash stored in the Blockchain. The patient's private key is needed to decrypt the data. This decrypted data is then represented to health professionals for future examination. Przytarski et al. proposed a trustworthy blueprint of a health record platform using IoT and Blockchain called SEAL. They primarily focused on the Authenticity, Confidentiality, and Integrity of data in the SEAL platform. The whole platform
Blockchain-Based Patient Health Record
Computational Intelligence for Data Analysis, Vol. 2 257
was divided into three parts: data producer, Blockchain platform layer, and data consumer layer. Using IoT devices, patient records are stored. It uses a private and public key in the lightweight attribute-based authentication mechanism to verify devices' authenticity and encrypt PHR data. The blockchain platform layer has provisions to store data, patient id, block hash, and encrypted data that anyone can search and verify. Access control, key management, and filtering are done in this layer. Data consumers such as physicians, researchers, and insurance agents can request data upon data retrieval request authentication. While data filtering and decryption occur before data delivery, as depicted in Fig. (3) [17].
Fig. (3). Data acquisition, management, provision, and verification via SEAL.
Banotra et al., in their research work, proposed layered architecture to use IoT, Cloud Computing, and Blockchain-based system to store and share PHR/MHR [18]. Their proposed model contains the following layers: 1. Physical Layer (Patient) 2. Device Layer (Mobiles, Wearables, Medical Devices) 3. Communication Layer (HTTPS/TCP-IP, API, GATEWAY (Lora, ZigBee, Profibus)) 4. Could Interface (Cloud Hub with MQTT/COAP support) 5. Platform Components (Low Level - Real-time data processing and Storage Engine, Blockchain Engine, Middle Level – API Management Layer, Government Healthcare Data System and HMS/HIS, Top-Level – Mobile/Web Application)
258 Computational Intelligence for Data Analysis, Vol. 2
Shrivas et al.
6. Stakeholder Layer (Governments, Regulators, Hospitals, Doctors, and Patients) PATIENTORY1 (Atlanta, Georgia) and MEDICALCHAIN2 (London, England) are two well-implemented Blockchain-based HDMS/HIS real-world products. The processes of collecting, processing, storing, retrieving, and completing/filtering data sharing among various stakeholders of healthcare ecosystems were similar, as discussed in earlier case studies, while using an APIbased data access model is a new addition. They also included multiple regulators and governments in their proposed model as stakeholders. How Does Blockchain Address Security, Consensus, and Data Manipulation Issues? Each wallet (client application) is secured using a private key and the passcode known to the wallet user. Guessing a private key using a brute-force attack is not feasible because of many key pools. In the case of a custodial wallet provider, the private key is managed by the wallet provider themselves, but then also no one can use their wallet without a passcode. However, in most cases, the attackers were found stealing individual private keys and passcode to steal information from the Blockchain client application. Transactions in Blockchain are highly secure because the communication channel is encrypted using public and private key cryptography. A participant in Blockchain is called a node, and each note digitally signs the transaction using its private key. After signing, the Block is broadcasted to various peers for confirmation, and based on consensus; the generated Block is verified and added to the distributed ledger. The fact is that no authentic party can sign these transactions without an approved private key verified using a public key, and any manipulation in a block during propagation would result in failure during the decryption process. Hence communication between nodes in a peer-to-peer network is highly secure [19]. After receiving the broadcasted block participant node verify the transaction and regenerates the block hash values using hash functions like SHA-256, Keccak256, SHA3, SHAKE256, etc., and matches with hash values stored in the Block. If the value is matched, then it is a valid block; else, the Block is discarded, and if a majority of the nodes (generally 51% nodes) in the Blockchain network confirm the correctness of the Block, then this Block is added to the Blockchain. This block verification process and the consensus is called the mining process, and the scheme is called Proof of Work (PoW). In the PoW, hash verification is done more time, and computing is resource-consuming [20]. Various Blockchain platform uses different-different Consensus Algorithms (CA) [20] like Proof of
Blockchain-Based Patient Health Record
Computational Intelligence for Data Analysis, Vol. 2 259
Stake (PoS), Proof of Importance (PoI), Byzantine Fault Tolerance (BFT), and Proof of Elapsed Time (PoET) [13]. In Blockchain, each block stores the hash value of its Block and the previous block hash. Any changes to any block generate a different hash result. The Blockchain network can easily track these manipulated blocks, and manipulated blocks get discarded. To change any block, one has to change the current block hash and all hash values of all forward blocks. An attacker could change the block value of any single node, but to reflect changes, the attacker has to change the Blockchain of more than 51% of nodes, which is almost impossible to do without being caught. As Blockchain is distributed, each node maintains a separate copy of Blockchain. It is also tough to do so because it requires a massive amount of computing resources, and the cost of doing so would be huge. How Does Blockchain Address Privacy Issues? The Blockchain supports anonymity and eliminates the need for a trusted third party, but anonymous transactions could be troublesome if this technology is used for wrongdoing. This is the biggest concern of various law enforcement authorities, but there are ways to address it without compromising anonymity using Blockchain technology. Without revealing the private key, a wallet application only exposes the wallet address, not the private key or any personal identity of the wallet holder, to protect the privacy of wallet owners. The wallet application providers can facilitate the Know Your Customer (KYC) process and issue digital identities in the form of wallet addresses without compromising owners' privacy. The digital identity can only be revealed to law enforcement agencies when needed to avoid misuse of Blockchain Technology. Preventing PHR, EMR Manipulation, and Sharing Records Securely using Blockchain In Blockchain-based PHR/EMR systems transaction process, reliability and security are maintained at the following level: 1. Submission of transaction or origination of data-source from trusted and authentic uses. A private key is issued to the user to sign messages digitally. Nodes verify the digital signature of the messages against the public key to ensure a trusted party initiates the transaction. 2. Inter-Node communication is done through remote procedure calls in peer-topeer distributed Blockchain networks. The communication channel is protected using a secure socket layer to ensure end-to-end encryption.
260 Computational Intelligence for Data Analysis, Vol. 2
Shrivas et al.
3. Each block is timestamped and stores current and previous block hash values in the distributed Blockchain. Any changes to any block would result in a different block hash value that can be tracked. To modify one Block, one has to change all the blocks that come after the changed blocks and make the same changes in the Blockchain of all participating nodes on Blockchain networks. It is almost impossible to do so in secret without reaching a consensus, and the cost of doing so would be very costly as it requires enormous computing resources. ISSUES, CHALLENGES, AND RECOMMENDATIONS The researchers proposed a systemic approach using Blockchain to collect, store, retrieve, and share PHR and MHR data. However, there are various issues and challenges, as listed below: 1. Real-time data collection is not possible in many cases due to networks, faulty devices, and power issues. 2. If security keys are compromised by the patient or custodial key service provider or key repositories, the attackers could alter the Blockchain data and change data blocks of PHR/MHR in the ledger. 3. To decrypt encrypted PHR/MHR records, the patient has to share their private key, which can be stolen if the health professional uses a compromised system. 4. The Pharmaceutical companies would also need the patient's private key to decrypt the patient's data, so they could also create a local store of private keys and PHR/MHR records and misuse it. 5. There is no need to share the complete PHR/MHR with all stakeholders. 6. Blockchain offers better transparency, trust, tamper-proofing, consensus, and security, but it has another set of challenges and issues [21] which can be addressed by adopting the recommended best practices of Shrivas et al. [13]. 7. Due to the invention of Quantum Computing, the entire Blockchain can be compromised, as Quantum computers are much faster than any known supercomputers of the current time. Most Blockchains and other systems, along with HDMS/HIS, use RSA-based private-public keys that come under prequantum cryptography and can easily be broken with the help of Quantum Computers [22, 23]. The patient should have complete ownership of the private key, and there could be intelligent contracts to share selected, not all, data among healthcare providers and pharmaceutical companies. The Patients should have the option to securely
Blockchain-Based Patient Health Record
Computational Intelligence for Data Analysis, Vol. 2 261
share complete PHR/MHR records with their doctors but share limited data with other healthcare ecosystem stakeholders. Qi Xia et al., in their research paper, proposed a better model for data sharing using Smart Contracts [24]. Anyone in the healthcare ecosystem wanted to access data each time data. The patient should give their consent and get access to the nonfiction of data. The data should be shared with various parties in read-only mode. After taking permission and then sharing PHR/MHR each time must be mandatory and the data must be well protected. Post-quantum cryptography should be adopted to build such systems to ensure higher protection. If any breach happens, there must be disclosure and intimations to each stakeholder in a data breach incident. There should be a standard working procedure for IT and Healthcare service providers, and their legal obligation. There should be provision to record data in manual paper format to address realtime data collection in case of a faulty device, power, and network issues, which can be later entered into the system by either patient or any health professional or just scanning the paper and storing the record in image or pdf format in a document data store. The future HDMS/HIS should address the above recommendations to store and share PHR/MHR with various stakeholders in healthcare ecosystems. We strongly recommend that the suggested changes should be incorporated in current and future system models to increase the adoption, and the same can be implemented not only to administer sleep medicines but also can be generalized to administer various other medications to treat patients like blood pressure, sugar, HIV, malaria, cancer, etc. Future systems should use distributed Blockchain with postquantum cryptography and a well-defined, more efficient consensus mechanism. CONCLUSION This work details a review of the issues and challenges faced by various healthcare stakeholders in PHRs/MHRs-based HDMS/HIS systems. Blockchainpowered patient health record and sharing schemes can be used to ensure the integrity and safety of healthcare data and share data among various healthcare ecosystem stakeholders using smart contracts to promote transparency, tamperproofing, and consented access to data in distributed multi-stakeholder environment. This chapter highlighted the need for post-quantum cryptography and recommendations for future improvements in Blockchain-based patient health records and sharing systems.
262 Computational Intelligence for Data Analysis, Vol. 2
Shrivas et al.
CONSENT FOR PUBLICATON Declared none. CONFLICT OF INTEREST The author declares no conflict of interest, financial or otherwise. ACKNOWLEDGEMENT Declared none. REFERENCES [1]
R.B. Korgaonkar, "Adoption of Information System by Indian Hospitals; Challenges and Roadmap", Int. J. Sci. Eng. Res., vol. 5, no. 2, pp. 473-476, 2014.
[2]
R. Batra, and A.S. Pall, "Adoption and Assessment of Hospital Information Systems: A Study of Hospitals in Jalandhar", Asia-Pac. J. Manag. Res. Innov., vol. 11, no. 3, pp. 205-218, 2015. [http://dx.doi.org/10.1177/2319510X15588383]
[3]
B.A. Ampofo, M.K. Shrivas, and S.V. Singh, "Public Cloud Computing; Benefits and Risks to Users and Potential Users", International Journal of Management & Scientific Research, vol. 1, no. 1, pp. 15, 2015.
[4]
K.K. Hiran, A. Henten, M.K. Shrivas, and R. Doshi, "Hybrid EduCloud Model in Higher Education: The case of Sub-Saharan Africa, Ethiopia", In: 2018 IEEE 7th International Conference on Adaptive Science & Technology (ICAST), Accra, Ghana, 2018. [http://dx.doi.org/10.1109/ICASTECH.2018.8507113]
[5]
L. Ismail, H. Materwala, A.P. Karduck, and A. Adem, "Requirements of Health Data Management Systems for Biomedical Care and Research: Scoping Review", J. Med. Internet Res., vol. 22, no. 7, p. e17508, 2020. [http://dx.doi.org/10.2196/17508] [PMID: 32348265]
[6]
M.K. Shrivas, A. Amoako, S.O. Boateng, and T. Yeboah, "Migration Model for unsecured Database driven Software System to Secure System using Cryptography", International Journal of ICT and Management, vol. III, no. 2, pp. 1-8, 2015.
[7]
M.K.S. Mark Osei Gyabi, "Data Security in Rural Banking Sector: A Case Study in Ashanti Region", International Journal of Advanced Research in Computer Science & Technology, vol. 4, no. 2, pp. 99106, 2016.
[8]
A. Gatouillat, Y. Badr, B. Massot, and E. Sejdić, "Internet of Medical Things: A Review of Recent Contributions Dealing With Cyber-Physical Systems in Medicine", IEEE Internet Things J., vol. 5, no. 5, pp. 3810-3822, 2018. [http://dx.doi.org/10.1109/JIOT.2018.2849014]
[9]
M. K. Shrivas and D. T. Yeboah, "A Critical Review of Cryptocurrency Systems," Texila Int J. Acad. Res., Vol. 4, no. 2, pp. 116-131, 2017.
[10]
M. K. Shrivas and T. Yeboah, "The Disruptive Blockchain: Types, Platforms, and Application," Texila Int. J. Acad. Res., pp. 17-39, 2019.
[11]
A. A. Monrat, O. Schelén and K. Andersson, "A Survey of Blockchain From the Perspectives of Applications, Challenges, and Opportunities," IEEE Open Access J., pp. 117134-117151, 2019.
[12]
D. Yaga, P. Mell, N. Roby, and K. Scarfone, "Blockchain Technology Overview," National Institute of Standards and Technology Internal Report, 2019.
Blockchain-Based Patient Health Record
Computational Intelligence for Data Analysis, Vol. 2 263
[13]
M.K. Shrivas, T. Yeboah, and S.S. Brunda, "Hybrid Security Framework for Blockchain Platforms", 2020 First International Conference on Power, Control and Computing Technologies (ICPC2T). Raipur, India, 2020.
[14]
A.H. Cristina Devecchi, and A. Scott Turner, Armellini and I. Brooks, “The application of the blockchain technology to track, register, certify and enhance the utilization of acquired learning assets. CCEG Blockchain UN Lab, 2017, pp. 1-34.
[15]
S. Bittins, G. Kober, A. Margheri, M. Masi, A. Miladi, and V. Sassone, Healthcare Data Management by Using Blockchain Technology.Applications of Blockchain in Healthcare. Springer: New Delhi, India, 2021, pp. 1-28. [http://dx.doi.org/10.1007/978-981-15-9547-9_1]
[16]
S.R. Pandi-Perumal, S. Santaji, V.V. Desai, T. Sunil, and V.K. Chattu, "The Blockchain Never Sleeps: How Can Blockchain Technology Transform Sleep Medicine?", Sleep Vigil., vol. 5, no. 1, pp. 17-27, 2021. [http://dx.doi.org/10.1007/s41782-021-00132-w]
[17]
D. Przytarski, C. Stach, C. Gritti, and B. Mitschang, "A Blueprint for a Trustworthy Health Data Platform Encompassing IoT and Blockchain Technologies", 29th International Conference on Software Engineering and Data Engineering (SEDE’ 20). Las Vegas, NV, United States, 2020.
[18]
A. Banotra, J.S. Sharma, S. Gupta, S.K. Gupta, and M. Rashid, Use of Blockchain and Internet of Things for Securing Data in Healthcare Systems. Multimedia Security, Algorithms for Intelligent Systems, 2021, pp. 255-267. [http://dx.doi.org/10.1007/978-981-15-8711-5_13]
[19]
M. Andrychowicz, S. Dziembowski, D. Malinowski, and Ł. Mazurek, "On the malleability of bitcoin transactions", International Conference on Financial Cryptography and Data Security. Berlin, Heidelberg, 2015. [http://dx.doi.org/10.1007/978-3-662-48051-9_1]
[20]
M. Hölbl, M. Kompara, A. Kamišalić, and L. Nemec Zlatolas, "A Systematic Review of the Use of Blockchain in Healthcare", Symmetry (Basel), vol. 10, no. 10, p. 470, 2018. [http://dx.doi.org/10.3390/sym10100470]
[21]
M.K. Shrivas, T. Yeboah, and S.S. Brunda, "The Disruptive Blockchain Security Threats and Threat Categorization", 2020 First International Conference on Power, Control and Computing Technologies (ICPC2T). Raipur, India, 2020. [http://dx.doi.org/10.1109/ICPC2T48082.2020.9071475]
[22]
S. Gupta, K.K. Gupta, P.K. Shukla, and M.K. Shrivas, "Blockchain-based Voting System Powered by Post-Quantum Cryptography (BBVSP-PQC", 2022 Second International Conference on Power, Control and Computing Technologies (ICPC2T). Raipur, India, 2022. [http://dx.doi.org/10.1109/ICPC2T53885.2022.9776966]
[23]
T. M. Fernández-Caramès and P. Fraga-Lamas, "Towards Post-Quantum Blockchain: A Review on Blockchain Cryptography Resistant to Quantum Computing Attacks," Emerging Approaches to Cyber Security, vol. Volume: 8, pp. 2169-3536, 2020.
[24]
"Q. Xia, E.B. Sifah, K.O. Asamoah, J. Gao, X. Du, and M. Guizani, "MeDShare: Trust-Less Medical Data Sharing Among Cloud Service Providers via Blockchain", IEEE Access, vol. 5, no. 2017, pp. 14757-14767, 2017.
264
Computational Intelligence for Data Analysis, 2023, Vol. 2, 264-268
SUBJECT INDEX A Artificial 107, 108, 110, 111, 114, 141, 142, 146, 166, 168, 182, 183, 184, 187, 215, 231 internet of things (AIoT) 107, 108, 110, 111, 114 neural networks (ANN) 141, 142, 146, 166, 168, 182, 183, 184, 187, 215, 231
B Bayesian 127, 130, 214, 217 ELM (BELM) 127, 130, 214, 217 method 130 network 127 Beacon delivery ratio (BDR) 71 Bilinear interpolation method 125 Binary conversion 131 Blockchain 250, 253, 254, 255, 258, 259, 260 networks 250, 253, 254, 255, 258, 259, 260 platform 253, 255, 258 Bluetooth technology 69, 74 Bone fractures 120, 122, 123, 124, 126, 127, 128, 129, 132, 133 femur 124 tibia 129 Brain 143, 210, 211, 212, 215, 216, 218, 219, 220, 221, 222 computer interface (BCIs) 210, 211, 212, 215, 216, 218, 219, 220, 221, 222 diseases 143 Byzantine fault tolerance (BFT) 259
C Cameras, visual 207 Cardiac hypertrophy 183 Cardiovascular 177, 185 disorders 177 heart diseases 185
Chronic inflammation 162 Cirrhosis 143, 144, 163 Cloud 11, 12, 19, 23, 56 based temperature monitoring technology 23 server 56 services 12, 19, 56 storage technology 11 CMOS technology 77 Cognitive 22, 114 IoMT (CIoMT) 22 machine learning 114 Computed tomography (CT) 37, 85, 120, 133, 144, 186, 187 Conditions 44, 75, 143, 154, 156, 160, 167, 178, 180, 183, 185 arrhythmic 185 inflammatory 178 Consensus algorithms (CA) 258 Convolution neuronal networks 185 Coronary infarction 186 Coronavirus disease 11 COVID-19 pandemic 11, 22, 23, 25, 83, 91, 96, 97, 110 Cryptography 43, 248, 260, 261 post-quantum 248, 261 pre-quantum 260
D Data augmentation techniques 245 Data mining 162, 163 method 162 techniques 163 Decryption process 258 Deep convolutional 204, 215, 217 generative adversarial networks 215, 217 neural network 204 Deep learning 158, 168 based transfer learning techniques 168 technique 158 Deep neural network (DNN) 85, 144
Mariya Ouaissa, Mariyam Ouaissa, Zakaria Boulouad, Inam Ullah Khan, Sailesh Iyer (Eds.) All rights reserved-© 2023 Bentham Science Publishers
Subject Index
Dengue fever 150, 166, 167 Detection 1, 131 of osteoporosis 131 system, intrusion 1 Device(s) 4, 5, 17, 18, 19, 20, 21, 41, 42, 43, 44, 65, 69, 70, 85, 87, 90, 108, 189, 195, 196, 223 brain-interfacing 223 electronic aid 195, 196 fabrication 43 mobile 4, 5, 90, 108 sensor 19 Diabetes disease 160, 161 Diabetic retinopathy 194 Disaster aid network 69 Discrete wavelet transform (DWT) 216, 234 Diseases 107, 137, 140, 147, 154, 186, 194, 211 coronary artery 154 eye 194 heart-related 140 infectious 107 life-threatening 137, 140 neuroanatomical 211 pulmonary 186 viral 147 Disorders 13, 48, 167, 169, 177, 212, 217 chronic brain 217 inflammatory 167 lung 169 motor system 212 Drone technology 91 Dysembroplastic neuroepithelial tumor (DNT) 240
E Echocardiography 186, 188 Electrocardiogram 52, 94, 181, 188, 231 automatic 231 Electrocardiography 187 Electroencephalography 210 Electronic medical record (EMR) 188, 255 Electrooculography 212
Computational Intelligence for Data Analysis, Vol. 2 265
Emergency 5, 72, 77 access period (EAP) 72, 77 response sensors 5 Energy consumption 60, 71, 77 Ensemble empirical mode decomposition (EEMD) 217, 218 Eye tracking 212
F FAWT technique 214 Flexible analytic wallet transform (FAWT) 214 Fuzzy logic system (FLSs) 241
G GDPR principles 40 Genetic 73, 88, 164, 180, 222 algorithms 73, 88, 164, 180 sequencing technology 222 Geographic information system (GIS) 91
H Healthcare 12, 14, 15, 54, 64, 87, 92, 112, 248, 249, 250, 251 Information System 248, 249, 250 management systems 251 monitoring 12, 87 services 14, 15, 54, 92 technologies 64, 112 Healthcare data 33, 49, 55, 57, 248, 249, 250, 255, 261 machines 55 management 248, 249, 250, 255 system (HDMS) 248, 249, 250 Heart diseases 4, 32, 149, 154, 155, 156, 180, 181, 183, 184, 189 chronic 156 coronary 154, 184 Hepatocellular carcinoma 144 Hough transform technique 128, 132 Hybrid lean-agile technique 54
266 Computational Intelligence for Data Analysis, Vol. 2
I Image 128, 129, 244 fusion process 244 processing techniques 128, 129 segmentation techniques 128 Image analysis 185, 187 learning-based texture 185 Imaging 88, 93 processes 88 scans 93 Intelligence-based techniques 2 IoMT 13, 17, 18, 21, 38, 41, 42, 43, 49, 50, 53, 60 architecture 49, 50, 60 devices 13, 17, 18, 21, 38, 41, 42, 43 utility services 53 IoMT-based 17, 18, 22 devices 17 methods for detecting infections 22 systems 18 IoMT cloud 54, 57 applications 57 architecture 54 IoMT edge cloud 57, 58, 59 applications 59 architecture 57 technologies 58 IoT 19, 49, 109 key technologies 49 networks 19 sensors 109 IoT-based 5, 93 intelligent diagnostic system 93 smart medical system 5
L Liver 143, 144, 149, 163 cancer 143, 144, 149, 163 disorders and cirrhosis 163 Liver disease 143, 144, 162, 163, 165 chronic 165 detection 143, 144, 162, 163
Ouaissa et al.
Logistic regression (LR) 4, 113, 132, 138, 139, 144, 146, 161, 165, 179, 182, 184 method 179 Low-energy security network method 52 Lung 142, 169 diseases detection 142 sickness 169
M Machine 1, 2, 7, 32, 88, 140, 149, 150, 153, 160, 185, 187 anaesthetic 32 Machine intelligence 1 applications 1 techniques 1 Machine learning 1, 2, 3, 4, 7, 83, 85, 95, 96, 97, 99, 107, 112, 113, 115, 137, 138, 139, 140, 143, 146, 149, 151, 153, 163, 164, 167, 169, 180, 184, 229 algorithms 7, 112, 113, 137, 138, 139, 140, 143, 146, 164, 167, 184 classification techniques 149 techniques 1, 2, 4, 7, 96, 97, 112, 115, 138, 140, 151, 153, 163 Magnetic resonance imaging (MRI) 37, 120, 122, 133, 144, 181, 186, 230, 233 Mammography 122, 156 MATLAB software 213 Medicine, cardiovascular 177, 181, 188, 189 ML-based technologies 98 Mobile 4, 64 edge computing (MEC) 64 technology 4 Multi-factor authentication (MFA) 41 Multilayer perceptron neural network (MLPNN) 216, 217, 218 Myocyte loss 183
N Natural 56, 93, 146 language processing (NLP) 56, 93, 146
Subject Index
Neural 143, 144, 146, 150, 151, 154, 157, 162, 183, 215, 216, 217, 230, 231 networks (NNs) 143, 144, 146, 150, 151, 154, 157, 162, 215, 216, 217, 230, 231 hormonal compensatory responses 183 Neural network 94, 180, 230 architecture 230 integration 180 migration 94 Neurons 143, 183, 222, 231, 238 artificial 231
O Obstacle detection techniques 196 Osteopenia disease risk 131 Osteoporosis 126, 131
P Phyllodes tumors 156 Polysomnography 256 Power 18, 66, 203, 222 bank device 203 consumption 18, 222 devices 66 Primitive neuroectodermal tumors (PNET) 230, 239 Principal component analysis (PCA) 215 PSO technique 183 Pulse wave velocity (PWV) 21
R Radio frequency identification (RFID) 12, 15, 38, 50, 51, 52, 53 Radiological images 85 Random 72, 77, 161, 162 access period (RAP) 72, 77 forest classifier (RFC) 161, 162 Reinforcement learning (RL) 113, 150, 152 Remote health monitoring (RHM) 12, 13 RF algorithms 217
Computational Intelligence for Data Analysis, Vol. 2 267
S Sequential pattern mining (SPM) 98 Skin 21, 132, 142 cancer 132, 142 condition monitoring systems 21 disease detection 142 Smart machine learning-based 146 Sparse bayesian ELM (SBELM) 214, 217 Supervised learning (SL) 138, 139, 150, 151, 152, 155, 156, 157, 161, 165, 166, 167, 170, 183, 184 techniques 155, 183 Supervised machine learning 138 Support vector machine (SVM) 128, 129, 130, 132, 145, 157, 163, 166, 180, 183, 184, 185, 214, 215, 217 Surgeons, neurological 230 Sustained data transmission speeds 70
T Techniques 65, 71, 113, 126, 133, 152, 155, 156, 157, 158, 160, 165, 169, 206, 229, 230, 231, 245 algorithmic 158 auto-coding 229 imaging 133 Technologies 12, 16, 53, 100, 137 block-chain 16 computational 137 intelligent 100 semantic 53 sensing 12 Telecare monitoring system 21 Tele-health monitoring systems 110 Thermal imaging 89, 91 process 91 system 91 Thoracic echocardiography 188 Tonsillitis 90 Trabecular density technique 126 Tracking 52, 128 image-based 128
268 Computational Intelligence for Data Analysis, Vol. 2
waste 52 Transmission, video 71 Tumor(s) 138, 157, 158, 230, 234, 240 detection 157 dysembryoplastic neuroepithelial 230 metastatic bronchogenic carcinoma 234 pituitary 230, 240 Tuneable-Q wavelet transform (TQWT) 217, 218
U Ultrasonic sensors 194, 195, 196, 197, 199, 202, 203, 207 Ultrasonography 12
V Variables 126, 132, 156, 166, 184 anthropometric 126 Ventricular dysfunction 187 Virtual 23, 24, 54, 98, 230 reality (VR) 23, 24, 230 resources (VRM) 54 Virus pathogenicity 98 Vision 194, 219, 231 impairment problems 194 robotic 231 Voice recognition 153
W Wavelet packet decomposition (WPD) 215 WBAN(s) 64, 66, 67, 68, 73, 76, 79 architecture 67, 68 mitigation method 79 sensors 76 process and arrangement condition 66 technologies 64, 79 transmission intervals 73 WBAN applications 65, 68, 69, 71, 79 non-medical 71 WBASN, social network 74 Wearable 5, 12, 23, 69, 86, 90, 107, 109, 111
Ouaissa et al.
AI-IoT sensors 111 devices 5, 12, 86, 90, 107 device technologies 90 health monitoring system (WHMS) 69 systems 109 technology 23 Wi-fi wireless technologies 64 Wireless 6, 11, 15, 31, 38, 43, 49, 52, 64, 65, 66, 67, 69, 70, 71, 76, 107 body networks 65 communication technologies 6, 107 low-power 38 networking technology 70 networks (WSN) 15, 38, 49, 52, 64, 66, 67, 69, 70, 71, 76 sensor network 49, 52, 67 technology research 67 Wk-NN algorithm 213 WLAN 70, 76, 126 network 76 noise transmission 70 Women, post-menopausal 126
Z ZigBee 70, 76 device 70 interference and WiFi confirmation method 76 network 76