Trends and Innovations in Information Systems and Technologies: Volume 3 (Advances in Intelligent Systems and Computing, 1161) 303045696X, 9783030456962

Citation preview

Advances in Intelligent Systems and Computing 1161

Álvaro Rocha · Hojjat Adeli ·  Luís Paulo Reis ·  Sandra Costanzo · Irena Orovic ·  Fernando Moreira   Editors

Trends and Innovations in Information Systems and Technologies Volume 3

Advances in Intelligent Systems and Computing Volume 1161

Series Editor Janusz Kacprzyk, Systems Research Institute, Polish Academy of Sciences, Warsaw, Poland Advisory Editors Nikhil R. Pal, Indian Statistical Institute, Kolkata, India Rafael Bello Perez, Faculty of Mathematics, Physics and Computing, Universidad Central de Las Villas, Santa Clara, Cuba Emilio S. Corchado, University of Salamanca, Salamanca, Spain Hani Hagras, School of Computer Science and Electronic Engineering, University of Essex, Colchester, UK László T. Kóczy, Department of Automation, Széchenyi István University, Gyor, Hungary Vladik Kreinovich, Department of Computer Science, University of Texas at El Paso, El Paso, TX, USA Chin-Teng Lin, Department of Electrical Engineering, National Chiao Tung University, Hsinchu, Taiwan Jie Lu, Faculty of Engineering and Information Technology, University of Technology Sydney, Sydney, NSW, Australia Patricia Melin, Graduate Program of Computer Science, Tijuana Institute of Technology, Tijuana, Mexico Nadia Nedjah, Department of Electronics Engineering, University of Rio de Janeiro, Rio de Janeiro, Brazil Ngoc Thanh Nguyen , Faculty of Computer Science and Management, Wrocław University of Technology, Wrocław, Poland Jun Wang, Department of Mechanical and Automation Engineering, The Chinese University of Hong Kong, Shatin, Hong Kong

The series “Advances in Intelligent Systems and Computing” contains publications on theory, applications, and design methods of Intelligent Systems and Intelligent Computing. Virtually all disciplines such as engineering, natural sciences, computer and information science, ICT, economics, business, e-commerce, environment, healthcare, life science are covered. The list of topics spans all the areas of modern intelligent systems and computing such as: computational intelligence, soft computing including neural networks, fuzzy systems, evolutionary computing and the fusion of these paradigms, social intelligence, ambient intelligence, computational neuroscience, artificial life, virtual worlds and society, cognitive science and systems, Perception and Vision, DNA and immune based systems, self-organizing and adaptive systems, e-Learning and teaching, human-centered and human-centric computing, recommender systems, intelligent control, robotics and mechatronics including human-machine teaming, knowledge-based paradigms, learning paradigms, machine ethics, intelligent data analysis, knowledge management, intelligent agents, intelligent decision making and support, intelligent network security, trust management, interactive entertainment, Web intelligence and multimedia. The publications within “Advances in Intelligent Systems and Computing” are primarily proceedings of important conferences, symposia and congresses. They cover significant recent developments in the field, both of a foundational and applicable character. An important characteristic feature of the series is the short publication time and world-wide distribution. This permits a rapid and broad dissemination of research results. ** Indexing: The books of this series are submitted to ISI Proceedings, EI-Compendex, DBLP, SCOPUS, Google Scholar and Springerlink **

More information about this series at http://www.springer.com/series/11156

Álvaro Rocha Hojjat Adeli Luís Paulo Reis Sandra Costanzo Irena Orovic Fernando Moreira •

•

•

•

•

Editors

Trends and Innovations in Information Systems and Technologies Volume 3

123

Editors Álvaro Rocha Departamento de Engenharia Informática Universidade de Coimbra Coimbra, Portugal

Hojjat Adeli College of Engineering The Ohio State University Columbus, OH, USA

Luís Paulo Reis FEUP Universidade do Porto Porto, Portugal

Sandra Costanzo DIMES Università della Calabria Arcavacata, Italy

Irena Orovic Faculty of Electrical Engineering University of Montenegro Podgorica, Montenegro

Fernando Moreira Universidade Portucalense Porto, Portugal

ISSN 2194-5357 ISSN 2194-5365 (electronic) Advances in Intelligent Systems and Computing ISBN 978-3-030-45696-2 ISBN 978-3-030-45697-9 (eBook) https://doi.org/10.1007/978-3-030-45697-9 © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2020 This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, express or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. This Springer imprint is published by the registered company Springer Nature Switzerland AG The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland

Preface

This book contains a selection of papers accepted for presentation and discussion at the 2020 World Conference on Information Systems and Technologies (WorldCIST’20). This conference had the support of the IEEE Systems, Man, and Cybernetics Society (IEEE SMC), Iberian Association for Information Systems and Technologies/Associação Ibérica de Sistemas e Tecnologias de Informação (AISTI), Global Institute for IT Management (GIIM), University of Montengero, Mediterranean University and Faculty for Business in Tourism of Budva. It took place at Budva, Montenegro, during 7–10 April 2020. The World Conference on Information Systems and Technologies (WorldCIST) is a global forum for researchers and practitioners to present and discuss recent results and innovations, current trends, professional experiences and challenges of modern information systems and technologies research, technological development and applications. One of its main aims is to strengthen the drive towards a holistic symbiosis between academy, society and industry. WorldCIST’20 built on the successes of WorldCIST’13 held at Olhão, Algarve, Portugal; WorldCIST’14 held at Funchal, Madeira, Portugal; WorldCIST’15 held at São Miguel, Azores, Portugal; WorldCIST’16 held at Recife, Pernambuco, Brazil; WorldCIST’17 held at Porto Santo, Madeira, Portugal; WorldCIST’18 held at Naples, Italy and WorldCIST’19 which took place at La Toja, Spain. The program committee of WorldCIST’20 was composed of a multidisciplinary group of almost 300 experts and those who are intimately concerned with information systems and technologies. They have had the responsibility for evaluating, in a ‘blind review’ process, the papers received for each of the main themes proposed for the conference: (A) Information and Knowledge Management; (B) Organizational Models and Information Systems; (C) Software and Systems Modelling; (D) Software Systems, Architectures, Applications and Tools; (E) Multimedia Systems and Applications; (F) Computer Networks, Mobility and Pervasive Systems; (G) Intelligent and Decision Support Systems; (H) Big Data Analytics and Applications; (I) Human–Computer Interaction; (J) Ethics, Computers and Security; (K) Health Informatics; (L) Information Technologies in

v

vi

Preface

Education; (M) Information Technologies in Radiocommunications; (N) Technologies for Biomedical Applications. The conference also included workshop sessions taking place in parallel with the conference ones. Workshop sessions covered themes such as (i) Innovative Technologies Applied to Rural; (ii) Network Modelling, Learning and Analysis; (iii) Intelligent Systems and Machines; (iv) Healthcare Information Systems Interoperability, Security and Efficiency; (v) Applied Statistics and Data Analysis using Computer Science; (vi) Cybersecurity for Smart Cities Development; (vii) Education through ICT; (viii) Unlocking the Artificial Intelligence Interplay with Business Innovation (ix) and Pervasive Information Systems. WorldCIST’20 received about 400 contributions from 57 countries around the world. The papers accepted for presentation and discussion at the conference are published by Springer (this book) in three volumes and will be submitted for indexing by ISI, EI-Compendex, SCOPUS, DBLP and/or Google Scholar, among others. Extended versions of selected best papers will be published in special or regular issues of relevant journals, mainly SCI/SSCI and Scopus/EI-Compendex indexed journals. We acknowledge all of those that contributed to the staging of WorldCIST’20 (authors, committees, workshop organizers and sponsors). We deeply appreciate their involvement and support that were crucial for the success of WorldCIST’20. April 2020

Álvaro Rocha Hojjat Adeli Luís Paulo Reis Sandra Costanzo Irena Orovic Fernando Moreira

Organization

Conference General Chair Álvaro Rocha

University of Coimbra, Portugal

Co-chairs Hojjat Adeli Luis Paulo Reis Sandra Costanzo

The Ohio State University, USA University of Porto, Portugal University of Calabria, Italy

Local Organizing Committee Irena Orovic (Chair) Milos Dakovic Andjela Draganic Milos Brajovic Snezana Scepanvic Rade Ratkovic

University of Montenegro, Montenegro University of Montenegro, Montenegro University of Montenegro, Montenegro University of Montenegro, Montenegro Mediterranean University, Montenegro Faculty of Business and Tourism, Montenegro

Advisory Committee Ana Maria Correia (Chair) Benjamin Lev Chatura Ranaweera Chris Kimble Erik Bohlin Eva Onaindia Gintautas Dzemyda

University of Sheffield, UK Drexel University, USA Wilfrid Laurier University, Canada KEDGE Business School and MRM, UM2, Montpellier, France Chalmers University of Technology, Sweden Polytechnical University of Valencia, Spain Vilnius University, Lithuania

vii

viii

Janusz Kacprzyk Jason Whalley João Tavares Jon Hall Justin Zhang Karl Stroetmann Kathleen Carley Keng Siau Manlio Del Giudice Michael Koenig Miguel-Angel Sicilia Reza Langari Vedat Verter Vishanth Weerakkody Wim Van Grembergen

Organization

Polish Academy of Sciences, Poland Northumbria University, UK University of Porto, Portugal The Open University, UK University of North Florida, USA Empirica Communication and Technology Research, Germany Carnegie Mellon University, USA Missouri University of Science and Technology, USA University of Rome Link Campus, Italy Long Island University, USA University of Alcalá, Spain Texas A&M University, USA McGill University, Canada Bradford University, UK University of Antwerp, Belgium

Program Committee Abdul Rauf Adnan Mahmood Adriana Peña Pérez Negrón Adriani Besimi Agostinho Sousa Pinto Ahmed El Oualkadi Ahmed Rafea Alberto Freitas Aleksandra Labus Alexandru Vulpe Ali Idri Amélia Badica Amélia Cristina Ferreira Silva Almir Souza Silva Neto Amit Shelef Ana Isabel Martins Ana Luis Anabela Tereso Anacleto Correia Anca Alexandra Purcarea Andjela Draganic Aneta Polewko-Klim Aneta Poniszewska-Maranda Angeles Quezada

RISE SICS, Sweden Waterford Institute of Technology, Ireland Universidad de Guadalajara, Mexico South East European University, Macedonia Polytechnic of Porto, Portugal Abdelmalek Essaadi University, Morocco American University in Cairo, Egypt FMUP, University of Porto, Portugal University of Belgrade, Serbia University Politehnica of Bucharest, Romania ENSIAS, University Mohammed V, Morocco Universti of Craiova, Romania Polytechnic of Porto, Portugal IFMA, Brazil Sapir Academic College, Israel University of Aveiro, Portugal University of Coimbra, Portugal University of Minho, Portugal CINAV, Portugal University Politehnica of Bucharest, Romania University of Montenegro, Montenegro University of Białystok, Institute of Informatics, Poland Lodz University of Technology, Poland Instituto Tecnologico de Tijuana, Mexico

Organization

Anis Tissaoui Ankur Singh Bist Ann Svensson Antoni Oliver Antonio Jiménez-Martín Antonio Pereira Armando Toda Arslan Enikeev Benedita Malheiro Boris Shishkov Borja Bordel Branko Perisic Bruno Veloso Carla Pinto Carla Santos Pereira Catarina Reis Cengiz Acarturk Cesar Collazos Christophe Feltus Christophe Soares Christos Bouras Christos Chrysoulas Christos Troussas Ciro Martins Claudio Sapateiro Costin Badica Cristian García Bauza Cristian Mateos Daria Bylieva Dante Carrizo Dayana Spagnuelo Dušan Barać Edita Butrime Edna Dias Canedo Eduardo Santos Egils Ginters Ekaterina Isaeva Elena Mikhailova Eliana Leite Erik Fernando Mendez Garcea Eriks Sneiders Esteban Castellanos

ix

University of Jendouba, Tunisia KIET, India University West, Sweden University of the Balearic Islands, Spain Universidad Politécnica de Madrid, Spain Polytechnic of Leiria, Portugal University of São Paulo, Brazil Kazan Federal University, Russia Polytechnic of Porto, ISEP, Portugal ULSIT/IMI-BAS/IICREST, Bulgaria Universidad Politécnica de Madrid, Spain Faculty of Technical Sciences, Serbia INESC TEC, Portugal Polytechnic of Porto, ISEP, Portugal Universidade Portucalense, Portugal Polytechnic of Leiria, Portugal Middle East Technical University, Turkey Universidad del Cauca, Colombia LIST, Luxembourg University Fernando Pessoa, Portugal University of Patras, Greece London South Bank University, UK University of Piraeus, Greece University of Aveiro, Portugal Polytechnic of Setúbal, Portugal University of Craiova, Romania PLADEMA-UNICEN-CONICET, Argentina ISISTAN-CONICET, UNICEN, Argentina Peter the Great St.Petersburg Polytechnic University, Russia Universidad de Atacama, Chile Vrije Universiteit Amsterdam, Netherlands University of Belgrade, Serbia Lithuanian University of Health Sciences, Lithuania University of Brasilia, Brazil Pontifical Catholic University of Paraná, Brazil Riga Technical University, Latvia Perm State University, Russia ITMO University, Russia University of Minho, Portugal Autonomous Regional University of the Andes, Ecuador Stockholm University, Sweden ESPE, Ecuador

x

Faisal Musa Abbas Fatima Azzahra Amazal Fernando Almeida Fernando Bobillo Fernando Molina-Granja Fernando Moreira Fernando Ribeiro Filipe Caldeira Filipe Portela Filipe Sá Filippo Neri Firat Bestepe Francesco Bianconi Francisco García-Peñalvo Francisco Valverde Galim Vakhitov Gayo Diallo George Suciu Gheorghe Sebestyen Ghani Albaali Gian Piero Zarri Giuseppe Di Massa Gonçalo Paiva Dias Goreti Marreiros Graciela Lara López Habiba Drias Hafed Zarzour Hamid Alasadi Hatem Ben Sta Hector Fernando Gomez Alvarado Hélder Gomes Helia Guerra Henrique da Mota Silveira Henrique S. Mamede Hing Kai Chan Hugo Paredes Ibtissam Abnane Igor Aguilar Alonso

Organization

Abubakar Tafawa Balewa University Bauchi, Nigeria Ibn Zohr University, Morocco INESC TEC and University of Porto, Portugal University of Zaragoza, Spain National University of Chimborazo, Ecuador Portucalense University, Portugal Polytechnic Castelo Branco, Portugal Polytechnic of Viseu, Portugal University of Minho, Portugal Polytechnic of Viseu, Portugal University of Naples, Italy Republic of Turkey Ministry of Development, Turkey Università degli Studi di Perugia, Italy University of Salamanca, Spain Universidad Central del Ecuador, Ecuador Kazan Federal University, Russia Univsersity of Bordeaux, France BEIA Consult International, Romania Technical University of Cluj-Napoca, Romania Princess Sumaya University for Technology, Jordan University Paris-Sorbonne, France University of Calabria, Italy University of Aveiro, Portugal ISEP/GECAD, Portugal University of Guadalajara, Mexico University of Science and Technology Houari Boumediene, Algeria University of Souk Ahras, Algeria Basra University, Iraq University of Tunis at El Manar, Tunisia Universidad Tecnica de Ambato, Ecuador University of Aveiro, Portugal University of the Azores, Portugal University of Campinas (UNICAMP), Brazil University Aberta, Portugal University of Nottingham Ningbo China, China INESC TEC and University of Trás-os-Montes e Alto Douro, Portugal Mohamed V University in Rabat, Morocco Universidad Nacional Tecnológica de Lima Sur, Peru

Organization

Imen Ben Said Inês Domingues Isabel Lopes Isabel Pedrosa Isaías Martins Issam Moghrabi Ivan Dunđer Ivan Lukovic Jaime Diaz Jan Kubicek Jean Robert Kala Kamdjoug Jesús Gallardo Casero Jezreel Mejia Jikai Li Jinzhi Lu Joao Carlos Silva João Manuel R. S. Tavares João Paulo Pereira João Reis João Reis João Rodrigues João Vidal Carvalho Joaquin Nicolas Ros Jorge Barbosa Jorge Buele Jorge Esparteiro Garcia Jorge Gomes Jorge Oliveira e Sá José Álvarez-García José Braga de Vasconcelos Jose Luis Herrero Agustin José Luís Reis Jose Luis Sierra Jose M. Parente de Oliveira José Machado José Paulo Lousado Jose Torres José-Luís Pereira Juan M. Santos Juan Manuel Carrillo de Gea Juan Pablo Damato Juncal Gutiérrez-Artacho Kalinka Kaloyanova

xi

Université de Sfax, Tunisia University of Coimbra, Portugal Polytechnic of Bragança, Portugal Coimbra Business School ISCAC, Portugal University of Leon, Spain Gulf University for Science and Technology, Kuwait University of Zabreb, Croatia University of Novi Sad, Serbia University of La Frontera, Chile Technical University of Ostrava, Czech Republic Catholic University of Central Africa, Cameroon University of Zaragoza, Spain CIMAT, Unidad Zacatecas, Mexico The College of New Jersey, USA KTH Royal Institute of Technology, Sweden IPCA, Portugal University of Porto, FEUP, Portugal Polytechnic of Bragança, Portugal University of Aveiro, Portugal University of Lisbon, Portugal University of the Algarve, Portugal Polytechnic of Coimbra, Portugal University of Murcia, Spain Polytechnic of Coimbra, Portugal Technical University of Ambato, Ecuador Polytechnic Institute of Viana do Castelo, Portugal University of Lisbon, Portugal University of Minho, Portugal University of Extremadura, Spain Universidade New Atlântica, Portugal University of Extremadura, Spain ISMAI, Portugal Complutense University of Madrid, Spain Aeronautics Institute of Technology, Brazil University of Minho, Portugal Polytechnic of Viseu, Portugal Universidty Fernando Pessoa, Portugal Universidade do Minho, Portugal University of Vigo, Spain University of Murcia, Spain UNCPBA-CONICET, Argentina University of Granada, Spain Sofia University, Bulgaria

xii

Kamel Rouibah Khalid Benali Korhan Gunel Krzysztof Wolk Kuan Yew Wong Laila Cheikhi Laura Varela-Candamio Laurentiu Boicescu Leonardo Botega Leonid Leonidovich Khoroshko Lia-Anca Hangan Lila Rao-Graham Łukasz Tomczyk Luis Alvarez Sabucedo Luis Cavique Luis Gouveia Luis Mendes Gomes Luis Silva Rodrigues Luiz Rafael Andrade Luz Sussy Bayona Oré Maksim Goman Manal el Bajta Manuel Antonio Fernández-Villacañas Marín Manuel Silva Manuel Tupia Manuel Au-Yong-Oliveira Marciele Bernardes Marco Bernardo Marco Ronchetti Mareca María PIlar Marek Kvet María de la Cruz del Río-Rama Maria João Ferreira Maria João Varanda Pereira Maria José Angélico Maria José Sousa María Teresa García-Álvarez Mariam Bachiri

Organization

Kuwait University, Kuwait LORIA University of Lorraine, France Adnan Menderes University, Turkey Polish-Japanese Academy of Information Technology, Poland Universiti Teknologi Malaysia (UTM), Malaysia University Mohammed V, Rabat, Morocco Universidade da Coruña, Spain E.T.T.I. U.P.B., Romania University Centre Eurípides of Marília (UNIVEM), Brazil Moscow Aviation Institute (National Research University), Russia Technical University of Cluj-Napoca, Romania University of the West Indies, Jamaica Pedagogical University of Cracow, Poland University of Vigo, Spain University Aberta, Portugal University Fernando Pessoa, Portugal University of the Azores, Portugal Polythencic of Porto, Portugal Tiradentes University, Brazil Universidad Nacional Mayor de San Marcos, Peru JKU, Austria ENSIAS, Morocco Technical University of Madrid, Spain

Polytechnic of Porto and INESC TEC, Portugal Pontifical Catholic University of Peru, Peru University of Aveiro, Portugal University of Minho, Brazil Polytechnic of Viseu, Portugal Universita’ di Trento, Italy Universidad Politécnica de Madrid, Spain Zilinska Univerzita v Ziline, Slovakia University of Vigo, Spain Universidade Portucalense, Portugal Polytechnic of Bragança, Portugal Polytechnic of Porto, Portugal University of Coimbra, Portugal University of A Coruna, Spain ENSIAS, Morocco

Organization

Marijana Despotovic-Zrakic Mário Antunes Marisa Maximiano Marisol Garcia-Valls Maristela Holanda Marius Vochin Marlene Goncalves da Silva Maroi Agrebi Martin Henkel Martín López Nores Martin Zelm Mawloud Mosbah Michal Adamczak Michal Kvet Miguel António Sovierzoski Mihai Lungu Mircea Georgescu Mirna Muñoz Mohamed Hosni Monica Leba Mu-Song Chen Natalia Grafeeva Natalia Miloslavskaya Naveed Ahmed Neeraj Gupta Nelson Rocha Nikolai Prokopyev Niranjan S. K. Noemi Emanuela Cazzaniga Noureddine Kerzazi Nuno Melão Nuno Octávio Fernandes Olimpiu Stoicuta Patricia Zachman Patrick C.-H. Soh Paula Alexandra Rego Paulo Maio Paulo Novais

xiii

Faculty Organizational Science, Serbia Polytechnic of Leiria and CRACS INESC TEC, Portugal Polytechnic Institute of Leiria, Portugal Polytechnic University of Valencia, Spain University of Brasilia, Brazil E.T.T.I. U.P.B., Romania Universidad Simón Bolívar, Venezuela University of Polytechnique Hauts-de-France, France Stockholm University, Sweden University of Vigo, Spain INTEROP-VLab, Belgium University 20 Août 1955 of Skikda, Algeria Poznan School of Logistics, Poland University of Zilina, Slovakia Federal University of Technology - Paraná, Brazil University of Craiova, Romania Al. I. Cuza University of Iasi, Romania Centro de Investigación en Matemáticas A.C., Mexico ENSIAS, Morocco University of Petrosani, Romania Da-Yeh University, China Saint Petersburg University, Russia National Research Nuclear University MEPhI, Russia University of Sharjah, United Arab Emirates KIET Group of Institutions Ghaziabad, India University of Aveiro, Portugal Kazan Federal University, Russia JSS Science and Technology University, India Politecnico di Milano, Italy Polytechnique Montréal, Canada Polytechnic of Viseu, Portugal Polytechnic of Castelo Branco, Portugal University of Petrosani, Romania Universidad Nacional del Chaco Austral, Argentina Multimedia University, Malaysia Polytechnic of Viana do Castelo and LIACC, Portugal Polytechnic of Porto, ISEP, Portugal University of Minho, Portugal

xiv

Paulvanna Nayaki Marimuthu Paweł Karczmarek Pedro Rangel Henriques Pedro Sobral Pedro Sousa Philipp Brune Piotr Kulczycki Prabhat Mahanti Rabia Azzi Radu-Emil Precup Rafael Caldeirinha Rafael M. Luque Baena Rahim Rahmani Raiani Ali Ramayah T. Ramiro Gonçalves Ramon Alcarria Ramon Fabregat Gesa Renata Maria Maracho Reyes Juárez Ramírez Rui Jose Rui Pitarma Rui S. Moreira Rustam Burnashev Saeed Salah Said Achchab Sajid Anwar Sami Habib Samuel Sepulveda Sanaz Kavianpour Sandra Patricia Cano Mazuera Savo Tomovic Sassi Sassi Seppo Sirkemaa Sergio Albiol-Pérez Shahed Mohammadi Shahnawaz Talpur

Organization

Kuwait University, Kuwait The John Paul II Catholic University of Lublin, Poland University of Minho, Portugal University Fernando Pessoa, Portugal University of Minho, Portugal Neu-Ulm University of Applied Sciences, Germany Systems Research Institute, Polish Academy of Sciences, Poland University of New Brunswick, Canada Bordeaux University, France Politehnica University of Timisoara, Romania Polytechnic of Leiria, Portugal University of Malaga, Spain University Stockholm, Sweden Hamad Bin Khalifa University, Qatar Universiti Sains Malaysia, Malaysia University of Trás-os-Montes e Alto Douro & INESC TEC, Portugal Universidad Politécnica de Madrid, Spain University of Girona, Spain Federal University of Minas Gerais, Brazil Universidad Autonoma de Baja California, Mexico University of Minho, Portugal Polytechnic Institute of Guarda, Portugal UFP & INESC TEC & LIACC, Portugal Kazan Federal University, Russia Al-Quds University, Palestine Mohammed V University in Rabat, Morocco Institute of Management Sciences Peshawar, Pakistan Kuwait University, Kuwait University of La Frontera, Chile University of Technology, Malaysia University of San Buenaventura Cali, Colombia University of Montenegro, Montenegro FSJEGJ, Tunisia University of Turku, Finland University of Zaragoza, Spain Ayandegan University, Iran Mehran University of Engineering & Technology Jamshoro, Pakistan

Organization

Silviu Vert Simona Mirela Riurean Slawomir Zolkiewski Solange N. Alves-Souza Solange Rito Lima Sonia Sobral Sorin Zoican Souraya Hamida Sümeyya Ilkin Syed Nasirin Taoufik Rachad Tatiana Antipova Teresa Guarda Tero Kokkonen The Thanh Van Thomas Weber Timothy Asiedu Tom Sander Tomaž Klobučar Toshihiko Kato Tzung-Pei Hong Valentina Colla Veronica Segarra Faggioni Victor Alves Victor Georgiev Victor Kaptelinin Vincenza Carchiolo Vitalyi Igorevich Talanin Wafa Mefteh Wolf Zimmermann Yadira Quiñonez Yair Wiseman Yuhua Li Yuwei Lin Yves Rybarczyk Zorica Bogdanovic

xv

Politehnica University of Timisoara, Romania University of Petrosani, Romania Silesian University of Technology, Poland University of São Paulo, Brazil University of Minho, Portugal Portucalense University, Portugal Polytechnic University of Bucharest, Romania Batna 2 University, Algeria Kocaeli University, Turkey Universiti Malaysia Sabah, Malaysia University Mohamed V, Morocco Institute of Certified Specialists, Russia University Estatal Peninsula de Santa Elena, Ecuador JAMK University of Applied Sciences, Finland HCMC University of Food Industry, Vietnam EPFL, Switzerland TIM Technology Services Ltd., Ghana New College of Humanities, Germany Jozef Stefan Institute, Slovenia University of Electro-Communications, Japan National University of Kaohsiung, Taiwan Scuola Superiore Sant’Anna, Italy Private Technical University of Loja, Ecuador University of Minho, Portugal Kazan Federal University, Russia Umeå University, Sweden University of Catania, Italy Zaporozhye Institute of Economics and Information Technologies, Ukraine Tunisia Martin Luther University Halle-Wittenberg, Germany Autonomous University of Sinaloa, Mexico Bar-Ilan University, Israel Cardiff University, UK University of Roehampton, UK Dalarna University, Sweden University of Belgrade, Serbia

Contents

Health Informatics A Product and Service Concept Proposal to Improve the Monitoring of Citizens’ Health in Society at Large . . . . . . . . . . . . . . . . . . . . . . . . . . Luís Fonseca, João Barroso, Miguel Araújo, Rui Frazão, and Manuel Au-Yong-Oliveira Artificial Neural Networks Interpretation Using LIME for Breast Cancer Diagnosis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Hajar Hakkoum, Ali Idri, and Ibtissam Abnane Energy Efficiency and Usability of Web-Based Personal Health Records . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . José Alberto García-Berná, Sofia Ouhbi, José Luis Fernández-Alemán, Juan Manuel Carrillo-de-Gea, and Joaquín Nicolás A Complete Prenatal Solution for a Reproductive Health Unit in Morocco . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Mariam Bachiri, Ali Idri, Taoufik Rachad, Hassan Alami, and Leanne M. Redman Machine Learning and Image Processing for Breast Cancer: A Systematic Map . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Hasnae Zerouaoui, Ali Idri, and Khalid El Asnaoui A Definition of a Coaching Plan to Guide Patients with Chronic Obstructive Respiratory Diseases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Diogo Martinho, Ana Vieira, João Carneiro, Constantino Martins, Ana Almeida, and Goreti Marreiros Reviewing Data Analytics Techniques in Breast Cancer Treatment . . . . Mahmoud Ezzat and Ali Idri

3

15

25

36

44

54

65

xvii

xviii

Contents

Enabling Smart Homes Through Health Informatics and Internet of Things for Enhanced Living Environments . . . . . . . . . . Gonçalo Marques and Rui Pitarma MyContraception: An Evidence-Based Contraception mPHR for Better Contraceptive Fit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Manal Kharbouch, Ali Idri, Taoufiq Rachad, Hassan Alami, Leanne Redman, and Youssef Stelate Predictors of Acceptance and Rejection of Online Peer Support Groups as a Digital Wellbeing Tool . . . . . . . . . . . . . . . . . . . . . . . . . . . . John McAlaney, Manal Aldhayan, Mohamed Basel Almourad, Sainabou Cham, and Raian Ali

76

86

95

Assessing Daily Activities Using a PPG Sensor Embedded in a Wristband-Type Activity Tracker . . . . . . . . . . . . . . . . . . . . . . . . . . 108 Alexandra Oliveira, Joyce Aguiar, Eliana Silva, Brígida Mónica Faria, Helena Gonçalves, Luís Teófilo, Joaquim Gonçalves, Victor Carvalho, Henrique Lopes Cardoso, and Luís Paulo Reis Simulation of a Robotic Arm Controlled by an LCD Touch Screen to Improve the Movements of Physically Disabled People . . . . . . . . . . . 120 Yadira Quiñonez, Oscar Zatarain, Carmen Lizarraga, Juan Peraza, Rogelio Estrada, and Jezreel Mejía Information Technologies in Education Performance Indicator Based on Learning Routes: Second Round . . . . 137 Franklin Chamba, Susana Arias, Gustavo Alvarez, and Héctor Gómez Evaluating the Acceptance of Blended-Learning Tools: A Case Study Using SlideWiki Presentation Rooms . . . . . . . . . . . . . . . . 142 Anne Martin, Bianca Bergande, and Roy Meissner Adaptivity: A Continual Adaptive Online Knowledge Assessment System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152 Miran Zlatović and Igor Balaban The First Programming Language and Freshman Year in Computer Science: Characterization and Tips for Better Decision Making . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162 Sónia Rolland Sobral Design of a Network Learning System for the Usage of Surgical Instruments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175 Ting-Kai Hwang, Bih-Huang Jin, and Su-Chiu Wang CS1 and CS2 Curriculum Recommendations: Learning from the Past to Try not to Rediscover the Wheel Again . . . . . . . . . . . . 182 Sónia Rolland Sobral

Contents

xix

On the Role of Python in Programming-Related Courses for Computer Science and Engineering Academic Education . . . . . . . . . 192 Costin Bădică, Amelia Bădică, Mirjana Ivanović, Ionuţ Dorinel Murareţu, Daniela Popescu, and Cristinel Ungureanu Validating the Shared Understanding Construction in Computer Supported Collaborative Work in a Problem-Solving Activity . . . . . . . . 203 Vanessa Agredo-Delgado, Pablo H. Ruiz, Alicia Mon, Cesar A. Collazos, Fernando Moreira, and Habib M. Fardoun Improving Synchrony in Small Group Asynchronous Online Discussions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215 Samuli Laato and Mari Murtonen Academic Dishonesty Prevention in E-learning University System . . . . . 225 Daria Bylieva, Victoria Lobatyuk, Sergei Tolpygin, and Anna Rubtsova Curriculum for Digital Culture at ITMO University . . . . . . . . . . . . . . . 235 Elena Mikhailova, Anton Boitsev, Olga Egorova, Natalia Grafeeva, Aleksei Romanov, and Dmitriy Volchek ICT Impact in Orientation and University Tutoring According to Students Opinion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245 Antonio Pantoja Vallejo, Beatriz Berrios Aguayo, and María Jesús Yolanda Colmenero Ruiz Blockchain Security and Privacy in Education: A Systematic Mapping Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253 Attari Nabil, Khalid Nafil, and Fouad Mounir The Development of Pre-service Teacher’s Reflection Skills Through Video-Based Classroom Observation . . . . . . . . . . . . . . . . . . . . 263 Ana R. Luís Formative Assessment and Digital Tools in a School Context . . . . . . . . 271 Sandra Paiva, Luís Paulo Reis, and Lia Raquel Information Technologies in Radiocommunications Compact Slotted Planar Inverted-F Antenna: Design Principle and Preliminary Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 287 Sandra Costanzo and Adil Masoud Qureshi Technologies for Biomedical Applications Statistical Analysis to Control Foot Temperature for Diabetic People . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 295 José Torreblanca González, Alfonso Martínez Nova, A. H. Encinas, Jesús Martín-Vaquero, and A. Queiruga-Dios

xx

Contents

Sensitive Mannequin for Practicing the Locomotor Apparatus Recovery Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 307 Cosmin Strilețchi and Ionuț Dan Cădar Pervasive Information Systems Data Intelligence Using PDME for Predicting Cardiovascular Predictive Failures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 317 Francisco Freitas, Rui Peixoto, Carlos Filipe Portela, and Manuel Santos Design of a Microservices Chaining Gamification Framework . . . . . . . . 327 Ricardo Queirós PWA and Pervasive Information System – A New Era . . . . . . . . . . . . . 334 Gisela Fernandes, Filipe Portela, and Manuel Filipe Santos Inclusive Education through ICT Young People Participation in the Digital Society: A Case Study in Brazil . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 347 Everton Knihs and Alicia García-Holgado Blockchain Technology to Support Smart Learning and Inclusion: Pre-service Teachers and Software Developers Viewpoints . . . . . . . . . . 357 Solomon Sunday Oyelere, Umar Bin Qushem, Vladimir Costas Jauregui, Özgür Yaşar Akyar, Łukasz Tomczyk, Gloria Sanchez, Darwin Munoz, and Regina Motz Digital Storytelling in Teacher Education for Inclusion . . . . . . . . . . . . . 367 Özgür Yaşar Akyar, Gıyasettin Demirhan, Solomon Sunday Oyelere, Marcelo Flores, and Vladimir Costas Jauregui In Search of Active Life Through Digital Storytelling: Inclusion in Theory and Practice for the Physical Education Teachers . . . . . . . . . 377 Burcu Şimşek and Özgür Yaşar Akyar Accessibility Recommendations for Open Educational Resources for People with Learning Disabilities . . . . . . . . . . . . . . . . . . . . . . . . . . . 387 Valéria Farinazzo Martins, Cibelle Amato, Łukasz Tomczyk, Solomon Sunday Oyelere, Maria Amelia Eliseo, and Ismar Frango Silveira Digital Storytelling and Blockchain as Pedagogy and Technology to Support the Development of an Inclusive Smart Learning Ecosystem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 397 Solomon Sunday Oyelere, Ismar Frango Silveira, Valeria Farinazzo Martins, Maria Amelia Eliseo, Özgür Yaşar Akyar, Vladimir Costas Jauregui, Bernardo Caussin, Regina Motz, Jarkko Suhonen, and Łukasz Tomczyk

Contents

xxi

Aggregation Bias: A Proposal to Raise Awareness Regarding Inclusion in Visual Analytics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 409 Andrea Vázquez-Ingelmo, Francisco J. García-Peñalvo, and Roberto Therón A Concrete Action Towards Inclusive Education: An Implementation of Marrakesh Treaty . . . . . . . . . . . . . . . . . . . . . . . . 418 Virginia Rodés and Regina Motz Intelligent Systems and Machines Cloud Computing Customer Communication Center . . . . . . . . . . . . . . . 429 George Suciu, Romulus Chevereșan, Svetlana Segărceanu, Ioana Petre, Andrei Scheianu, and Cristiana Istrate International Workshop on Healthcare Information Systems Interoperability, Security and Efficiency A Study on CNN Architectures for Chest X-Rays Multiclass Computer-Aided Diagnosis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 441 Ana Ramos and Victor Alves A Thermodynamic Assessment of the Cyber Security Risk in Healthcare Facilities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 452 Filipe Fernandes, Victor Alves, Joana Machado, Filipe Miranda, Dinis Vicente, Jorge Ribeiro, Henrique Vicente, and José Neves How to Assess the Acceptance of an Electronic Health Record System? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 466 Catarina Fernandes, Filipe Portela, Manuel Filipe Santos, José Machado, and António Abelha An Exploratory Study of a NoSQL Database for a Clinical Data Repository . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 476 Francini Hak, Tiago Guimarães, António Abelha, and Manuel Santos Clinical Decision Support Using Open Data . . . . . . . . . . . . . . . . . . . . . . 484 Francini Hak, Tiago Guimarães, António Abelha, and Manuel Santos Spatial Normalization of MRI Brain Studies Using a U-Net Based Neural Network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 493 Tiago Jesus, Ricardo Magalhães, and Victor Alves Business Analytics for Social Healthcare Institution . . . . . . . . . . . . . . . . 503 Miguel Quintal, Tiago Guimarães, Antonio Abelha, and Manuel Filipe Santos Step Towards Monitoring Intelligent Agents in Healthcare Information Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 510 Regina Sousa, Diana Ferreira, António Abelha, and José Machado

xxii

Contents

Network Modeling, Learning and Analysis A Comparative Study of Representation Learning Techniques for Dynamic Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 523 Carlos Ortega Vázquez, Sandra Mitrović, Jochen De Weerdt, and Seppe vanden Broucke Metadata Action Network Model for Cloud Based Development Environment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 531 Mehmet N. Aydin, Ziya N. Perdahci, I. Safak, and J. (Jos) van Hillegersberg Clustering Foursquare Mobility Networks to Explore Urban Spaces . . . 544 Olivera Novović, Nastasija Grujić, Sanja Brdar, Miro Govedarica, and Vladimir Crnojević Innovative Technologies Applied to Rural Regions The Influence of Digital Marketing Tools Perceived Usefulness in a Rural Region Destination Image . . . . . . . . . . . . . . . . . . . . . . . . . . . 557 Filipa Jorge, Mário Sérgio Teixeira, and Ramiro Gonçalves Ñawi Project: Visual Health for Improvement of Education in High Andean Educational Communities in Perú . . . . . . . . . . . . . . . . 570 Xavi Canaleta, Eva Villegas, David Fonseca, Rafel Zaragoza, Guillem Villa, David Badia, and Emiliano Labrador Building Smart Rural Regions: Challenges and Opportunities . . . . . . . . 579 Carlos R. Cunha, João Pedro Gomes, Joana Fernandes, and Elisabete Paulo Morais The Power of Digitalization: The Netflix Story . . . . . . . . . . . . . . . . . . . . 590 Manuel Au-Yong-Oliveira, Miguel Marinheiro, and João A. Costa Tavares An Online Sales System to Be Managed by People with Mental Illness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 600 Alicia García-Holgado, Samuel Marcos-Pablos, and Francisco J. García-Peñalvo Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 613

Health Informatics

A Product and Service Concept Proposal to Improve the Monitoring of Citizens’ Health in Society at Large Luís Fonseca1, João Barroso1, Miguel Araújo1, Rui Frazão1, and Manuel Au-Yong-Oliveira2(&) 1

Department of Electronics, Telecommunications and Informatics, University of Aveiro, Aveiro, Portugal {luismiguel.fonseca,joao.barroso,mdaraujo, ruifilipefrazao}@ua.pt 2 GOVCOPP, Department of Economics, Management, Industrial Engineering and Tourism, University of Aveiro, Aveiro, Portugal [email protected]

Abstract. Nowadays wearable devices are very popular. The reason for that is the sudden reduction in pricing and the increase in functionalities. Healthcare services have been greatly benefiting from the emergence of these devices since they can collect vital signs and help healthcare professionals to easily monitor patients. Medical wellness, prevention, diagnosis, treatment and monitoring services are the main focus of Healthcare applications. Some companies have already invested in this market and we present some of them and their strategies. Furthermore, we also conducted a group interview with Altice Labs in order to better understand the critical points and challenges they encountered while developing and maintaining their service. With the purpose of comprehending users’ receptiveness to mHealth systems (mobile health systems which users wear - wearables) and their opinion about sharing data, we also created a questionnaire (which had 114 valid responses). Based on the research done we propose a different approach. In our product and service concept solution, which we share herein, we consider people of all ages to be targets for the product/service and, beyond that, we consider the use of machine learning techniques to extract knowledge from the information gathered. Finally, we discuss the advantages and drawbacks of this kind of system, showing our critical point of view. Keywords: Healthcare
mHealth monitoring
Health tracking


Wearable technology
Biomedical

1 Introduction The healthcare system is intended to efficiently provide healthcare services so as to meet the health needs and demands of individuals, families and the community. In the last few years, technology and healthcare have been combined to improve the quality of life of the population around the world. The interest in the mixing of these fields is to © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2020 Á. Rocha et al. (Eds.): WorldCIST 2020, AISC 1161, pp. 3–14, 2020. https://doi.org/10.1007/978-3-030-45697-9_1

4

L. Fonseca et al.

understandingly be capable of improving healthcare systems by taking advantage of the ubiquitous and powerful mobile devices that everyone carries with them, on a daily basis, that are able to gather information during users’ daily activities and correlate it with their health status. The term mHealth (mobile Health) comes with the development and use of less intrusive and more comfortable mobile devices, like smartphones, wearable devices incorporated in clothes, bracelets, necklaces, watches, or many others, or sensors and applications (apps) that track users’ physiological information. This concept is capable of revolutionizing the healthcare service delivery by preventing and diagnosing earlystage medical conditions, leading to an increase in its quality and efficiency. The World Health Organization (WHO) states that mHealth systems in high-income countries allow for the reducing of healthcare costs whereas developing countries tend to provide access to primary healthcare services [1]. With the rising concern about healthcare and mHealth topics and the increase in the amount of data gathered, computational techniques started being used to process the data to extract information and knowledge from it and try to improve healthcare systems and services provided to its users. Since 2017, the Machine Learning for Healthcare annual research meeting has as its objective to bring together two usually insular disciplines: computer science and medical research [2]. The contribution that this paper offers is the analysis of some already existing mHealth systems and understanding the contributing/decisive factors, concerns and different approaches to future systems. Based on research about wearable technology devices and on the market, and on organizations which act towards the improvement of healthcare systems and lifestyle, it is easily noticed that most products are developed to operate in a more specific field, like the tracking of physical activity or biometrics measurement, instead of crossreferencing data from other biometric markers or even other individuals. To clarify some concepts and in order to understand the strategic point of view of a mHealth services provider and how their product contributes to its users’ lifestyles, a group interview with Altice Labs (in the Telecom industry), an Altice branch in Aveiro, Portugal, was performed to talk about their solution, SmartAL [3]. During the interview, aspects like their system’s features and the possible applications of machine learning techniques on the gathered data to predict medical conditions were discussed. After that a questionnaire was created and administered with the objective of understanding people’s knowledge on monitoring devices and their opinions about mHealth systems and the sharing of data. Finally, the conclusions taken from the analyzed solutions are presented and a newly proposed system is described with regards to the related work as well as to the benefits and drawbacks it may bring. A positive impact on the community may well be the result.

A Product and Service Concept Proposal

5

2 Literature Review In this section we shall describe what has already been researched about wearable devices and medical data processing and we shall explain some mHealth systems implemented by enterprises or organizations that are similar to our proposed system. The literature review will be the basis of the development of a possible solution to improve the quality of life of the population while at the same time reducing healthcare costs. 2.1

Wearable Devices

Wearables are electronic technologies or devices that are incorporated into items of clothing and personal accessories. This technology has seen significant growth in the last few years [4]. Comparatively to smartphones, these devices have a great advantage because they are used close to the human body, so they can read biometric data more accurately and beyond that can read a greater diversity of biometrical signs. In some daily activities, such as sleeping, as well as during a lot of physical exercise, people do not usually have their smartphones close by, so they cannot continuously read the users’ signs, unlike what occurs with wearables that are always in contact with the user during every task throughout the day [5]. Actually, wearables can read a great diversity of biometrical signs including: heartrate, motion (acceleration and direction), earth magnet field, ambient light, steps, distance, calories burned, active minutes, hourly activity and stationary time, sleep monitoring, movement, muscle fatigue, joints’ pressure, heart rate, breathing, temperature, light, falls, fatigue, among others, and users can be warned if some irregular markers occur [6]. According to Berglund et al.’s research [7], most of this technology is applied in sports/fitness and lifestyle/fashion. In this same article it is mentioned that technology is incorporated mainly in watches, jewelry and shirts [7]. Over time this technology has become more popular and the main reason for that is comfort, as they have become more sophisticated and more accessible [8]. Some researchers believe that in the future doctors will prescribe a treatment that includes both medicine and devices such as wearables [9]. However, despite the increased popularity of these devices, Kolasinska et al. refer that few people use smartwatches; albeit, if they were to use sensors they prefer that they would to be incorporated in smartwatches or clothes [10]. An example of wearable clothes are the “MagIC System” that is a prototype of a t-shirt with sensors [11] and a smart bra with a mechanism to detect early signs of breast cancer through thermodynamic sensors [12]. 2.2

Data Processing and Analytics

Data produced by monitoring devices can be stored in local devices, but normally is saved in the cloud in order to be easily accessed [13]. That is important for some

6

L. Fonseca et al.

applications so that medical staff, trusted relatives and even the patient might access that data. These devices provide high-density data [14] (e.g., 10–500 times per second) which can be processed using algorithms emerging in the machine learning field. Machine learning is one of the many areas of artificial intelligence oriented towards the study, processing and analysis of large amounts of data with the objective of developing computational models capable of automatic learning [15]. These models are able to detect relationships in the data, that would be difficult for humans to perceive. Algorithms that make classification decisions have a tremendous dependency on the quantity of the data needed in order to learn, so there is an opportunity to use data from mHealth devices. It is possible to draw conclusions about individuals such as type of physical activity, level of stress, or intensity of pain [14]. For physical activity classifications Knearest neighbors (KNN) have been used as well as Bayes techniques from either single accelerometer or multiple types of sensors; Artificial neural network (ANN) and decision tree modelling are used to recognize these activities, fusing data from accelerometers and GPS [16]. As fall detection is a major concern for elderly people, support vector machines (SVM) have been studied to detect those events, as well as gesture classification [16]. By using different data sources, it is possible to improve the validity of the estimates with data fusion. For example, the accelerometer information can improve the interpretation of the raw electrocardiogram (ECG) data while a person is exercising, because the ECG signal is strongly affected. Improving the interpretation of data, combining different signs, can reduce incorrect clinical evaluations that can lead to false alarms. In practice, fusion can be really challenging because the spatial and temporal resolutions of different data sources can be different [14]. 2.3

Enterprises and Services

Nowadays, it is possible to find solutions, mostly for personal use only or in an organizational context. With the appearance and rising in popularity of the concept of mHealth, some enterprises began to invest in services within this concept and in this section we shall present some of them. Systems such as smartwatches can help users by informing them about their health status during daily activities such as counting steps, monitoring sleep, among many others. Apple has invested in the development of the Apple Watch aiming for personal use. This device offers the possibility to take an electrocardiogram (ECG) [17] and specifically for women, makes it easy to log information about their menstrual cycle [18]. Garmin also developed a fitness tracker capable of preventing the risk of injury. The tracker issues warnings when the risk of injury is high and gives an estimate of how long the user should rest [19]. In existing markets it is possible to find a great diversity of these devices from the cheapest, such as smartbands, to more expensive ones, such as powerful smartwatches.

A Product and Service Concept Proposal

7

In response to a specific use case, in this case diabetes, Medtronic created MiniMed, a device with the ability to automatically adjust basal insulin based on patient’s CGM readings. It also keeps record of the last 90 days of pump history and generates reports [20]. In an organizational context solutions are starting to be implemented, for example in hospitals, clinics and private entities. An example of that occurred from 2009 to 2011, in London’s Chelsea and Westminster Hospital. They invested in an e-Health pilot project that consists in recording and storing patients’ activities on a platform so that doctors can easily access users’ data to analyze it and reach conclusions [13]. Altice developed a pilot project where the intended users are the elderly. Smart Assisted Living (SmartAL) is a social and health support solution that includes the telemonitoring of vital signs, such as weight, blood pressure, pulse rate, accessible via TV with an interactive IPTV service, Android app or web browser. This system makes it possible to configure threshold values of biometric data to emit alerts to healthcare professionals, family and friends [3]. On the one hand, most of these kinds of organization take advantage of the increasing ease of access to wearable devices, mostly for fitness and lifestyle purposes. On the other hand, in many cases the integration of the elderly in this type of systems has been a common concern.

3 Methodology Based on the literature review section, the strategies of some organizations that develop mhealth systems were analyzed. A group interview with Altice Labs was scheduled in order to gain more knowledge about this kind of system. Before the group interview with the company we created an interview script with the aim to understand what were the main challenges that they encountered while developing the product, which were its weaknesses and also if the product had a margin of progress with the improvement of technology. In other words, we intended to do a high-level SWOT analysis of their Assisted Living product. The group interview was performed on the 17th of October 2019, for around thirty minutes, in Aveiro, using an interview script, and involved two employees of the firm (both from the product development department). Notes were taken on the topics discussed. Authorization for the use of the material discussed was a topic of the small group interview (at the end). Some material was deemed as not being able to be used by the research group. The interview led to important conclusions on the product concept developed during this research study. During the group interview, we discussed aspects such as their system’s features, people’s receptivity to it and other issues duly uncovered. Further to that, other approaches and scenarios where their system could be integrated were discussed with the objective of contributing to a better solution. In an ideal system, it is essential that the system be capable of answering all people’s needs. We also performed a questionnaire with the purpose of perceiving citizens’ knowledge of this kind of device, their opinion about sharing personal data and their

8

L. Fonseca et al.

receptiveness to this kind of system. Thus, what we really wanted to understand with this questionnaire was also if the younger generation was receptive to use a mHealth product, thus expanding the market and increasing the target audience. In order to reach a group of heterogeneous participants concerning age level, we distributed the questionnaire at a high school to receive feedback from younger people, in a nursing home to get responses from older people, while we also visited a company to get responses from middle-aged people. In the survey, 114 people agreed to participate, whereby 51.8% were male and 48.2% female. With regards to the age group 14.9% were up to 17 years old, 40.4% were between 18 and 35 years old, 28.9% were between 36 and 60 years old and the remaining 15.8% were more than 60 years old. As concerns academic qualifications, 14.9% had attended primary school, 3.5% had attended the 2nd cycle of school, 2.6% had attended the 3rd cycle of school, 24.6% had been to high school, 27.2% had a licentiate degree, 26.3% had a Master’s degree and the remaining 0.9% had a PhD degree. It was important to discriminate the age of the participants as well as their academic qualifications in order to understand if they had an influence on their choices. The questionnaire contained the following questions: • Do you know any kind of device for measuring vital signs (example: bracelets and smart watches, chest bands, etc.)? • Would you be willing to use one of these devices to monitor your vital signs? • Which biometric signs are you most interested in monitoring? • Would you be willing to send your data to an outside entity and thereby benefit from a closer monitoring of your health? • Are you aware of the General Data Protection Regulation (GDPR)? • Would you be willing to pay for a service that uses the information of your vital signs, and thus enjoy a closer monitoring of your health?

4 Results In this section, we shall present the conclusions obtained from the group interview with Altice Labs as well as the results from the questionnaire. 4.1

SmartAL – Software Platform Developed by Altice Labs for Monitoring the Elderly

As referred to above, SmartAL’s focus is the monitorization of the elderlies’ vital signs while raising alerts to family, friends and healthcare professionals. For that it is necessary to define threshold values for each vital signal. Another functionality of this system is to allow healthcare professionals and relatives to remotely access user data. In this application each user is individually analyzed and there is no direct comparison between the user and groups of similar users.

A Product and Service Concept Proposal

9

According to Altice Labs’ research, they claim vital signs are not enough to predict medical conditions. For a precise diagnostic it is mandatory to have complementary exams, such as blood analyses and sonographies, among others. Another point mentioned was that medical staff are usually not in favour of giving a computer system the possibility of diagnosis, as they believe that the human factor is extremely important to infer the final decision based on data semantics. One of the main points that we gathered concerning their solution is that it is necessary to manually insert users’ vital signs on the SmartAL platform. We believe that a significative improvement would be an automatic collection of these signs using mHealth devices. 4.2

Survey

In this section we shall present the conclusions after the questionnaire referred to above (114 responses). Considering the global results obtained, it was found that: • Around 75% of the participants know about wearable devices; • Although 25% do not know about wearable devices, only 16.7% would not use them, i.e., 83.3% would accept to use these devices for monitoring purposes; • 70.2% would be willing to share their data; • 72.8% of the participants have knowledge about laws on data protection and privacy (GDPR); • Only 21.1% would pay and around 49% of all participants are undecided; • In general, people are mostly preoccupied in measuring their heart rate (74.6%) and their stress levels (64.9%). Analyzing each age group, the following was observed: • There is a tendency for the older generation to know less about the monitoring devices (see Fig. 1); • Although the elderly do not know about these devices, they would be willing to use them. In the other younger age groups people would generally use them (85%– 95%); • Few young and elderly people know about GDPR (less than 50%), however people between 18 and 60 years are aware of it (around 85%); • In general, those who know about GDPR would not be comfortable sharing their data; on the contrary, those who do not know about data privacy regulations would not mind sharing; • No one under 17 years of age answered they would pay for the service. The interval of people from 18–60 years of age had a similar relative percentage, where 30% said they would not pay, while 20% would, and the remaining 50% are undecided. In the elderly group, only 20% are undecided and 50% of them would definitely be willing to pay.

10

L. Fonseca et al.

Fig. 1. Knowledge about wearable devices

5 Discussion of a New Possible Solution for the General Population and for All Age Groups In accordance with the questionnaire results, the literature review and the group interview with Altice Labs, we present a possible solution. According to the questionnaire results, it will be interesting to apply this kind of system to the whole population, as all age groups would be willing to use wearable devices for monitoring and not only elderly people. Although most systems are mainly focused on elderly people, we noticed that most of this group answered that they are not aware of this kind of device. It is crucial that users have some wearables that are continuously sending their biomedical signs such as heart rate (pulse), stress levels, body temperature, among others, to a platform data management system hosted in the cloud. The use of wearable devices during long periods of time produces a huge volume of data. That data can be used by algorithms to detect long term patterns and notice when the pattern changes, raising alerts to healthcare entities. To improve diagnostics and the quality of the data, all exams done by patients such as blood collection, ultrasounds, among others, should be stored. Hospital entity staff have the responsibility to analyze the received alert and to make a decision. If they consider a possible problem, they should notify the patient to

A Product and Service Concept Proposal

11

come to the hospital entity to do more precise exams and consequently verify if, in fact, something is wrong with the patient. Briefly, the aim of this system is to transform data into information and that information into knowledge which may be used as a support decision system by the medical community. The ecosystem explained before is illustrated in Fig. 2.

Fig. 2. Conceptual architecture of a healthcare system

Regarding the availability to pay for this kind of service, around 49% of all participants are undecided, and 30% will not pay, so we believe it would be necessary to have more concrete features that would potentially improve quality of life, for these people to change their position. This kind of system can have several drawbacks. As stated before, the collected data is stored in a cloud platform, which normally replicates the data, to better protect it, which can lead to considerable expenses, because it is necessary to have a big infrastructure to store this data. Due to the sensitivity associated with this kind of data, it is not possible to use existing cloud service providers, so it would be necessary to build one from scratch. The use of that data for other purposes is also a major concern because some companies could take advantage of this information, for example by selling it or customizing product advertisements. We further noticed that participants who know about GDPR tend not to feel comfortable about sharing their data. Despite seeming contradictory, this can be explained by their knowledge about related risks. Despite the disadvantages, the information could also be relevant for research purposes while considering the fact that participants must be anonymized.

6 Conclusion In this paper, we start with a researching of the literature about devices, data processing and existing healthcare systems. We had the opportunity to perform a group interview with Altice Labs which gave us insights about the challenges that they had before such

12

L. Fonseca et al.

a prediction of medical conditions with only biometrical data was possible. In their perspective, medical staff would not accept the possibility of full prediction by the system and, on the other hand, they defend that these systems should only be for decision support. Furthermore, from the questionnaire, we can conclude that most people would be willing to use healthcare systems. Based on these results, we discussed and proposed an approach that would be able to solve some of the existing issues, although we also state that some disadvantages are not easily overcome. In conclusion, we note that there exist a lot of solutions related to healthcare, especially for the elderly. However, there are still some problems, such as the elderlies’ unawareness of wearable devices and services related to them. According to our research, we perceive that this solution could be applied to people of all ages, taking advantage of the popularity of wearables since they are cheaper and more robust than ever before. In general, people are receptive to this kind of system, but it is evident that there are concerns about privacy issues. In addition to the advantages for citizens, we believe that countries’ health systems would reduce costs and improve health service qualities. Furthermore, the scientific community may benefit from this solution using data gathered for researchers in various areas.

7 Contribution and Suggestions for Future Research The solution set forth is an incremental contribution, in terms of the innovation involved – and when comparing our solution to what already exists in the market. This is due to the fact that it adds an analytical component to the data on society in general (instead of being an individual solution) and the focus commences to be on all people, instead of only being on the elderly. In order for the solution to be a radical innovation it would have to have, besides the monitoring of people, some type of reaction, as a consequence to a malady detected. In sum, if something bad was to happen to someone (for example, if someone fainted, or had an epileptic seizure) he or she would be helped automatically – by a drone or by a system alerting the health services, which would then act accordingly. It would thus be interesting to study what type of automatic reactions could occur with existing technology, while also thinking of new systems to be put in place. Regarding the Professional Performance Framework, designed by the Medical Board of Australia, our solution goes according to the initiatives stated therein, such as: Guidance to support practitioners - regularly updated professional standards that support good medical practice and collaborations to foster a culture of medicine that is focused on patient safety [21]. Acknowledgements. We would like to thank Telma Mota and Ricardo Machado from Altice Labs for having agreed to be interviewed and for all the information provided during the group interview. For the dissemination of the questionnaire, we would like to thank Patrícia Gonçalves and Graça Ferraz, from the Recesinhos Social Center, for having helped in the gathering of data from the elderly; and Ana Araújo for having helped in the gathering of data from the younger age groups.

A Product and Service Concept Proposal

13

References 1. World Health Organization: mHealth: New horizons for health through mobile technologies. Observatory 3, 66–71 (2011). http://www.webcitation.org/63mBxLED9 2. Machine Learning for Healthcare. https://www.mlforhc.org/. Accessed 19 Oct 2019 3. AlticeLabs: SmartAL – Smart Assisted Living. http://www.alticelabs.com/site/smartal/. Accessed 16 Oct 2019 4. Wearable Devices: Wearable Technology and Wearable Devices: Everything You Need to Know. http://www.wearabledevices.com/what-is-a-wearable-device/. Accessed 17 Oct 2019 5. Hänsel, K.: Wearable sensing approaches for stress recognition in everyday life. In: Proceedings of the 2017 Workshop on MobiSys 2017 Ph.D. Forum - Ph.D. Forum 2017, pp. 1–2 (2017). http://dl.acm.org/citation.cfm?doid=3086467.3086470. Accessed 19 Oct 2019 6. Qiu, H., Wang, X., Xie, F.: A survey on smart wearables in the application of fitness. In: 2017 IEEE 15th International Conference on Dependable, Autonomic and Secure Computing, 15th International Conference on Pervasive Intelligence and Computing, 3rd International Conference on Big Data Intelligence and Computing and Cyber Science and Technology Congress (DASC/PiCom/DataCom/CyberSciTech), pp. 303–307 (2017). http:// ieeexplore.ieee.org/document/8328407/. Accessed 19 Oct 2019 7. Berglund, M.E., Duvall, J., Dunne, L.E.: A survey of the historical scope and current trends of wearable technology applications. In: Proceedings of the 2016 ACM International Symposium on Wearable Computers - ISWC 2016, pp. 40–43 (2016). http://dl.acm.org/ citation.cfm?doid=2971763.2971796. Accessed 19 Oct 2019 8. Sultan, N.: Reflective thoughts on the potential and challenges of wearable technology for healthcare provision and medical education. Int. J. Inf. Manag. 35(5), 521–526 (2015). https://www.sciencedirect.com/science/article/pii/S0268401215000468. Accessed 19 Oct 2019 9. Fletcher, R.R., Poh, M.-Z., Eydgahi, H.: Wearable sensors: opportunities and challenges for low-cost health care. In: 2010 Annual International Conference of the IEEE Engineering in Medicine and Biology, pp. 1763–1766 (2010). http://ieeexplore.ieee.org/document/5626734/. Accessed 19 Oct 2019 10. Kolasinska, A., Quadrio, G., Gaggi, O., Palazzi, C.E.: Technology and aging: users’ preferences in wearable sensor networks. In: Proceedings of the 4th EAI International Conference on Smart Objects and Technologies for Social Good - Goodtechs 2018, pp. 77– 81 (2018). http://dl.acm.org/citation.cfm?doid=3284869.3284884. Accessed 19 Oct 2019 11. Di Rienzo, M., Rizzo, F., Parati, G., Brambilla, G., Ferratini, M., Castiglioni, P.: MagIC system: a new textile-based wearable device for biological signal monitoring. Applicability in daily life and clinical setting. In: Annual International Conference of the IEEE Engineering in Medicine and Biology - Proceedings, vol. 7, pp. 7167–7169 (2005) 12. Innovatemedtec: This smart bra can detect breast cancer much earlier than existing screening tests - innovatemedtec content library. https://innovatemedtec.com/content/smart-bra?fbclid= IwAR1IKHXd7ih9JnwVDQvL6wkUCpltayE-nsIHkv_9zADrYXhyI9bzKz56D5Q. Accessed 19 Oct 2019 13. Sultan, N.: Making use of cloud computing for healthcare provision: opportunities and challenges. Int. J. Inf. Manag. 34(2), 177–184 (2014). https://www.sciencedirect.com/ science/article/pii/S0268401213001680. Accessed 19 Oct 2019 14. Kumar, S., et al.: Mobile health technology evaluation: the mHealth evidence workshop. Am. J. Prev. Med. 45(2), 228–236 (2013). https://www.sciencedirect.com/science/article/pii/ S0749379713002778. Accessed 13 Oct 2019

14

L. Fonseca et al.

15. Géron, A.: Hands-on machine learning with Scikit-Learn and TensorFlow: concepts, tools, and techniques to build intelligent systems. https://www.oreilly.com/library/view/hands-onmachine-learning/9781492032632/. Accessed 15 Oct 2019 16. Qi, J., Yang, P., Newcombe, L., Peng, X., Yang, Y., Zhao, Z.: An overview of data fusion techniques for Internet of Things enabled physical activity recognition and measure. Inf. Fusion 55, 269–280 (2020). https://www.sciencedirect.com/science/article/pii/S1566253 519302258?via%3Dihub. Accessed 05 Oct 2019 17. Apple: Efetuar um ECG com a app ECG no Apple Watch Series 4ou posterior - Suporte Apple. https://support.apple.com/pt-pt/HT208955. Accessed 14 Oct 2019 18. Apple: Apple Watch Series 5 - Apple. https://www.apple.com/apple-watch-series-5/. Accessed 14 Oct 2019 19. IMTInnovation: IMT Innovation Digital Health Incubator Wearable Technology to Minimise Injury Risk. https://imtinnovation.com/2018/11/10/wearable-technology-to-mini mise-injury-risk/. Accessed 19 Oct 2019 20. Medtronic: MiniMed 670G Insulin Pump System—Medtronic Diabetes. https://www. medtronicdiabetes.com/products/minimed-670g-insulin-pump-system. Accessed 19 Oct 2019 21. Medical Board of Australia: Building a professional performance framework. https://www. racgp.org.au/getmedia/d810a609-c344-4e97-b615-1f83b6e504eb/Medical-Board-Report-Bu ilding-a-professional-performance-framework.PDF.aspx. Accessed 04 Jan 2020

Artificial Neural Networks Interpretation Using LIME for Breast Cancer Diagnosis Hajar Hakkoum1 , Ali Idri1,2(B) , and Ibtissam Abnane1 1

Software Project Management Research Team, ENSIAS, Mohammed V University, Rabat, Morocco [email protected] 2 Complex Systems Engineering and Human Systems, Mohammed VI Polytechnic University, Ben Guerir, Morocco

Abstract. Breast Cancer (BC) is the most common type of cancer among women. Thankfully early detection and treatment improvements helped decrease its number of deaths. Data Mining techniques (DM), which discover hidden and potentially useful patterns from data, particularly for breast cancer diagnosis, are witnessing a new era, where the main objective is no longer replacing humans or just assisting them in their tasks but enhancing and augmenting their capabilities and this is where interpretability comes into play. This paper aims to investigate the Local Interpretable Model-agnostic Explanations (LIME) technique to interpret a Multilayer perceptron (MLP) trained on the Wisconsin Original Data-set. The results show that LIME explanations are a sort of real-time interpretation that helps understanding how the constructed neural network “thinks” and thus can increase trust and help oncologists, as the domain experts, learn new patterns.

Keywords: Interpretability

1

· Breast Cancer · Diagnosis · LIME

Introduction

Breast cancer is a phenotypically diverse population of breast cancer cells [1]. It is affecting about 1 out of 8 women over the world and is considered as the leading cause of cancer death among women at the age between 40 and 59 [2]. Its causes are not yet fully known, although age, genetic risk, smoking, eating unhealthy food, overweight, late menopause, and late age at first childbirth were identified as risk factors [3]. Several Data Mining techniques, whether relying on artificial intelligence or statistics, were used to help discover new patterns and new high-level information from historical databases of patients that had Breast Cancer [3–6]. It is thus a powerful tool to analyze and deal with BC challenges, capable of reducing the deaths rate caused by this disease [3]. c The Editor(s) (if applicable) and The Author(s), under exclusive license  to Springer Nature Switzerland AG 2020 ´ Rocha et al. (Eds.): WorldCIST 2020, AISC 1161, pp. 15–24, 2020. A. https://doi.org/10.1007/978-3-030-45697-9_2

16

H. Hakkoum et al.

A systematic mapping study by Idri et al. [3] was conducted on 403 articles treating DM techniques in BC showed that most of the articles focused on the diagnosis task using DM techniques with 78.63%, 7.63%, 9.16% and 4.58% for classification, regression, clustering, and association respectively. They also showed that the use of black-box models such as Support Vector Machines or Neural Nets was very high followed by Decision Trees, which are explainable models and highlighted that the use of Neural Nets has decreased over time which is, very probably, due to their inexplicable behavior [3]. Interpretability is one of the most common reasons limiting artificial neural networks, and black-box in general, to be accepted and used in critical domains such as the medical one. Indeed, healthcare offers more challenges to Machine learning (ML) by being more demanding for interpretability [7]. Model interpretability is thus often chosen over its accuracy. Therefore, understanding black-box models can help assist and augment the provision of better care while doctors remain integral to their role. It could also improve human performance, extract insights, gain new knowledge about the disease which may be used to generate hypotheses [7]. Skater is an Oracle unified framework for Model Interpretation [8] that used LIME explanations to interpret a basic MLP, with 100 hidden nodes, trained on the Wisconsin (Diagnostic) Database that has 30 attributes. This paper aims to apply and evaluate the local interpretability technique LIME on an MLP. The main contribution is the LIME interpretation of the best of two neural networks: a basic MLP and a Deep MLP (four layers) trained on the Breast Cancer Wisconsin (Original) data-set that has 9 attributes. The rest of this paper is structured as follows: Sect. 2 presents some important concepts related to this paper. Section 3 presents some related work dealing with the use of interpretation techniques. Section 4 describes the database as well as the performance measures used to select the best performing model. Section 5 presents the experimental design followed in this empirical evaluation. Section 6 discusses the obtained results. The threats to the validity of this paper are given in Sect. 7. Section 8 presents conclusions and future work.

2

Background

This section presents an overview of the feed-forward neural networks that were constructed and evaluated in our experiments. As well as the interpretability techniques that were applied to their best variants. 2.1

Artificial Neural Networks: MLP

Artificial neural networks are a set of algorithms designed to mimic the brainbehavior [9]. There are multiple types of neural networks, the most basic type is the feed-forward where information travels in one direction from input to output. A popular example of this type is MLP, it is composed of an input layer to receive the signal, an output layer that makes a decision, and in between, an

ANN Interpretation Using LIME For BC Diagnosis

17

arbitrary number of hidden layers. Each layer is a set of neurons interconnected with the other layers by weights. It represents a non-linear mapping between an input vector and an output vector [10]. In the forward pass, the information travels from the input layer through the hidden layers to the output layer, and the decision of the output layer is compared against the ground truth labels. In the backward pass, using backpropagation, partial derivatives of the error function, the various weights and biases are backpropagated through the MLP [11]. That act of differentiation gives us a gradient, or a landscape of error, along which the parameters may be adjusted as they move the MLP one step closer to the error minimum. 2.2

Local Interpretability

Interpretability can be defined as the degree to which a human can understand the cause of a decision [12] or the degree to which a human can consistently predict the model’s result [13]. The more faithful and intelligible a model’s explainability is, the easier it is for someone to trust it and comprehend why certain decisions or predictions have been made [14]. For instance, doctors can not trust blindly a black-box prediction, not understanding how it provides the results. There are two options to make ML interpretable [15]: Using interpretable models (white-boxes), such as linear models or decision trees and here data scientists are faced with the accuracy-interpretability trade-off, or using interpretation tools. Such tools can either explain model behavior across all data instances or individual predictions. Thus two types of interpretability techniques can be identified: Global (i.e. They consider all instances and give a statement about the global relationship of a feature with the predicted outcome) and Local (i.e. They explain the conditional interaction between the predictions and the predictors/attributes concerning a single prediction). This study deals with local interpretation techniques, in particular, LIME. In 2016, Ribeiro et al. [14] explained the predictions of any classifier by learning an interpretable model (linear models) locally around a prediction. They called this explanation system LIME, Local Interpretable Model-agnostic Explanations, where model-agnostic refers to the system treating the model as a black box and not needing any prior information about its architecture. What LIME does concretely is it tests how the predictions change when variations of the data are given to the model. It perturbs the data-set and gets the black box predictions for these new points. On this new data-set, LIME then trains a weighted, interpretable model like a linear classifier. The linear classifier will then be the learned explanation that is locally (but not globally) faithful. This kind of accuracy is also called local fidelity [14]. Mathematically, local surrogate models with interpretability constraint can be expressed as follows: explanation(x) = arg min L(f, g, πx ) + Ω(g) g∈G

(1)

The explanation model for an instance x is the model g (linear regression) that minimizes loss L (mean squared error), which measures how close the

18

H. Hakkoum et al.

explanation is to the prediction of the original model f, while the model complexity Ω(g) is kept low (prefer fewer features). G is the family of possible explanations, for example, all possible linear regression models [15], and the proximity measure πx defines how large the neighborhood around instance x is that we consider for the explanation. In practice, LIME only optimizes the loss part and the user has to determine the complexity by selecting the maximum number of features that the linear regression model may use. Although the explanation of a single prediction provides some understanding of the model, it is not sufficient to evaluate and assess trust in the model as a whole. Therefore, Ribeiro et al. [14] proposed to explain a judiciously picked set of individual instances using the SubmodularPick, an algorithm that helps to select a representative set to simulate a global understanding of the model. For example, if an explanation A was explained relying on two features x1 and x2 , there is no need to show the end-user another explanation that focused on the same features x1 and x2 .

3

Related Work

A variety of work has been done on ML models’ interpretation. Since the ML community has noticed the importance of explaining which features a model did or did not take into account more than which parameter increased its accuracy [16]. Interpretability aims increasing model trustworthiness so that it can be used to make high-stakes decisions, especially in domains such as the medical one [17]. ML explainability is thus a major concern since it has the power to break even the models with the highest accuracy. When it comes to using a deployed model to make decisions, end-users often ask the almighty question: “Why should I trust it?”. In 2002, Idri et al. [11] asked a slightly different question: “Can neural networks be easily interpreted in software cost estimation?” and they used the method of [18] to map MLP to a fuzzy rule-based system which expresses the information encoded in the architecture of the network and which can be easily interpreted. Although they found the i-or operator that connected rules inappropriate. LIME framework was applied in different domains such as medicine and finance [16]. Particularly, it was applied by the Skater team [8] in different fields one of them was breast cancer diagnosis where they used the Breast Cancer Wisconsin (Diagnostic) Database, available on the UCI repository, to train four models including an MLP. They discussed how sensitive each classifier is to the attributes to show how interpretation techniques can help with model understanding for model selection. Puri et al. [16] proposed an approach that can be thought of as an extension to LIME. It learns if-then rules that represent the global behavior of a model that solves a classification problem. Then validated their approach using different data-sets, particularly the Wisconsin Breast Cancer data-set, to train a random

ANN Interpretation Using LIME For BC Diagnosis

19

forest classifier where they got an accuracy of 98%. After running their technique, they compared the model predictions to the predictions of the resulted if-then rules, and they computed a metric they introduced: Imitation@K where K is the number of rules. Their approach generated rules able to imitate the model behavior for a large fraction of the data-set.

4

Database Description and Performance Criteria

In this section, the database used to train the models is described. A set of performance measures is then defined to help select the best classifier. 4.1

WISCONSIN Database

In this work, the Breast Cancer Wisconsin (Original) data-set was used [19]. The data-set is available on the UCI repository and has 9 attributes which values are in the bucket 1.10. It has 458 benign and 241 malignant cases which shows an imbalance problem. The imbalanced distribution of classes constitutes a challenge for standard learning algorithms because they are biased towards the majority classes. Different resampling methods modify the original class distributions to tackle the imbalance issue. In particular, Synthetic Minority Oversampling Technique (SMOTE) algorithm [20], which was used in this study to balance the data-set. Also, the data-set had 16 missing values which were removed from the set before resampling. 4.2

Performance

The best model isn’t necessarily the one with higher accuracy. Therefore, multiple performance classification measures were used to select the best MLP architecture: – Accuracy: The most intuitive performance measure and it is simply a ratio of correctly predicted observation to the total observations. – Precision: The ratio of correctly predicted positive observations to the total predicted positive observations. – Recall (Sensitivity): The ratio of correctly predicted positive observations to the all observations in actual class ‘yes’. – F1-Score: The weighted average of Precision and Recall. Therefore, this score takes both false positives and false negatives into account. Accuracy =

TP + TN TP + FP + FN + TN

(2)

TP TP + FP

(3)

P recision = Recall =

TP + TN TP + FN

(4)

20

H. Hakkoum et al.

F 1 − Score = 2 ∗

5

P recision ∗ Recall P recision + Recall

(5)

Experimental Design

This Section presents the experimental process followed in this empirical evaluation. It consists of choosing the parameters for models construction, selecting the best performing model and fixing the number of LIME explanations to discuss. 5.1

Models Construction

Two MLP architectures were adopted and since our aim was interpretability and not performance, hyperparameters were chosen randomly. We constructed a basic MLP with 200 hidden nodes and a deep MLP where another hidden layer of 128 nodes was added. To make training faster, the non-linearity activation function ReLU (Rectified Linear Units) was used, since it makes training several times faster than with other function such as tanH [21]. As to the output layer, we opted for a softmax activation function and thus we had two neurons since we have two classes (Malignant/Benign). The dropout technique was used twice in the deep MLP to avoid over-fitting, in the first hidden layer with a 0.5 probability and in the second layer with a probability of 0.8. The neurons which are dropped out in this way do not contribute to the forward pass and do not participate in back-propagation. So every time an input is presented, the neural network samples a different architecture, but all these architectures share weights. This technique reduces complex coadaptations of neurons since a neuron can not rely on the presence of particular other neurons [21]. After the training of 500 epochs (training cycles) and a batch size of 128, the Accuracy (Eq. 2) and F1-Score (Eq. 5) were voters to choose the best MLP. Here BordaCount [22], a voting system was used. When each individual of a group ranks m candidates, Borda has each assign 0 to its last-ranked candidate, 1 to the second-to-the-last-ranked, until it assigns n − 1 to its top-ranked, and then, for each of the candidates, sums over those numbers to determine the group ranking [23]. K-folds Cross-validation is a validation technique that ensures that the model is low on bias and that it can work well for the real unseen data. The data is divided into k subsets, where one of the k subsets is used as the test/validation set and the other (k − 1) subsets form the training set [24]. After k iterations, the error and performance metrics are computed by getting the average over all k folds. As a general rule and empirical evidence, 5 or 10 is generally preferred for k [24]. For the experiments of the present study 10 folds cross validation was performed.

ANN Interpretation Using LIME For BC Diagnosis

5.2

21

Interpretability Techniques

As cross validation’s purpose is model checking and not model building, after selecting the best model to interpret in terms of Accuracy and F1-Score, the chosen model was fitted with a normal split, feeding the maximum amount to the training leaving 0.05 for the testing. This phase aimed to interpret the selected model locally and define on which features it relies on the most and how each feature affects the final decision. A four instances set was chosen using the SubmodularPick algorithm provided with LIME. LIME explanations were then computed for the four instances.

6

Results and Discussion

This section shows the results of the empirical evaluations of this study and discusses the LIME provided local explanations. 6.1

Models Construction

After 10 folds cross validation training with 500 epochs, two models were constructed: – MLP: A basic three layers MLP with 200 nodes/neurons with ReLu activation function in the hidden layer, and two nodes for the output with a Softmax activation function. – Deep MLP (DMLP): Another hidden layer of 128 nodes was added to the basic MLP. The dropout technique was used for both hidden layers. Table 1 shows the accuracy and F1-Score values of the two models MLP and DMLP. Note that F1-Score is the weighted average of precision and recall. Table 1. Models performances Model Accuracy F1-Score MLP

71.20%

60.24%

DMLP 92.95%

98.74%

DMLP did better, which was expected since the added layer helped in doing more computations and recognizing more useful patterns to distinguish between the two classes [25]. Both, the accuracy and F1-Score were voters for the two MLP candidates to choose the best MLP. Therefore, the BordaCount voting system was used to select the best MLP architecture which turned out to be the DMLP.

22

H. Hakkoum et al.

Fig. 1. LIME explanations of the first four instances by SubmodularPick

6.2

Interpretability Results

LIME was used for the local interpretability of the best MLP model. It focuses on training local surrogate models (interpretable models) to explain individual predictions and so the decider can understand why the model predicted a certain class for a particular instance. A set of four instances was thus chosen using the SubmodularPick algorithm for the best model, then the plotted explanations are further interpreted to understand how the best model uses the features. In Fig. 1, we notice in the first upper explanations how the cellSizeUniformity, the barNuclei as well as the cellShapeUniformity and normalNucleoil switched the prediction when they all increased which shows that instances tend to be classified benign more when the uniformities are very low. In the third explanation, the barNuclei being in the 2.9 bucket and the uniformities in 3.6 had a huge impact on classifying the instance as malignant and although the blandChromatin was higher than 6 which voted for the benignity of the instance, the model’s decision was affected more by the first three features. In the fourth explanation, it was again the uniformalities features as well as the barNuclei and normalNucleoil that affected the prediction the most but differently, some “voted” for malignant such as normalNucleoil and cellShapeUniformity since the first was high and the second was average, while the others, barNuclei, and cellSizeUniformity “voted” for benign since they were very low.

ANN Interpretation Using LIME For BC Diagnosis

7

23

Threats to Validity

In this work, parameter tuning was ignored since we focused on interpretability, so no parameter tuning technique was used, which can be interesting and might give better results [5]. The medical domain is a very critical one, using one data-set is not enough to select the best classifier nor to define its trustworthiness. Moreover, to check the reliability of explanations, using other interpretability techniques than LIME would also be helpful to understand the model. The interpretability techniques were applied to an MLP. Constructing and interpreting other types of black-box models such as Support Vector Machine will help understand those techniques more and generalize their effectiveness in interpretation.

8

Conclusion and Future Work

The present study constructed and evaluated two MLPs where the deeper one did better, from which we explained four instances using LIME. LIME interpretability framework proved to be highly useful in understanding the model’s behavior and increasing trustworthiness which can help domain experts understand or discover hidden patterns. We noticed that the model focused more on the uniformities and barNuclei features to decide on the prediction. Ongoing work will focus on checking other interpretability techniques whether global or local since it will be interesting to enhance the trustworthiness of models as well as the understanding of how those techniques work. If a model is not highly interpretable, the domain might not be legally permitted to use its insights to make changes to processes. ML explainability is defying black box non-transparency to gain both, high accuracy and high interpretability.

References 1. Al-Hajj, M., Wicha, M.S., Benito-Hernandez, A., Morrison, S.J., Clarke, M.F.: Prospective identification of tumorigenic breast cancer cells. Proc. Nat. Acad. Sci. 100(11), 6890 (2003). Correction to 100(7):3983 2. Solanki, K.: Application of data mining techniques in healthcare data, vol. 148, no. 2, p. 1622 (2016) 3. Idri, A., Chlioui, I., El Ouassif, B.: A systematic map of data analytics in breast cancer. In: ACM International Conference Proceeding Series. Association for Computing Machinery (2018) 4. Hosni, M., Abnane, I., Idri, A., de Gea, J.M.C., Fernandez Aleman, J.L.: Reviewing ensemble classification methods in breast cancer. Comput. Methods Programs Biomed. 177, 89–112 (2019) 5. Idri, A., Hosni, M., Abnane, I., de Gea, J.M.C., Fernandez Aleman, J.L.: Impact of parameter tuning on machine learning based breast cancer classification. In: Advances in Intelligent Systems and Computing, vol. 932, pp. 115–125. Springer (2019)

24

H. Hakkoum et al.

6. Chlioui, I., Idri, A., Abnane, I., de Gea, J.M.C., Fernandez Aleman, J.L.:. Breast cancer classification with missing data imputation. In: Advances in Intelligent Systems and Computing, vol. 932, pp. 13–23. Springer (2019) 7. Aurangzeb, A.M., Eckert, C., Teredesai, A.: Interpretable machine learning in healthcare. In: Proceedings of the 2018 ACM International Conference on Bioinformatics, Computational Biology, and Health Informatics, BCB 2018, pp. 559–560. ACM Press, New York (2018) 8. Oracle’s unified framework for Model Interpretation. https://github.com/oracle/ Skater 9. Thomas, A.: An introduction to neural networks for beginners. Technical report in Adventures in Machine Learning (2017) 10. Gardner, M.W., Dorling, S.R.: Artificial neural networks (the multilayer perceptron) - a review of applications in the atmospheric sciences. Atmos. Environ. 32(14– 15), 2627–2636 (1998) 11. Idri, A., Khoshgoftaar, T., Abran, A.: Can neural networks be easily interpreted in software cost estimation? In: 2002 IEEE World Congress on Computational Intelligence. IEEE International Conference on Fuzzy Systems, FUZZ-IEEE 2002. Proceedings (Cat. No.02CH37291), vol. 2, pp. 1162–1167. IEEE (2002) 12. Miller, T.: Explanation in artificial intelligence: insights from the social sciences. Artif. Intell. J. 267, 1–38 (2017) 13. Kim, B., Khanna, R., Koyejo, O.: Examples are not enough, learn to criticize! Criticism for interpretability. In: Advances in Neural Information Processing Systems (NIPS 2016), vol. 29 (2016) 14. Ribeiro, M.T., Singh, S., Guestrin, C.: “Why should I trust you?” Explaining the predictions of any classifier. In: Proceedings of the ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 13–17 August 2016, pp. 1135–1144. Association for Computing Machinery (2016) 15. Molnar, C.: Interpretable Machine Learning. A Guide for Making Black Box Models Explainable (2018). https://christophm.github.io/book/ 16. Puri, N., Gupta, P., Agarwal, P., Verma, S., Krishnamurthy, B.: MAGIX: model agnostic globally interpretable explanations (arXiv) (2017) 17. Lazzeri, F.: Automated and Interpretable Machine Learning - Microsoft Azure Medium (2019) 18. Benitez, J.M., Castro, J.L., Requena, I.: Are artificial neural networks black boxes? IEEE Trans. Neural Netw. 8(5), 1156–1164 (1997) 19. https://archive.ics.uci.edu/ml/datasets/breast+cancer+wisconsin+(original) 20. Chawla, N., Bowyer, K., Hall, L., Kegelmeyer, W.: SMOTE: synthetic minority over-sampling technique. J. Artif. Intell. Res. 16, 321–357 (2002) 21. Krizhevsky, A., Sutskever, I., Hinton, G.E.: ImageNet classification with deep convolutional neural networks. In: Advances in Neural Information Processing Systems, vol. 25, no. 2 (2012) 22. de Borda, J.C.: M´emoire sur les ´elections au scrutin, M´emoire de l’Acad´emie Royale. Histoire de l’Acad´emie des Sciences, Paris, pp. 657–665 (1781) 23. Risse, M.: Why the count de Borda cannot beat the Marquis de Condorcet. Soc. Choice Welfare 25(1), 95–113 (2005) 24. Gupta, P.: Cross-Validation in Machine Learning - Towards Data Science (2017) 25. Reed, R., MarksII, R.J.: Neural Smithing: Supervised Learning in Feedforward Artificial Neural Networks, p. 38 (1999)

Energy Efficiency and Usability of Web-Based Personal Health Records Jos´e Alberto Garc´ıa-Bern´a1(B) , Sofia Ouhbi2 , Jos´e Luis Fern´ andez-Alem´an1 , 1 as1 Juan Manuel Carrillo-de-Gea , and Joaqu´ın Nicol´ 1

Department of Informatics and Systems, Faculty of Computer Science, University of Murcia, Murcia, Spain {josealberto.garcia1,aleman,jmcdg1,jnr}@um.es 2 Department of Computer Science and Software Engineering, CIT, United Arab Emirates University, Al Ain, UAE [email protected]

Abstract. Usability is a critical aspect in the adoption of e-health applications. However, its impact on energy consumption has not been thoroughly studied in e-health domain. The aim of this paper is to investigate the relationship between energy efficiency and usability in the context of personal health records (PHRs). A total of 5 web-based PHRs out of 19 were selected for this study and the energy consumption of these PHRs was assessed when performing 20 tasks. The results showed that the PHRs with best practices of usability consumed more energy than the others. Based on the findings of this study, recommendations for practitioners on how to trade-off between usability and energy efficiency were proposed. Keywords: Energy efficiency · Usability · Software sustainability Green software · Personal health records · e-health

1

·

Introduction

Energy efficiency and energy consumption are considered as fundamental sustainability characteristics in architectural design [29]. Whilst energy consumption is the amount of power used to operate a technology, energy efficiency refers to the use of as little energy as possible in a particular system. Software sustainability is attracting attention of researchers [3,27,28]. Sustainability in e-health technologies is an important challenge for the healthcare industry [20]. Energy efficiency will need improving due to the large amount of health data that will This research is part of the BIZDEVOPS-GLOBAL-UMU (RTI2018-098309-B-C33) project, and the Network of Excellence in Software Quality and Sustainability (TIN2017-90689-REDT). Both projects are supported by the Spanish Ministry of Science, Innovation and Universities and the European Regional Development Fund (ERDF). c The Editor(s) (if applicable) and The Author(s), under exclusive license  to Springer Nature Switzerland AG 2020 ´ Rocha et al. (Eds.): WorldCIST 2020, AISC 1161, pp. 25–35, 2020. A. https://doi.org/10.1007/978-3-030-45697-9_3

26

J. A. Garc´ıa-Bern´ a et al.

be accumulated due to advancement in Information and Communications Technology (ICT) [5]. Personal health records (PHRs) are a promising solution to the sustainability gap in the public healthcare systems. There is a rapid growth in literature related to PHRs in spite of the fact that this technology is in its beginnings [15]. PHRs are tools that allow people to store and manage their health data [10]. PHRs are managed by the users themselves and have the potential to empower patients [11]. Providing preventive measures and self-treatment instructions to people could lead to decrease the demand of healthcare [22]. The usability of PHRs plays an important role for a good acceptance among the users. Poor usability and functionality have been proven to result in a low utility, which affects the enrollment’s rates by both patients and clinicians alike [6]. Patients have a positive attitude towards sharing medical information as they consider it a way to promote a better healthcare service [14], but they are usually concerned with the usability and privacy issues when accessing e-health applications [25]. Usability helps healthcare organizations in the customization of e-health interventions [17]. To this end, several approaches to evaluate usability in e-health applications were used, such as: end-user surveys [9], and thinkaloud protocol [30]. Usability requirements were proposed in a reusable requirements catalogue for sustainable connected health applications [21]. Although some studies about energy efficiency and usability in e-health can be found, the relationship of these two factors in PHRs has not been studied before to the best of our knowledge. Motivated by the lack of research that evaluates the relationship of these two important factors, this paper studied the relation between energy consumption and usability gathered from a set of tasks performed in a group of web-based PHRs. The remainder of this paper is organized as follows: Sect. 2 presents the materials and method used in this study, Sect. 3 presents the results, Sect. 4 presents recommendations based on the results, and Sect. 5 concludes with principal findings and future works.

2 2.1

Materials and Methods PHRs Under Study

The PHRs were selected from a previous study [10]. The method proposed by the Preferred Reporting Items for Systematic reviews and Meta-Analysis (PRISMA) group [19] was employed for accurateness and impartiality of the selection. The search was performed at myPHR.com and in ACM Digital Library, IEEE Digital Library, Med-line and ScienceDirect. Web-based format was the inclusion criterion (IC) with which a total of 19 PHRs were collected initially (see Fig. 1). A refinement of the results was performed with the following exclusion criteria: nonavailable PHRs (EC1), non-free PHRs (EC2), registration not possible (EC3), mal-functioning (EC4), only available in USA (EC5), and low-popularity PHRs (EC6). EC6 was applied through Alexa website (alexa.com/siteinfo), which is a sorting online tool to verify visits in web portals.

Energy Efficiency and Usability of Web-Based Personal Health Records

27

Fig. 1. PRISMA flow chart

From the results of 19 PHRs selected, those that met ECs were discarded. A first rejection of PHRs was carried out. HealthyCircles, Telemedical, Dr. I-Net, MedsFile.com, ZebraHealth, EMRySTICK and Dlife were dropped due to EC1, myMediConnect and Juniper Health because of EC2, RememberItNow! by EC3, WebMD HealthManager by EC4, and PatientPower by EC5. Finally another round to discard more PHRs was done. In this case, My Health Folders and My Doclopedia fulfilled EC6. The Alexa ranking mark exposed a very low popularity of these portals—the higher the mark, the less popular a website is—. The Alexa mark in some cases was not available which lead to a low popularity of the portal consideration. Finally, the PHRs selected were HealthVet, PatientsLikeMe, HealthVault, Health Companion, and NoMoreClipBoard. All these PHRs covered as many functionalities as possible provided by this type of tools. The use of PHRs was analyzed by carrying out a set of identified tasks with common needs detected for a better interaction among several usage profiles [4]. The recommendations for the development of a PHR from the American Health Information Management Association [2] were also taken into account to propose the tasks to be performed in the PHRs. Table 1 depicts a list of the 20 PHR common tasks identified.

28

J. A. Garc´ıa-Bern´ a et al. Table 1. Typical tasks identified in a PHR TASK TASK TASK TASK TASK TASK TASK TASK TASK TASK TASK TASK TASK TASK TASK TASK TASK TASK TASK TASK

2.2

01: 02: 03: 04: 05: 06: 07: 08: 09: 10: 11: 12: 13: 14: 15: 16: 17: 18: 19: 20:

Registration System access Add profile View profile Manage permissions to 3rd parties Add family history Add medication Add new allergy Add vaccine Add disease View medications Print report View glucose evolution Search for information about conditions Export health info Schedule appointments and medication reminder Send suggestion/contact See privacy policy Exit Forgotten password

Power Measurements Procurement

The power expenses during the performance of the tasks were measured with the Energy Efficient Tester (EET) [18]. This device is provided with sensors capable to measure the instant power consumption of the processor, hard disk drive, graphic card, monitor and total power supplied to a host machine. The experiment was carried out using EET connected to a thin film transistor-liquid crystal display (TFT-LCD) monitor Philips 170S6FS and a PC provided with a GigaByte GA-8I945P-G motherboard, an Intel Pentium D @ 3.0 Ghz processor, a set of 2 modules of 1 GB DDR2 @ 533 MHz RAM memory, a Samsung SP2004C 200 GB 7500 rpm hard disk drive, a Nvidia GeForce GTS 8600 graphics card, and a Aopen Z350-08FC 350 W power supply. Data were checked to ensure the absence of outliers. To this end, each task was carried out five times in order to average the results and to smooth any peak red of power consumption that may had occurred. The operating system installed in the PC was Microsoft Windows 7 Professional, which allowed to disable background running processes and reduce the resources required by the operating system.

Energy Efficiency and Usability of Web-Based Personal Health Records

2.3

29

Usability Assessment

In this paper, Usability is defined according to ISO/IEC 25010-2011 standard as “the degree to which a product or system can be used by specified users to achieve specified goals with effectiveness, efficiency and satisfaction in a specified context of use.” Usability is defined into six characteristics [1]: – Appropriateness recognizability: degree to which users can recognise whether a product or system is appropriate for their needs. – Learnability: degree to which a product or system can be used by specified goals of learning to use the product with effectiveness, efficiency, freedom from risk and satisfaction in a specified context of use. – Operability: degree to which a product or system has attributes that make it easy to operate and control. – User error protection: degree to which a system protects users against making errors. – User interface aesthetics: degree to which a user interface enables pleasing and satisfying interaction for the user. – Accessibility: degree to which a product or system can be used by people with the widest range of characteristics and capabilities to achieve a specified goal in a specified context of use. Because we are interested in energy efficiency and its relationship with usability, we will consider only the following two sub-characteristics: user interface aesthetics and operability. Other usability sub-characteristics, such as user error protection, were not considered because of the design of this study wish will not allow us to investigate thoroughly their impact on energy consumption. For Operability, the following quality factors will be analyzed in each web-based PHR selected in this study: Understandable categorization of information, Appearance consistency, Operational consistency, and Message clarity. For Interface aesthetics, one quality attribute is assigned to this sub-characteristic by ISO/IEC 25023, which is appearance aesthetics of user interface. This quality attribute can be used to determine to what extend the user interface and overall design are aesthetically pleasing in appearance.

3 3.1

Results Energy Consumption of the Selected Web-Based PHRs

The average energy consumption for each PHR and each sensor was calculated with the data available in this supplementary file (http://tiny.cc/9w72fz). In Table 2 the cells with the largest values were coloured in red whereas the ones with the smallest values in green. In NoMoreClipBoard the lowest energy consumption appeared in three sensors (monitor, processor and PC), in HealtVault in one sensor (hard disk) and in PatiensLikeMe in another one sensor (graphics card). In contrast, Health Companion spent the highest amount of energy in three of the sensors (graphics card, hard disk and monitor) and PatiensLikeMe in two of them (processor and the whole PC).

30

J. A. Garc´ıa-Bern´ a et al. Table 2. Average power consumption in Watts

Web-based PHR Graphics card HealthVet 1.4070701 HealthVault 1.4005134 NoMoreClipboard 1.3844928 PatientsLikeMe 1.3512662 Health Companion 1.4109732

3.2

Hard disk 14.387069 14.373165 14.408445 14.383003 14.425882

Monitor 59.249537 58.584966 57.699905 60.515900 62.705551

Processor 3.9319199 3.3334926 3.1624559 5.2833229 4.1794896

PC 193.34284 184.47266 181.11312 228.16677 204.83057

Operability and Energy Efficiency

Only Health Companion presented a processing icon while the page is loading in most of the cases. Although it is not an environment-friendly feature, loading icon or image can improve usability of the website. The scrollbars were present in all the PHRs. This widget was necessary due to the resolution of the display. Processing icons were less common but were found in Health Companion when performing task 15, export health info. The lowest energy consumptions were measured in NoMoreClipBoard for the same components, in which no loading icons appeared when performing this task. Scrollbars are widely used. Nevertheless, the need to use scrollbars might be determined by the dimensions and resolution of the display. Although they are not recommended for an efficient energy consumption, their availability simplifies portal’s usage [7,16]. NoMoreClipBoard showed all the health data at once in “Task 4: View profile”. In contrast, PatientsLikeMe had a wall of updates from the community, resembling a social network. This widget in PatientsLikeMe required refreshing the web page, which leaded to a high-energy demanding feature. In NoMoreClipBoard all the medical information was accessible at once in “Task 11: View Medications”, and thematic icons were present in this case. HealthVet was the opposite with no thematic icons in Task 11. Autocomplete allowed to find an illness name that may be difficult to remember at first. In the evaluation of “Task 14: Search for information about conditions”, extreme amounts of power were spent by the hard disk, the graphics card and the processor in HealthVet and PatientsLikeMe. HealthVet was the one with the maximum power measurements and PatientsLikeMe the one with the minimum values. This could be explain because in HealthVet a new tab is opened in the browser to complete the task, making it more complex. Although NoMoreClipBoard, HealthVet, HealthVault and PatientsLikeMe had an efficient user interface, these PHRs should improve the energy consumption by simplifying the interaction with the user, and reducing the time required to perform specific operations as shown in Table 3. 3.3

Interface Aesthetics and Energy Efficiency

In terms of power consumption, the lowest CPU energy consumption appeared in NoMoreClipBoard, whilst the highest necessities of power when carrying out

Energy Efficiency and Usability of Web-Based Personal Health Records

31

Table 3. Time required to complete the tasks in each PHR Task HealthVet

HealthVault

NoMoreClipBoard PatientsLikeMe Health Companion

#1

1’17.45”

1’36.79”

1.28”

1’13.05”

2’03.99”

#2

19.33”

45.36”

21.47”

20.26”

14.77”

#3

1’30.24”

26.28”

5’32.05”

1’56.26”

35.11”

#4

8,61”

2.30”

5.92”

10.14”

8.48”

#5

Not available 40.72”

17.54”

8.87”

19.82”

#6

2’05.68”

33.54”

3’44.35”

Not available

1’22.56”

#7

1’23.60”

49.66”

46.10”

50.63”

41.20”

#8

1’17.36”

53.51”

22.40”

Not available

34.20”

#9

50.56”

1’00.17”

29.85”

Not available

43.53”

# 10 47.13”

33”

25.32”

51.34”

42.62”

# 11 5.87”

4.61”

4.94”

5.30”

9.68”

# 12 18.85”

10.06”

22.54”

Not available

23.36”

# 13 11.31”

6.76”

18.45”

9.43”

37.22”

# 14 19.38”

Not available Not available

12.15”

Not available

# 15 33.79”

15.33”

19.92”

Not available

23.23”

# 16 51.29”

44.70”

Not available

Not available

28.16”

# 17 1’09.95”

1’59.10”

27.84”

22.77”

25.23”

# 18 5.95”

4.52”

6.34”

7.09”

4.30”

# 19 3”

10.62”

5.36”

2.69”

4.33”

Not available

20.71”

Not available

# 20 Not available 47.80”

the proposed tasks were generated in PatientsLikeMe, despite of having a good design in human-interface interaction. Both PHRs had an important difference concerning the user interface that could explain the different necessities of power. A total of two main graphic user interface (GUI) factors impacted on the display’s energy efficiency [26]: energy color scheme and screen changes. PHRs were divided according to GUI complexity. Health Companion’s GUI was the simplest one, HealthVet, HealthVault and PatientsLikeMe presented a middle complexity, and NoMoreClipBoard had the most overelaborate GUI. Bearing these groups in mind, the major the complexity, the greater amount of energy spent by the graphics card in “Task 6: Add family history”. This component spent 1.35 W in HealthCompanion and 1.41 W in NoMoreClipBoard. Significant differences in power consumed by the monitor were found between NoMoreClipBoard and Health Companion. This could be explained by the fact that both PHRs had a different color scheme. There were more dark areas in Health Companion than in NoMoreClipBoard, turning in a higher necessity of power for Health Companion. Solid colours stood out in Health Companion, whereas degraded colors were common in NoMoreClipBoard, defining the color scheme of the PHRs. Moreover, the GUI of NoMoreClipBoard was brighter than that of Health Companion, defining also the power needs by the TFT monitor

32

J. A. Garc´ıa-Bern´ a et al.

in the experiment. Solid colors appeared in most of the cases in the GUI of the PHRs. NoMoreClipBoard was the only one with degraded colors, impacting on the switching characteristic of the screen shots. The LCD-TFT employed in this experiment generated a higher power consumption when dark tonalities were shown in the monitor, which could be related to the textures in NoMoreClipBoard [26]. Biggest widgets were relevant in terms of power needs when performing the tasks. Biggest widget were showed from task 7 to 10 in Health Companion. In addition, the energy spent by the monitor in this PHR stood out. The solid color scheme and the dark tonalities of this PHR could explained a low energy need. In HealthVet big widgets also appeared, moreover, they were closer to each other in the aforementioned tasks. This PHR revealed the lowest power consumption for the hard disk drive.

4

Discussion

After analysing the results, several alternatives to be implemented in the PHRs have been identified, which may reduce time required to complete any task and consequently reduce power consumption. 1. Automatic jumping, Whenever the number of characters is known in some fields of a form (i.e. phone number, dates, insurance number, etc.), the automatic jumping of the cursor could be implemented in the PHRs. HealthVet had a form where the cursor moved to the next field when completing the Social Security Number. 2. Macros. The congregation of a set of actions in a macro produces a more efficient user interface [23]. Nevertheless, there were not found macros in any of the PHRs studied, and this feature could improve sustainability of these tools. 3. Autocompletion. Input caches are relevant when a reduced number of known inputs occurs very often. Previous experiments proved that autocompletion functions made more energy-efficient when typing consisted of, at least, three additional letters [26]. The PHRs PatientsLikeMe, HealthVet, NoMoreClipBoard and Health Companion had autocompletion by catching previous inputs to reduce the input time. This feature was present in the reason of a hospitalization, the name of a medical test, conditions, symptoms and treatments for PatientsLikeMe, and the name of the insurance company, medical providers, medications, illnesses, medical procedures, immunizations, allergies and conditions for NoMoreClipBoard. HealthVet autocomplete is available when searching for information about conditions and in Health Companion to add a condition, laboratory data, medications and vaccine. HealthVault did not have autofilling. Observe that this feature can be useful to start to familiarize with a particular medical situation, especially with health names that can be difficult to remember at first.

Energy Efficiency and Usability of Web-Based Personal Health Records

33

4. Hick-Hyman Law. This law postulates that the human cognitive process of taking decision can be accelerated [12,13]. People divide the number of options into groups, eliminating around half of the remaining choices at each step. A logarithmic relationship between reaction time and the number of choices available is proposed in this law. In addition, when each option one at a time must be considered by the users, the relationship between response time and the number of choices available has been found to be linear [8]. Therefore, few choices as possible should be available in a GUI to take advance of HickHyman Law. To this end, the most common functionality should be split into a smaller menu [24]. HealthVault, HealthVet and PatientsLikeMe proceeded with this law. The navigation to find the information in these PHRs was divided into drop-down menus. There was a left column with the main options of the PHR in HealthVault and a second level menu to view the medical information. In HealthVet and PatientsLikeMe there was a first level menu with the main options in the headline of the web. After clicking on this menu, a left column appeared to retrieve the medical information.

5

Conclusions and Future Works

This paper investigated the relationship between usability and energy efficiency of five web-based PHRs. The findings showed that meeting both usability and consuming less energy is challenging as it depends on factors related to the hardware employed and the users’ manipulation of the system. Results allowed us to suggest recommendations about energy efficiency of PHRs. However, the inclusion of only five PHRs might have impacted the results. Further studies with more PHRs and more tasks to be performed should be conducted to confirm our results. For future work, we intend to propose a reusable requirement catalogue for usable and energy efficient e-health applications and use energy management systems during the performance of tasks to validate our catalogue.

References 1. ISO/IEC 25010 standard. Systems and Software Engineering – Systems and Software Quality Requirements and Evaluation (SQuaRE) – System and Software Quality Models (2011) 2. AHIMA: American health information management association (2019). Accesses Dec 2019. http://www.ahima.org/ 3. Ahmad, R., Hussain, A., Baharom, F.: A systematic review on characteristic and sub-characteristic for software development towards software sustainability. Environment 20, 34 (2015) 4. Archer, N., Fevrier-Thomas, U., Lokker, C., McKibbon, K.A., Straus, S.E.: Personal health records: a scoping review. J. Am. Med. Inform. Assoc. 18(4), 515–522 (2011) 5. Bhatt, C., Dey, N., Ashour, A.S.: Internet of Things and Big Data Technologies for Next Generation Healthcare, vol. 23. Springer, Cham (2017)

34

J. A. Garc´ıa-Bern´ a et al.

6. Bidargaddi, N., van Kasteren, Y., Musiat, P., Kidd, M.: Developing a third-party analytics application using Australia’s national personal health records system: case study. JMIR Med. Inform. 6(2), e28 (2018) 7. Breuninger, J., Popova-Dlugosch, S., Bengler, K.: The safest way to scroll a list: a usability study comparing different ways of scrolling through lists on touch screen devices. IFAC Proc. Vol. 46(15), 44–51 (2013) 8. Cockburn, A., Gutwin, C.: A predictive model of human performance with scrolling and hierarchical lists. Hum.-Comput. Interact. 24(3), 273–314 (2009) 9. Farzandipour, M., Meidani, Z., Riazi, H., Sadeqi Jabali, M.: Task-specific usability requirements of electronic medical records systems: lessons learned from a national survey of end-users. Inform. Health Soc. Care 43(3), 280–299 (2018) 10. Fern´ andez-Alem´ an, J.L., Seva-Llor, C.L., Toval, A., Ouhbi, S., Fern´ andez-Luque, L.: Free web-based personal health records: an analysis of functionality. J. Med. Syst. 37(6), 9990 (2013) 11. Helmer, A., Lipprandt, M., Frenken, T., Eichelberg, M., Hein, A.: Empowering patients through personal health records: a survey of existing third-party webbased PHR products. Electron. J. Health Inform. 6(3), 26 (2011) 12. Hick, W.E.: On the rate of gain of information. Q. J. Exp. Psychol. 4(1), 11–26 (1952) 13. Hyman, R.: Stimulus information as a determinant of reaction time. J. Exp. Psychol. 45(3), 188 (1953) 14. Karampela, M., Ouhbi, S., Isomursu, M.: Exploring users’ willingness to share their health and personal data under the prism of the new GDPR: implications in healthcare. In: 41st Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), pp. 6509–6512. IEEE (2019) 15. Kelly, M.M., Coller, R.J., Hoonakker, P.: Inpatient portals for hospitalized patients and caregivers: a systematic review. J. Hosp. Med. 13(6), 405–412 (2018) 16. Leung, R., MacLean, K., Bertelsen, M.B., Saubhasik, M.: Evaluation of haptically augmented touchscreen GUI elements under cognitive load. In: 9th International Conference on Multimodal Interfaces, pp. 374–381. ACM (2007) 17. Lyerla, F., Durbin, C.R., Henderson, R.: Development of a nursing electronic medical record usability protocol. CIN: Comput. Inform. Nurs. 36(8), 393–397 (2018) 18. Mancebo, J., Arriaga, H.O., Garc´ıa, F., Moraga, M., Garc´ıa-Rodr´ıguez de Guzm´ an, I., Calero, C.: EET: a device to support the measurement of software consumption. In: 6th International Workshop on Green and Sustainable Software (GREENS), pp. 16–22 (2018) 19. Moher, D., Liberati, A., Tetzlaff, J., Altman, D.G.: Preferred reporting items for systematic reviews and meta-analyses: the prisma statement. Ann. Int. Med. 151(4), 264–269 (2009) 20. Ouhbi, S.: Sustainability and internationalization requirements for connected health services: method and applications. Proyecto de investigaci´ on (2018) 21. Ouhbi, S., Fern´ andez-Alem´ an, J.L., Toval, A., Rivera Pozo, J., Idri, A.: Sustainability requirements for connected health applications. J. Softw.: Evol. Process 30(7), e1922 (2018) 22. Rantanen, M.M., Koskinen, J.: Phr, we’ve had a problem here. In: IFIP International Conference on Human Choice and Computers, pp. 374–383. Springer (2018) 23. Savelyev, A., Brookes, E.: GenApp: extensible tool for rapid generation of web and native GUI applications. Future Gener. Comput. Syst. 94, 929–936 (2017) 24. Sears, A., Shneiderman, B.: Split menus: effectively using selection frequency to organize menus. ACM Trans. Comput.-Hum. Interact. (TOCHI) 1(1), 27–51 (1994)

Energy Efficiency and Usability of Web-Based Personal Health Records

35

25. Staccini, P., Lau, A.Y., et al.: Findings from 2017 on consumer health informatics and education: health data access and sharing. Yearb. Med. Inform. 27(01), 163– 169 (2018) 26. Vallerio, K.S., Zhong, L., Jha, N.K.: Energy-efficient graphical user interface design. IEEE Trans. Mob. Comput. 5(7), 846–859 (2006) 27. Venters, C., Lau, L., Griffiths, M., Holmes, V., Ward, R., Jay, C., Dibsdale, C., Xu, J.: The blind men and the elephant: towards an empirical evaluation framework for software sustainability. J. Open Res. Softw. 2(1), e8 (2014) 28. Venters, C.C., Capilla, R., Betz, S., Penzenstadler, B., Crick, T., Crouch, S., Nakagawa, E.Y., Becker, C., Carrillo, C.: Software sustainability: research and practice from a software architecture viewpoint. J. Syst. Softw. 138, 174–188 (2018) 29. Villa, L., Cabezas, I., Lopez, M., Casas, O.: Towards a sustainable architectural design by an adaptation of the architectural driven design method. In: International Conference on Computational Science and Its Applications, pp. 71–86. Springer (2016) 30. Yen, P.Y., Walker, D.M., Smith, J.M.G., Zhou, M.P., Menser, T.L., McAlearney, A.S.: Usability evaluation of a commercial inpatient portal. Int. J. Med. Inform. 110, 10–18 (2018)

A Complete Prenatal Solution for a Reproductive Health Unit in Morocco Mariam Bachiri1, Ali Idri1,2(&), Taoufik Rachad1, Hassan Alami3, and Leanne M. Redman4 1

Software Project Management Research Team, Department of Web and Mobile Engineering, ENSIAS, Mohammed V University in Rabat, Rabat, Morocco [email protected] 2 CSEHS, Mohammed VI Polytechnic University, Ben Guerir, Morocco 3 Faculty of Medicine, Mohammed V University in Rabat, Rabat, Morocco 4 Pennington Biomedical Research Center, Baton Rouge, LA 70808, USA

Abstract. A prenatal mobile Personal Health Records (mPHR), along with an Electronic Health Records (EHR) are, respectively, exploited in order to permit both the pregnant women and gynecologists or obstetricians monitor the pregnancy progress in the best conditions. For this intent, a complete solution consisting of a prenatal mPHR and an EHR were developed for the maternity “Les Orangers” of the Avicenne University Hospital in Rabat. The complete solution provides the main functionalities of a prenatal service. Thereafter, the solution will be validated by conducting an experiment for quality and potential assessment. Hence, a recruitment process has been determined to identify the eligibility criteria to enroll participants (pregnant women and gynecologists), in addition to planning the course of the experiment. Keywords: Mobile personal health records Experiment
Prenatal
Pregnancy


Electronic health records


1 Introduction A pregnancy can encounter disorders or medical conditions that might influence the progress of the pregnancy. It can be either related to the obstetrical history, medical complications, lifestyle choices, nutrition, cardiac diseases, diabetes or hypertensive disorders [1]. These factors can produce health troubles before, during and after delivery. This implies regular prenatal checkups with her obstetrician and gynecologist in order to track her health and the baby’s health [2]. During these checkups, health data related to the pregnant woman and her baby are registered in their health records. The classical form of these health records are paper-based health records. They allow pregnant women access their health data, by meticulously following the progress of their pregnancy [3]. Hence, pregnant women who can access their health records are more informed and aware about potential risks while being pregnant, which can help

© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2020 Á. Rocha et al. (Eds.): WorldCIST 2020, AISC 1161, pp. 36–43, 2020. https://doi.org/10.1007/978-3-030-45697-9_4

A Complete Prenatal Solution for a Reproductive Health Unit in Morocco

37

them take better decisions about their health [4]. However, the use of paper-based health records remains inadequate, since they are exposed to loss, they cannot be shared with healthcare providers and it is difficult to identify each information included in them. Hence, since pregnant women should monitor their health away from the hospital, there is a need to remotely communicate with the healthcare providers when there is a necessity [5]. Prenatal mobile Personal Health Records (mPHRs) are useful for these purposes. They are mobile applications available on the app stores, which can be installed on smartphones and allow pregnant women consult, record and control their health data whenever required [6]. Prenatal mPHRs can be connected to Electronic Health Records (EHRs), which are implemented in hospitals to be used only by obstetricians, gynecologists or their assistants. EHRs can be either web or desktop applications accessible from computers. Facilitating the access to health data, information about the progress of pregnancy and communication with healthcare providers promotes the improvement of the health status of pregnant women and their infants, and therefore indirectly lessening the mortality rates. As part of a collaboration with the maternity “Les Orangers” in Rabat, a prenatal mPHR and EHR were developed, based on the specifications that were extracted while visiting the maternity several times. Afterwards, these applications will be evaluated in the course of an experiment, which will be conducted among selected pregnant women and obstetricians or gynecologists who will participate in this experiment. The remainder of this paper is structured as follows: An overview of prenatal mobile personal health records is introduced in Sect. 2. The developed solution is presented in Sect. 3. The implementation of the solution is explained in Sect. 4. Section 5 describes the experiment design of the solution. Lastly, Sect. 6 provides conclusions and future work.

2 Prenatal Mobile Personal Health Records: An Overview Prenatal mPHRs are mobile health applications that allow a pregnant woman to access, record and share her health data with healthcare professionals, in order to carry on an accurate and consistent monitoring for her health and the baby’s health [7]. According to a previous study [6], these mobile apps generally include features such as: Calendar and reminders for follow-ups and important appointments, information about the progress of pregnancy regarding the mother and the baby’s health, health habits to be followed during pregnancy, in addition to recorders and counters for baby kicks and contractions. Furthermore, among the data that should be collected in the prenatal mPHRs are the pregnant woman’s personal details, the physical body information (e.g. weight, blood pressure, glucose, heart frequency), her medical history (e.g. allergies or immunizations) and her obstetrical history (e.g. contraception methods used, information about previous pregnancies and the status of her current pregnancy) [7].

38

M. Bachiri et al.

3 A New Prenatal Solution This Section presents the purpose of the new prenatal solution as well as its requirements’ specification. 3.1

Purpose

A prenatal mPHR was developed for the pregnant women to follow up their pregnancy and stay in touch with their doctors, while having a vision on their personal health records. This mobile application interacts with the EHR, which is a web application that will be implemented for the healthcare providers (doctors and assistants). Hence, it will permit the ease of interaction between the pregnant woman and her doctor. The doctor’s role is to fill out the health records of his patients during each consultation, to be able to follow their health state through the EHR. The management of appointments, patients, doctors and their availability are also handled in the EHR. As for pregnant women, they will be able to access and consult their personal health records through the prenatal mPHR, in addition to taking appointments according to the availability of doctors, consulting information about the progress of pregnancy as regards the mother and the baby, recording contractions, baby kicks and the measured weight and blood pressure, which will be accessible for doctors via the EHR. 3.2

Requirements Specification

Based on the requirements catalog of prenatal mPHRs conceived in a previous study [8], in addition to scheduled visits to the maternity “Les Orangers”, by assisting to realtime consultations and discussing with the medical staff, a set of requirements for both the prenatal mPHR and EHR were identified. Hence, the following functional requirements were implemented in the developed prenatal mPHR: • Displaying an up-to-date version of the health records received from the EHR. • Recording the weight and blood pressure measurements that have been taken by the pregnant woman, which will be available for the doctors through the EHR. • Consulting the availability of doctors for consultations, received from the EHR, and taking appointments. • Entering measurements (Weight/Blood pressure: Lists and graphs by date) and displaying these measurements to the doctor through the EHR. • Authenticating by email and password. The password is assigned to each patient at the creation of their personal health records, and can be changed later by the patient if needed. • Consulting information about the progress of pregnancy per week (1–40). • Contractions counter. • Baby kicks counter. • The application can support other languages in the future.

A Complete Prenatal Solution for a Reproductive Health Unit in Morocco

39

As regards the non-functional requirements mentioned in the previously conceived catalog [8], the mPHR should meet the Operability, Performance efficiency, Reliability, Functional suitability, Sustainability and I18n requirements. Moreover, the following functional requirements were implemented in the developed EHR: • • • •

Creating a PHR for the pregnant woman. Editing the PHR. Managing the available days for consultations. Consulting the measurements of the weight and blood pressure that were recorded by the patient from the mPHR. • Managing patients by adding, editing, deleting and searching. • Authenticating by email and password. The creation of the PHR consists in covering the following data: 1. Personal information: Last name and first name, date of birth, blood group, height, phone number, address, doctor’s name, family situation, age of marriage and profession. 2. Medical History: Number of children, number of previous pregnancies, chronic diseases, genetic diseases, contraceptive method used, surgical antecedents, gynecological antecedents and obstetrical antecedents. 3. Current pregnancy: Expected date of delivery, weight before pregnancy and last menstrual period date. 4. Measurements: Measurement date, weight and blood pressure. 5. Consultation: Consultation’s date, weight, blood pressure, uterine height, observations, treatment, pelvic exam, breast examination, conjunctive state, echography and blood analysis.

4 Implementation of the Prenatal Solution 4.1

The Prenatal mPHR

The prenatal mPHR is an Android mobile application developed using the Java language, since it is one of the official languages for developing native Android applications [9]. Native applications have the best performance, are more secure, interactive, intuitive and allow developers access the full features of the devices. In order to use the prenatal mPHR, the user is asked first to define a PIN code to secure the access to the application. Secondly, she is asked to sign in using her email and the password that has been assigned to her at the creation of her PHR. Once signed in, the user can access general and diverse weekly information about pregnancy, from week 1 to week 40, either related to the mother or the baby. She can also consult her PHR, which was filled in by the doctor or the assistant, via the EHR, including her personal and medical information. Moreover, the user can visualize a calendar displaying the available days for consultations. Once an available day is selected, the user can choose a convenient time for the consultation and take an appointment.

40

M. Bachiri et al.

Furthermore, the user can record baby kicks and contractions, using recorders that permit to save the history of the records in the application. For instance, in order to record contractions, the user marks the start of a session, and every time a contraction occurs, she clicks on a button to record it until contractions stop, she then ends the session and the records are automatically saved. Lastly, the user can enter their measured weight or blood pressure, according to a defined date and time, and then visualize the progress history of these variables as graphs or lists. Figure 1 (a–j) in Appendix demonstrates some screenshots of the developed prenatal mPHR. The Appendix is accessible via the following link: https://www.um.es/ giisw/prenatal/Appendix.pdf. 4.2

The EHR

The EHR is a web application that was developed using HyperText Markup Language (HTML), Cascading Style Sheets (CSS) and JavaScript. Google Firebase [10] was used as well to provide a real-time database and assure authentication of doctors in the EHR, in addition to hosting the EHR. Through the EHR, the doctors or assistants can manage the patients’ health records by consulting, adding, editing and deleting them. Note that deleting a patient’s health records leads to denying her access to the mPHR. Moreover, the list of doctors can be managed as well, by adding a new doctor, editing and deleting a specific one. Note that deleting a doctor leads to denying his access to the EHR. Furthermore, the set of consultations for each patient are managed, since the doctor can add, edit and delete a consultation. Doctors can also specify the days of consultations when they are available, or remove them. The doctors or assistants can consult, in real time, the list of appointments taken by patients as well. In this list, the date, time, and name of the patients are indicated. Lastly, the doctors or assistants can consult the weight and blood pressure measurements recorded by the pregnant women in the mPHR. Figure 2 (a–f) in Appendix shows some screenshots of the developed EHR. The Appendix is accessible at the following link: https://www.um.es/giisw/prenatal/ Appendix.pdf.

5 Experiment Design This section presents the experimental design we will follow to evaluate the quality and the usefulness of our prenatal solution. Firstly, we present the criteria we will use to recruit participants (pregnant women and gynecologists). Secondly, we describe the recruitment process. 5.1

Selection Criteria

The aim of this experiment is to evaluate the quality and potential of both the prenatal mPHR and EHR. For this purpose, pregnant women and obstetricians/gynecologists will be recruited in order to carry out this experiment.

A Complete Prenatal Solution for a Reproductive Health Unit in Morocco

41

A set of inclusion (ICs) and exclusion criteria (ECs) were defined for this selection. Hence, pregnant women are eligible for the experiment if they: • • • • • •

are aged between18 and 45 years. are resident in either Rabat or Casablanca. are currently pregnant. have a moderate level of experience with mobile applications. own a smartphone that runs a recent version of Android. are willing the comply with all study procedures. Otherwise, pregnant women are ineligible if they:

• • • • • • • • • •

are planning to relocate from the study area within the next two years. are currently smoking. use alcohol and drugs. have a sever debilitating illness preventing their participation. are infertile. are sterilized. are in premature or normal menopause. are not planning to deliver at public hospitals in Casablanca or Rabat. are willing to terminate their pregnancy. have history of three or more consecutive pregnancy complications.

As for obstetricians/gynecologists, they are eligible for the experiment only if they are practicing in public hospitals in Rabat or Casablanca. The experiment will be then carried out in two phases: (1) Before delivery: The selected pregnant women will be assigned one of the selected obstetricians/ gynecologists and will be asked to use the developed prenatal mPHR to monitor their pregnancy. Throughout their use of the prenatal mPHR, they will have to access and record their own health data (weight, blood pressure, baby kicks and contractions) and connect with their obstetrician/gynecologist in real-time. Moreover, they will have to set appointments for consultations. The obstetrician/gynecologist will guide the pregnant women all along the progress of their pregnancy until their due date, and will be at disposal if any complication occurs. (2) After delivery: After giving birth to the baby and getting enough rest, the new mother will be asked to fill in questionnaires in order to evaluate the potential and quality of the prenatal mPHR. 5.2

Recruitment Process

In this experiment, the enrollment of pregnant women and obstetricians/gynecologists will target local public hospitals in Rabat and Casablanca to directly recruit potential subjects. Hence, five obstetricians/gynecologists are expected to be enrolled in this experiment to fulfill reproductive health needs of the recruited pregnant women, through the EHR and the scheduled appointments.

42

M. Bachiri et al.

Obstetricians/gynecologists will be guided by providing them with a detailed overview of the conduct of the experiment. An informed consent form will be given in advance to the obstetricians/gynecologists to guide the recruitment discussion. Therefore, after obtaining their consent to be part of the study, the obstetricians/ gynecologists will be selected for the experiment. As regards pregnant women, the targeted number of participants is approximately 50. For this purpose, two research assistants will be in charge of the recruitment process. First, flyers and posters will be distributed over public hospitals in both Rabat and Casablanca to attract interested subjects, who can, thereafter, contact the research team for further details about the experiment if needed. Hence, if they seem willing to participate, they will be given a written informed consent document to cautiously read, understand and sign. Moreover, detailed information about the purpose of the study, the procedures to be followed, the risks and discomforts as well as potential benefits associated with participation will be explained to them. Furthermore, they will be invited to an interview during which eligibility criteria will be evaluated. Participants who meet these criteria will be then enrolled to the experiment. Afterwards, a meeting will be held to gather the registered participants and help them download and install the developed prenatal mPHR on their smartphones. Moreover, the main features and functionalities will be briefly presented to the participants, in addition to instructions about the conduct of the experiment.

6 Conclusion and Future Work In pursuance of facilitating the access to reproductive healthcare services for pregnant women, a prenatal mPHR and EHR have been developed, according to specifications based on our previous studies on prenatal health services [8, 11–16] and scheduled visits to the maternity “Les Orangers” in Rabat. The prenatal mPHR is intended to be used by pregnant women, while the EHR will be used by obstetricians/gynecologists or their assistants. Hence, an experiment is expected to be conducted among pregnant women and obstetricians/gynecologists, in order to assess the quality of these solutions and evaluate their potential as regards improving the reproductive healthcare services for tracking pregnancy. Moreover, an iOS version of the mPHR is intended to be developed in order to target a large number of users. Acknowledgments. This work was conducted within the research project PEER 7-246 supported by the US Agency for International Development (USAID). The authors would like to thank the NAS and USAID for their support.

A Complete Prenatal Solution for a Reproductive Health Unit in Morocco

43

References 1. Gu, B.D., Yang, J.J., Li, J.Q., Wang, Q., Niu, Y.: Using knowledge management and mhealth in high-risk pregnancy care: a case for the floating population in China. In: Proceedings of the IEEE 38th Annual International Computer Software Applications Conference Workshops COMPSACW 2014, pp. 678–683 (2014) 2. Oh, S., Sheble, L., Choemprayong, S.: Personal pregnancy health records (PregHeR): facets to interface design. Proc. Am. Soc. Inf. Sci. Technol. 43(1), 1–10 (2007) 3. Hoang, D.B., et al.: Assistive care loop with electronic maternity records. In: 2008 10th IEEE International. Conference e-Health Networking, Applications Services Health, pp. 118–123 (2008) 4. Homer, C.S.E., Davis, G.K., Everitt, L.S.: The introduction of a woman-held record into a hospital antenatal clinic: the bring your own records study. Aust. N. Z. J. Obstet. Gynaecol. 39(1), 54–57 (1999) 5. Shaw, E., et al.: Access to web-based personalized antenatal health records for pregnant women: a randomized controlled trial. J. Obstet. Gynaecol. Can. 30(1), 38–43 (2008) 6. Bachiri, M., Idri, A., Fernández-Alemán, J.L., Toval, A.: Mobile personal health records for pregnancy monitoring functionalities: analysis and potential. Comput. Methods Programs Biomed. 134, 121–135 (2016) 7. Idri, A., Bachiri, M., Fernández-Alemán, J.L.: A framework for evaluating the software product quality of pregnancy monitoring mobile personal health records. J. Med. Syst. 40(3), 50 (2016) 8. Bachiri, M., Idri, A., Redman, L.M., Fernández-Alemán, J.L., Toval, A.: A requirements catalog of mobile personal health records for prenatal care. In: Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics, vol. 11622 LNCS, pp. 483–495 (2019) 9. Application Fundamentals. https://developer.android.com/guide/components/fundamentals. Accessed 07 Nov 2019 10. Google Firebase. https://firebase.google.com/. Accessed 06 Nov 2019 11. Sardi, L., Idri, A., Redman, L.M., Alami, H., Bezad, R., Fernández-Alemán, J.L.: Mobile health applications for postnatal care: review and analysis of functionalities and technical features. Comput. Methods Programs Biomed. 184, 105114 (2020) 12. Bachiri, M., Idri, A., Abran, A., Redman, L.M., Fernández-Alemán, J.L.: Sizing prenatal mPHRs using COSMIC measurement method. J. Med. Syst. 43(10), 319 (2019) 13. Bachiri, M., Idri, A., Redman, L.M., Abran, A., de Gea, J.M.C., Fernández-Alemán, J.L.: COSMIC functional size measurement of mobile personal health records for pregnancy monitoring. Adv. Intell. Syst. Comput. 932, 24–33 (2019) 14. Idri, A., Bachiri, M., Fernández-Alemán, J.L., Toval, A.: Experiment design of free pregnancy monitoring mobile personal health records quality evaluation. In: 2016 IEEE 18th International Conference on e-Health Networking, Applications and Services (Healthcom), pp. 1–6 (2016) 15. Bachiri, M., Idri, A., Fernández-Alemán, J.L., Toval, A.: Evaluating the privacy policies of mobile personal health records for pregnancy monitoring. J. Med. Syst. 42(8), 144 (2018) 16. Bachiri, M., Idri, A., Fernández-Alemán, J.L., Toval, A.: A preliminary study on the evaluation of software product quality of pregnancy monitoring mPHRs. In: Proceedings of 2015 IEEE World Conference on Complex Systems WCCS 2015 (2016)

Machine Learning and Image Processing for Breast Cancer: A Systematic Map Hasnae Zerouaoui1, Ali Idri1,2(&), and Khalid El Asnaoui1 1

Complex Systems Engineering and Human Systems, Mohammed VI Polytechnic University, Ben Guerir, Morocco {Hasnae.zerouaoui,Khalid.elasnaoui}@um6p.ma 2 Software Project Management Research Team, ENSIAS, Mohammed V University in Rabat, Rabat, Morocco [email protected]

Abstract. Machine Learning (ML) combined with Image Processing (IP) gives a powerful tool to help physician, doctors and radiologist to make more accurate decisions. Breast cancer (BC) is a largely common disease among women worldwide; it is one of the medical sub-field that are experiencing an emergence of the use of ML and IP techniques. This paper explores the use of ML and IP techniques for BC in the form of a systematic mapping study. 530 papers published between 2000 and August 2019 were selected and analyzed according to 6 criteria: year and publication channel, empirical type, research type, medical task, machine learning objectives and datasets used. The results show that classification was the most used ML objective. As for the datasets most of the articles used private datasets belonging to hospitals, although papers using public data choose MIAS (Mammographic Image Analysis Society) which make it as the most used public dataset. Keywords: Breast cancer
Machine learning
Image processing
Systematic mapping study

1 Introduction One of the most common cancers for women in the world is Breast Cancer. It happens when the cell tissue of the breast cells grows abnormally and start to divide rapidly [1]. The BC disease is distinguished by an overgrowth of a malignant tumor in the breast [2]. The goal of BC screening is to achieve an early diagnosis, which aims to discern the Malignant and Benign tumor, as for prognosis helps to put a treatment plan. The use of medical image processing and machine learning for breast cancer diagnosis, prognosis and/or treatment is promising since it can help physicians, doctors and experts in detecting efficiently abnormalities [3]. To the extent of the authors’ knowledge, no Systematic Mapping Study (SMS) was carried out to summarize the findings of primary studies dealing with the use of machine learning and image processing techniques for any breast cancer medical tasks such as diagnosis, prognosis and treatment. However, Idri et al. [4] have carried out a SMS on the use of data mining techniques in BC, and Hosni et al. [5] conducted a SMS © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2020 Á. Rocha et al. (Eds.): WorldCIST 2020, AISC 1161, pp. 44–53, 2020. https://doi.org/10.1007/978-3-030-45697-9_5

Machine Learning and Image Processing for Breast Cancer

45

on the use of ensemble techniques in breast cancer. The present SMS searches the primary studies dealing with the application of machine learning and image processing for BC published between 2000 to August 2019 in six libraries: ScienceDirecte, IEEEXPLORE, Pubmed, Springer, ACM and Google Scholar. It provides a synthesis and a summary of 530 selected papers by means of six Research Questions (RQs): (1) determine the year, publication channels and sources of the selected papers, (2) identify the type of contributions and empirical methods, (3) examine the most used machine learning objective, (4) discover the datasets employed for ML and IP for BC. The paper is structured as follow: Sect. 2 describes the research methodology followed by this review. Section 3 reports the results of the four RQs. Section 4 discusses the results obtained. Section 5 concludes this SMS.

2 Research Methodology The purpose of a systematic mapping study is to offer an overview of a research area by identifying the research type and quantity of a research field and to describe broadly the methodologies and results of primary studies [6]. A SMS involves 5 steps which are: Defining the research questions, searching for relevant papers, screening the selected papers, keywording of abstract and data extracting and mapping the results. 2.1

Research Questions

The main goal of this paper is to provide an overview of the studies published from 2000 to August 2019 in the field of machine learning and image processing techniques applied to breast cancer. Therefore, we identify four research questions and their motivations as shown in Table 1. Table 1. Research questions N# RQ1

RQ2

RQ3 RQ4

Research question In which year, publication channels and sources were the selected papers related to machine learning and image processing in breast cancer published? What type of contributions and empirical methods is being made to the area of machine learning and image processing in breast cancer? Which is the most investigated machine learning objective? What are the most used datasets for ML and IP in BC?

Motivation Identify the publication trends, and the different publication channels and sources of the papers selected Identify the different type of studies performed in ML and IP applied to BC

Discover the most investigated ML objective in BC literature Identify the most used datasets

46

2.2

H. Zerouaoui et al.

Search Strategy

To formulate the search string, we used the principal key words and their synonyms extracted from the research questions. The Boolean AND was used to join the important parts and the Boolean OR was used to join alternative words. The finale search string was defined as followed: (Breast OR “Mammary gland”) AND (cancer* OR tumor OR malignancy OR masses) AND (Prognosis OR Predict* OR Diagnosis OR Identification OR Analysis OR monitoring OR treatment) AND (“data mining” OR intelligent OR classificat* OR cluster* OR associat* OR predict* OR “machine learning” OR “deep learning”) AND (model* OR algorithm* OR technique* OR rule* OR method* OR tool* OR framework*) AND (mammogr* OR ultrasound OR thermogra* OR “magnetic resonance imaging” OR tomosynthesis OR tomography OR imag* OR “image processing” OR “medical images” OR “computer vision”). We search the relevant papers in six digital libraries: Science Direct, IEEEXPLORE, PubMed, ACM, Springer and Google Scholar. These libraries offer many candidate papers, furthermore they index several journals, conferences, and books addressing the topic of this study. 2.3

Study Selection

In order to select the relevant papers for our SMS, we identified a set of inclusion and exclusion criteria (ICs/ECs) combined by the OR Boolean operator. The ICs/ECs we used are tabulated in Table 2. Three authors evaluated the candidate papers using these ICs/ECs to decide on including or excluding each paper; in case of a disagreement, a meeting took place between the three authors to reach a final decision. Table 2. Inclusion and exclusion criteria Inclusion criteria IC1: papers using new or proposing existing ML and IP techniques for BC IC2: papers presenting an overview on the use of ML and IP techniques in BC IC3: papers providing empirical/theoretical comparisons of ML and IP techniques in BC IC4: papers published between 2000 and later than 2019

2.4

Exclusion criteria EC1: Papers written in other languages than English EC2: Papers dealing with others cancer types EC3: duplicated papers EC4: Short papers with only (2–3 pages) EC5: Presentations or posters

Data Extraction Strategy and Synthesis

After selecting the relevant papers, we followed a form to extract the relevant data from the selected studies in order to answer the four RQs of Table 1. RQ1: Involves the identification of the year of publication, the publication channel (Journal, conference, book section), and publication source. RQ2: Identification of the research types [7]: Evaluation Research (ER), Solution Proposal (SP), Experience Papers, Review, Case

Machine Learning and Image Processing for Breast Cancer

47

study, Survey, and Historical based evaluation. RQ3: Identification of the machine learning objectives such as: classification, clustering, prediction and others. RQ4: Identification of the datasets employed [4]. 2.5

Threats to Validity

The main threats of validity for this study are presented below. Study Selection Bias: To choose the relevant papers for this study, we established a search string that contain all the important keywords to cover the maximum of primary studies from the digital libraries that were used (Science Direct, IEEEXPLORE, PubMed, ACM, Springer and Google Scholar). To prevent excluding relevant papers, selection criteria were defined to rigorously match the RQs. Data Extraction Bias: Data extraction is a crucial step in the process of the SMS, any inaccuracy may lead to incorrect results. Therefore, the extracted data is validated by the three authors. in case of a disagreement a discussion took place between the three authors to finally decide if the paper is relevant for the study or not.

3 Results This section presents an overview of the selection process results. Thereafter, we present the results of each RQ. 3.1

Studies Selection

As shown in the Fig. 1, 5817 candidate papers were extracted using the search string on the 6 digital libraries. When applying the exclusion criteria on the titles, keywords and eventually the abstracts of the candidate paper, 5028 papers were discarded. Then, we apply on the 789 remaining studies the inclusion criteria and we obtain 530 selected studies. The list of these 530 papers including all required information to answer the RQs of this SMS is available upon request by email to the authors of this study.

Fig. 1. Selection process

48

3.2

H. Zerouaoui et al.

RQ1: In Which Year, Publication Channels and Sources Were the Selected Papers Related to Machine Learning and Image Processing in Breast Cancer Published?

In Fig. 2 we can identify the number of the selected papers extracted from 2000 to August 2019. 71% of the papers were published in journals, as 27% were presented in conferences and only 1% is presented as book chapters. The most frequent journals are Expert Systems with Applications, Computers in Biology and Medicine, Computer Methods and Programs in Biomedicine, IEEE Access and Scientific reports. The most recurrent conferences are RACS Proceedings of the ACM Symposium on Research in Applied Computation, International Symposium on Biomedical Imaging (ISBI), IEEE International Conference on Big Data (Big Data), IEEE International Conference on Bioinformatics and Biomedicine (BIBM) and Image processing. In Fig. 2 we observe that the number of papers published before 2015 was very low compared with the number of papers published from 2015 to 2019.

Fig. 2. Number of papers published per year and publication channel

3.3

RQ2: What Type of Contributions and Empirical Methods is Being Made to the Area of Machine Learning and Image Processing in Breast Cancer?

We identify three main research types in our SMS: Evaluation research (ER), Solution Proposal (SP) and Review. 59% of the selected papers were SP proposing new or improving existing machine learning techniques based on image processing for breast cancer. 66% of the SP were also evaluated and classified as being ER studies, and only 34% papers proposed new ML techniques without evaluation. 31% of the papers were classified as ER for comparing or evaluating existing Ml techniques. 10% of selected papers were reviews. The evolution of the research type of the selected papers over

Machine Learning and Image Processing for Breast Cancer

49

Fig. 3. Evolution of research types identified over the years

years is presented in Fig. 3. We note that SP and ER come into sight in 2000 and rise over the years. Also reviews started to get more interested from 2017. The selected papers were empirically evaluated using three types of empirical evaluation: case study, historical based evaluation and survey [8]. As shown in Fig. 4, most of the solution proposal articles used historical based evaluation by using publicly available databases. For the evaluation research, most of the papers used a case study based empirical evaluation, and for the review they used survey empirical method. It’s noticed that researchers started to give more importance to reviews for the large number of articles published in the subject and the importance of information that needs to be summarized, hence the importance of this research type.

Fig. 4. Distribution of research types and empirical types

50

3.4

H. Zerouaoui et al.

RQ3: Which is the Most Investigated Machine Learning Objective?

The aim of the RQ3 is to discover the most investigated machine learning objective with image processing in Breast cancer. Figure 5 shows the distribution of the ML objectives. We observe that 89% of the selected papers dealt with classification which consists of classifying the tumor in malign or benign, while 6% treated the prediction objective, 4% dealt with clustering and only 1% for association.

Fig. 5. Distribution of machine learning objectives

3.5

RQ4: What are the Datasets Used for ML and IP in BC?

The aim of RQ6 is to identify the different datasets, the validation methods and the performance measures used to evaluate the use of machine learning and image processing in breast cancer. Table 3 shows the most used datasets in the 530 selected papers selected. It can be noticed that 47% of datasets are private, MIAS is used by 15% of selected studies, Digital Database for Screening Mammography (DDSM) (13%), Breast Cancer Histopathological (BREAKHIS) (5%), Breast Cancer Digital Repository (BCDR), WISCONSIN and INBREAST (3% each), Mytos (2%), 1% The Cancer Genome Atlas) TCGA. The remaining articles used other databases such as IMAGENET, ICIAR, Camelyon challenge, BUS, IRMA, and AMIDA. Table 3. Datasets used Datase t Private MIAS DDSM BREAKHIS BCDR

N# of pape rs 237 74 64 23 16

Datase t WISCONSIN INBREAST MYT OS T CGA OT HERS

N# of pape rs 13 13 11 6 42

Machine Learning and Image Processing for Breast Cancer

51

4 Discussion This section discusses the results of the 6 research questions of Table 1. 4.1

RQ1: In Which Year, Publication Channels and Sources Were the Selected Papers Related to Machine Learning and Image Processing in Breast Cancer Published?

From Fig. 2, it is noticeable that the number of publications significantly increased in 2016, since ML and IP are becoming an important issue and are increasingly used by researchers in the medical field, particularly in breast cancer. This is due to the effectiveness of ML and IP techniques in improving the performance of medical decisions. The selected papers were published in types of channels: Journals, conferences and book chapters. Furthermore, we notice that 71% of the papers were published in journals which reflect the importance of the research and the good scientific level of maturity, since it is in general more difficult to publish in journals than in conferences and symposiums. For the sources of publications, there is no specific publication source, but different ones were targeted such as medicine, computer science applied to medicine, computer science and artificial intelligence, and this is due to the multidisciplinary of the field ML and IP applied to breast cancer. 4.2

RQ2: What Type of Contributions and Empirical Methods is Being Made to the Area of Machine Learning and Image Processing in Breast Cancer?

The 530 selected papers can be classified in three types: evaluation research, solution proposal and review. We notice that the solution proposal papers used evaluation techniques to measure the performance of the proposed methods, the use of SP by researchers is due to the fact that the domain of ML and IP for BC still needs new and more effective solutions to offer better results. Also, the fact that most of the SP papers evaluated their techniques proves a good scientific maturity. As for the papers presenting a review gained more interested since 2017 due to the fact that the number of primary studies became important and therefore the need of synthesize and summarize their findings. For the empirical type, evaluation of solution proposal (SP) was in general done using historical data, since researchers choose to test their newly developed technique on publicly available databases; this comes down to the privacy of the data and the difficulty of collecting data from hospitals. As regards, the evaluation research (ER) studies used in general case study as an empirical type to test existing techniques on new databases; note that most of the ER studies were in collaboration with hospitals. 4.3

RQ3: Which is the Most Used Machine Learning Objective?

Figure 5 shows that 434 articles investigated the classification objective, 32 papers were for prediction (regression), 19 for clustering, and the remaining 5 articles investigated the association objective [7–11]. The use of classification methods is

52

H. Zerouaoui et al.

explained by the fact that image processing steps include image preprocessing, segmentation, feature extraction, feature selection and classification [12]. Therefore, classification is an important step in IP for classifying properly the medical images to detect the type of the tumor. 4.4

RQ4: What are the Datasets Used for ML and IP in BC?

Table 3 shows that 47% of the selected papers used private datasets collected from hospitals; this is due to the privacy of the medical images and the fact that not all patients want to share their medical images. Researchers are then encouraged to collaborate with clinics and medical centers to collect the required images to evaluate and their BC solutions. Moreover, the most used public datasets are MIAS (28%) and DDSM (25%) for mammographic images due to the fact mammographic images are still the most used for BC diagnosis [13–15]; Breakhis (9%) for histopathological images; and Wisconsin (5%), Inbreast (5%) and BCDR (6%) for other medical imaging types. We note that some studies used several datasets to compare their results [10, 16–19].

5 Conclusion and Future Work The purpose of this SMS was to present an overview of the use of ML and IP in breast cancer. 530 papers published from 2000 to August 2019 were selected and classified according to: year and source of publication, research type and empirical type, BC discipline, ML methods and techniques, validation techniques and performance measures. This paper discussed the results of the six RQs. The findings per RQ are: (RQ1) The use of ML and IP for BC is gaining more interest in the last years by researchers, the number of published articles has increased significantly since 2015 and the majority of the papers (71%) were published in journals. (RQ2) Most of the relevant papers were identified as solution proposal and evaluation research, and the majority of the articles used historical based evaluation. (RQ4) Classification is the most investigated objective in ML and IP for BC, and that is explained by the fact that classification is a component of any IP process. (RQ6) Private datasets are the most frequently used to evaluate ML and IP for BC, followed by two public datasets MIAS and DDSM. As future work we aim to: (1) use the results of the SMS study as the base to perform a systematic literature review concerning the use of ML and IP in Breast cancer, and (2) conduct an evaluation research using case study data collected from a Moroccan hospital to investigate the performance of different ML and IP techniques.

References 1. Metelko, Z., et al.: Pergamon the world health organization quality of life assessment. 41(10) (1995) 2. Bish, A., Ramirez, A., Burgess, C., Hunter, M.: Understanding why women delay in seeking help for breast cancer symptoms B. J. Psychosom. Res. 58, 321–326 (2005)

Machine Learning and Image Processing for Breast Cancer

53

3. Zhang, G., Wang, W., Moon, J., Pack, J.K., Jeon, S.I.: A review of breast tissue classification in mammograms. In: Proceedings of the 2011 ACM Research in Applied Computation Symposium, RACS 2011, pp. 232–237 (2011) 4. Idri, A., Chlioui, I., El Ouassif, B.: A systematic map of data analytics in breast cancer. In: ACM International Conference. Proceeding Series (2018) 5. Hosni, M., Abnane, I., Idri, A., Carrillo de Gea, J.M., Fernández Alemán, J.L.: Reviewing ensemble classification methods in breast cancer. Comput. Methods Programs Biomed. 177, 89–112 (2019) 6. Kofod-petersen, A.: How to do a structured literature review in computer science. Researchgate, no. May 2015, pp. 1–7 (2014) 7. Kitchenham, B., Pearl Brereton, O., Budgen, D., Turner, M., Bailey, J., Linkman, S.: Systematic literature reviews in software engineering - a systematic literature review. Inf. Softw. Technol. 51(1), 7–15 (2009) 8. Tonella, P., Torchiano, M., Du Bois, B., Systä, T.: Empirical studies in reverse engineering: state of the art and future trends. Empir. Softw. Eng. 12(5), 551–571 (2007) 9. Rampun, A., Wang, H., Scotney, B., Morrow, P., Zwiggelaar, R.: School of Computing, Ulster University, Coleraine, Northern Ireland, UK Department of Computer Science, Aberystwyth University, UK. In: 2018 25th IEEE International Conference Image Processing, pp. 2072–2076 (2018) 10. Agarap, A.F.M.: On breast cancer detection: an application of machine learning algorithms on the Wisconsin diagnostic dataset. In: ACM International Conference. Proceeding Series, no. 1, pp. 5–9 (2018) 11. Xiong, X., Kim, Y., Baek, Y., Rhee, D.W., Kim, S.H.: Analysis of breast cancer using data mining & statistical techniques. In: Proceedings of the Sixth International Conference on Software Engineering, Artificial Intelligence, Networking and Parallel/Distributed Computing and First ACIS International Workshop on Self-assembling Wireless Network, SNPD/SAWN 2005, vol. 2005, pp. 82–87 (2005) 12. Sadoughi, F., Kazemy, Z., Hamedan, F., Owji, L., Rahmanikatigari, M., Azadboni, T.T.: Artificial intelligence methods for the diagnosis of breast cancer by image processing: a review. Breast Cancer Targets Ther. 10, 219–230 (2018) 13. Wei, X., Ma, Y., Wang, R.: A new mammography lesion classification method based on convolutional neural network. In: ACM International Conference. Proceeding Series, pp. 39–43 (2019) 14. Ting, F.F., Tan, Y.J., Sim, K.S.: Convolutional neural network improvement for breast cancer classification. Expert Syst. Appl. 120, 103–115 (2019) 15. Torrents-Barrena, J., Puig, D., Melendez, J., Valls, A.: Computer-aided diagnosis of breast cancer via Gabor wavelet bank and binary-class SVM in mammographic images. J. Exp. Theor. Artif. Intell. 28(1–2), 295–311 (2016) 16. Hu, Z., Tang, J., Wang, Z., Zhang, K., Zhang, L., Sun, Q.: Deep learning for image-based cancer detection and diagnosis – a survey. Pattern Recogn. 83, 134–149 (2018) 17. Mini, M.G.: Neural network based classification of digitized mammograms. In: Proceedings of the 2nd Kuwait Conference on e-Services e-Systems, KCESS 2011, pp. 1–5 (2011) 18. Hamidinekoo, A., Dagdia, Z.C., Suhail, Z., Zwiggelaar, R.: Distributed rough set based feature selection approach to analyse deep and hand-crafted features for mammography mass classification. In: Proceedings of the 2018 IEEE International Conference on Big Data, Big Data 2018, pp. 2423–2432 (2019) 19. Mendel, K., Li, H., Sheth, D., Giger, M.: Transfer learning from convolutional neural networks for computer-aided diagnosis: a comparison of digital breast tomosynthesis and full-field digital mammography. Acad. Radiol. 26(6), 735–743 (2019)

A Definition of a Coaching Plan to Guide Patients with Chronic Obstructive Respiratory Diseases Diogo Martinho(&) , Ana Vieira , João Carneiro , Constantino Martins , Ana Almeida , and Goreti Marreiros Research Group on Intelligent Engineering and Computing for Advanced Innovation and Development (GECAD), Institute of Engineering, Polytechnic of Porto, Porto, Portugal {diepm,aavir,jomrc,acm,amn,mgt}@isep.ipp.pt

Abstract. With such a noticeable increase in the number of people with chronic obstructive respiratory diseases the effectiveness of traditional healthcare systems has worsened significantly over the last years. There is an opportunity to develop low cost and personalized solutions that can empower patients to selfmanage and self-monitor their health condition. In this context, the PHE project is present whose main goal is to develop coaching solutions for remote monitoring of patients and that can be provided through the exclusive use of the smartphone. In this work we explore how patients with chronic obstructive respiratory diseases can adopt healthier behaviors by following personalized healthcare coaching plans used throughout their daily lives. We explain how a coaching plan can be defined to guide the patient and explore the mechanisms necessary to operate automatically and adapt itself according to the interactions between the patient and the system. As a result, we believe to be possible to enhance user experience and engagement with the developed system and consequentially improve his/her health condition. Keywords: CORD


mHealth
Personal healthcare
Self-monitoring

1 Introduction Chronic obstructive respiratory diseases (CORD) affect a large percentage of the world’s population and is already the third leading cause of death in the world [1, 2]. Furthermore, is estimated that just in Europe, the cost of respiratory diseases exceeds the €380 billion [3]. Moreover, CORDs are progressive and worsen over time. This means that there is a high prevalence of CORD throughout a person’s life cycle (asthma starting in early ages and other chronic obstructive pulmonary diseases are detected from the middle-age onwards). The progressive deterioration of CORDs often leads to frequent exacerbations, which in turn results in frequent hospital admissions. As such, patients require regular medical consultations and constant monitoring of their health throughout their daily lives. Health care in the context of CORD management has traditionally been provided through either face-to-face interventions between the © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2020 Á. Rocha et al. (Eds.): WorldCIST 2020, AISC 1161, pp. 54–64, 2020. https://doi.org/10.1007/978-3-030-45697-9_6

A Definition of a Coaching Plan to Guide Patients

55

patient and the healthcare professional, separated by periods without structured support or by the use of self-monitoring tools (such as flow meters, handheld spirometers, oximeters) and self-management tools (such as symptom diaries, manuals, pamphlets and web resources) between consultations. The reality, however, is that the constant monitoring of patients’ condition has become a burden on the healthcare providers [4] and traditional healthcare delivered through health professionals’ face-to-face interactions becomes more difficult to achieve. As such, the necessity to develop cost-effective solutions to monitor and treat patients with CORD has increased significantly in recent years [5]. In this scope, concepts such as mobile health (mHealth) have emerged towards the self-management of the patient’s disease, by providing mobile systems that are capable of monitoring patients’ health status and giving customized feedback about activities and behaviors that can be done to improve health and wellbeing [6, 7]. Furthermore, mobile devices now offer a wide set of features and embedded sensors and the development of solutions that can exploit these components without or with minimal access to external devices other than the smartphone itself seem to be adequate and easy to integrate in the daily lives of patients to measure and monitor patients’ current health condition and support them in the management of their diseases [8]. Therefore, coaching solutions delivered through smartphones (mCoaching) that can combine data gathering and processing, gamification elements for user engagement and support to behaviors change seem to be an ideal platform to deliver both simple and effective self-management interventions, while maintaining or improving quality of care and reducing costs, specially in the context of CORD management [9–12]. The work here proposed is part of the PHE project1 which aims to empower people to monitor and improve their health using personal data and technology assisted coaching. To achieve this goal, PHE will apply innovative and intelligent measuring and monitoring tools for preventive healthcare and allow cost-saving and self and home-care solutions with increased patient involvement. Furthermore, PHE project will exclusively use the smartphone and its embedded sensors to acquire all the necessary data to provide personalized support to the CORD patient. In this work we explore the personalization given to the CORD patient by providing him/her a coaching plan to follow and to adopt healthier behaviors throughout his/her daily life. A conceptual definition of the coaching plan is presented which includes four different phases of operation (initialization, execution, completion and post completion). We describe each of these phases and explain how the coaching plan can enhance the personalized healthcare provided to the patient and define a proactive mechanism which is not completely dependent on user input but also capable of adapting itself based on the data collected over time while the patient uses and interacts with the PHE system.

1

https://itea3.org/project/personal-health-empowerment.html.

56

D. Martinho et al.

2 Proposed Model The work here proposed has been extended from [13] in which an architecture for the coaching module to support self-monitoring of CORD patients was defined. This coaching module is responsible for processing patient data and generate recommendations to improve patient’s health condition accordingly. Furthermore, and as will be explained, the proposed model can operate independently from the PHE system due its generic structure. Three main type of users have been identified which interact with the PHE system: patient, healthcare professional and health manager. The first user is the main user and will interact with the developed system by inserting clinical information and receiving recommendations to adopt healthier behaviors and improve health condition and wellbeing. The healthcare professional can access patient clinical information and provide specific guidelines (through coaching plans). The health manager can access and update available domain knowledge (which includes rules and associated variables, recommendations, user profiles and non-specific coaching plans). In this section, we first describe the architecture of the defined coaching module considered for the CORD Management in the PHE system and associated components. The Coaching Plans component is then discussed in more detail as it represents the novel feature proposed in this work. 2.1

PHE Coaching Module

According to Fig. 1, three main layers have been identified for the considered architecture: Service Layer, Business Layer and Data Access Layer. Within the Service Layer, a Web API has been developed to provide a set of services that can be accessed internally within the PHE system, but also externally by other systems. The Business Layer includes four main components which combined allow the definition and model of knowledge regarding a certain domain. The Rules component specifies the set of conditions associated to the patient’s clinical data and that are necessary to identify possible recommendations to send to the patient. These conditions require the

Fig. 1. Coaching module architecture for self-monitoring of CORD

A Definition of a Coaching Plan to Guide Patients

57

validation of different health variables, which in the case of this work correspond to both patient demographic data and health state (for example, gender and weight, smoke exposure, etc.). Besides that, each health variable has an associated periodicity and to measure/collect its current value a mechanism has to also be defined to promote a specific interaction between the patient and the system (for example, to know if an exacerbation was detected within the last week, the associated health variable has to be updated weekly using an interaction mechanism such as a visual notification). The definition of each rule and corresponding recommendation is structured in a clinical matrix format and are based on scientific evidence. Figure 2 shows an example of a rule that was defined for a recommendation to send to the patient.

Fig. 2. Rule example for CORD

User Profiles specifies all the characteristics that can identify a certain profile which is assigned to the patient. So far two main groups have already been identified (Asthma and Rhinitis). Furthermore, the remaining groups will be defined using clustering techniques to identify users sharing characteristics related to patient’s demographic data, context, etc. The Coaching Plans component includes the selected recommendations to be provided to the patient in a given time frame. Furthermore, each coaching plan is related to a specific health topic which has been identified according to the literature on clinical evidence and medical guidelines. This component will be discussed in the following subsection of this work. The recommendations component verifies and processes the received data according to the defined Rules, User Profiles and Coaching Plans and selects suitable recommendations. The Data Access Layer serves as a middle layer between the Business Logic Layer and the different data sources and controls all the read, insert, update and delete operations on the database. The database contains information regarding the patient’s clinical data, health variables associated with recommendations, the history of provided recommendations and respective feedback. It also contains knowledge provided by health professionals, rules used for the generation of recommendations, the defined user profiles and respective characteristics, and the coaching plans provided to each patient. The coaching module has been developed using JBoss Drools framework as it provides intuitive rule language for non-developers, supports flexible and adaptive process, enhances intelligent process automation and complex event processing and is easy to integrate with web services.

58

2.2

D. Martinho et al.

Coaching Plan

According to the literature [1], and in the context of CORD, coaching plans refer to different topics related with the management of the disease. In the case of this work, we have identified 15 minor and 5 major topics through the study of current medical guidelines and clinical evidence to drive individualized coaching. For that we considered the results published in American College of Sports Medicine, American College of Rheumatology, Allergic Rhinitis and its Impact on Asthma, British Thoracic Society, The Association of Chartered Physiotherapists in Respiratory Care, Australian and New Zealand guidelines for the management of chronic obstructive pulmonary disease, Global Initiative for Asthma, Global Initiative for Chronic Obstructive Lung Disease, Royal Dutch Society for Physical Therapy, National Asthma Education and Prevention Program, National Institute for Health and Care Excellence, Direção-Geral da Saúde. Norma sobre Diagnóstico e Tratamento da Doença Pulmonar Obstrutiva Crónica, Portuguese Ministry of Health and U.S. Department of Health and Human Services. Each considered topic is presented in Table 1. Table 1. Topics for CORD management Major topic Chronic Respiratory Diseases Concomitant Diseases Exposition to External Agents Non-pharmacological Therapies Pharmacological Therapies Other

Minor topic(s) Symptoms Respiratory Infections; Sleep Disorders; Rhinitis; Food Allergy Smoking Habits; Occupational Hazards; Allergens Physical Activity and Exercise; Breathing Exercises and Airway Clearance Techniques Adherence and Inhaler Techniques; Devices and Active Principles; Vaccinations Anxiety; Depression; Stress; Nutrition

Four steps have been identified to define a coaching plan in the context of the PHE system: Plan Initialization, Plan Execution, Plan Completion and Plan Post Completion.

A Definition of a Coaching Plan to Guide Patients

59

Fig. 3. Manual coaching plan and goal definition

Plan Initialization. The coaching plan initialization is a process that can configured manually or automatically by the user. Manual coaching plans are defined either by the healthcare professional or the health manager and differ by the fact that they can target a specific patient (coaching plans created by the healthcare professional) or not (coaching plans created by the health manager). Automatic coaching plans are created by the patient himself/herself and are based on the coaching plans defined for the associated user profile. As can be seen in Fig. 3, coaching plan has an associated periodicity which can be weekly, monthly or non-repetitive. The user must then select the topics and intended goals to be achieved with the coaching plan. We define a goal as a desired state regarding a specific patient-related variable according to a certain topic. For example, in the context of smoking habits, one objective could be to decrease the number of cigarettes smoked per day. Furthermore, to achieve a certain goal a list of intermediate goals can also be defined. Following the given example, intermediate goals which would allow the patient to decrease the number of cigarettes smoked per day could be to start the coaching plan and smoke a maximum of 3 cigarettes in the morning, 3 cigarettes in the afternoon and finally 3 cigarettes in the evening/night. This means that when defining a goal and its associated intermediate goals, the user should also define a deadline to achieve each identified goal. The flowchart presented in Fig. 4 shows the coaching plan initialization process that was described.

60

D. Martinho et al.

Fig. 4. Manual coaching plan initialization flowchart

Plan Execution. In the second step, the coaching plan is put into practice and the targeted patient is monitored according to the goals identified. As such, all patientrelated variables considered for the coaching plan are collected through patient interaction with the PHE system by having the patient insert new records and values for those variables. The coaching framework will process those values and whenever a recommendation is verified (if those values trigger all the conditions necessary to activate a recommendation) it will be sent to the smartphone and provided to the patient in different formats (such as an alert or a notification). In parallel, the coaching framework will also verify if any goal established for the coaching plan was achieved and update the coaching plan accordingly. The flowchart presented in Fig. 5 shows the coaching plan execution process that was described.

A Definition of a Coaching Plan to Guide Patients

61

Fig. 5. Coaching plan evaluation flowchart

The ideal time to provide recommendations to the patient will depend on the feedback provided while using the developed system. Several feedback mechanisms are defined to identify the best moments during the day to provide recommendations to the patient and to filter positive recommendations among all the available recommendations: • Recommendation Evaluation – Whenever a detected recommendation is provided to the patient, he/she can rate the same whether they liked or disliked it. This way, unwanted recommendations can be filtered in future similar scenarios. • Goal Evaluation – Whenever patient data is inserted which can modify the current state of a defined goal, it will be evaluated to understand whether the patient was capable of achieving the desired state configured in the coaching plan or if the state associated to an already achieved goal was deteriorated into a previous state. • Patient and System Interaction Evaluation – Different data can be obtained from the interaction between the patient and the system. In this case, it is considered both system utilization rate (which corresponds to utilization times and frequency of use of the system, and response time (verify whether the patient answered a provided recommendation or not and the corresponding response time). This information can then be used to readjust deadlines and understand the most adequate times during the day to interact with the user.

62

D. Martinho et al.

This way it will be possible to avoid unnecessary and very repetitive interactions with the patient which may tire him/her and only increase his/her disinterest to keep using the developed system. All the previous feedback mechanisms are considered in the adaptative goal setting procedure that is executed automatically every day to evaluate and readjust goals based on user performance for that day. For this, we have taken into account the model proposed by Akker and colleagues in [14] where they defined an automated personalized goal-setting feature in the context of physical activity coaching in which they determined the goal line for an upcoming day by combining either stored data from that day of the week or in default parameters defined by the healthcare professional with the new acquired data. We have considered a similar process which updates the coaching plan goals automatically every single day by comparing the current acquired data from that day with the historical data (or with the default parameters in case no data was provided by the user until then) for the same day. We have considered both goals completion rate and average goals’ difficulty as performance measures to identify if the user improved or worsened and depending on the difference between both values the goals for the upcoming days will be updated accordingly. After that we will consider the data obtained from patient and system interaction to measure if an established deadline to achieve a certain goal could also be adapted depending on the average utilization rate and response time that is obtained. Plan Completion. The third step considered is the completion of the defined coaching plan. The condition necessary to complete a defined plan, and as explained above, is whenever the defined goals (excluding all the intermediate goals) have been achieved. After this the patient is provided with a report containing all the information on his/her performance while executing the coaching plan which includes the total number of goals achieved (including all the intermediate goals) and other metrics such as the time needed to achieve those goals, the number of deteriorations verified, the number of generated recommendations while following the coaching plan, the number of approved and disapproved recommendations, among others. Plan Post Completion. The last step is the coaching plan completion in which the achieved results are verified after the plan has been completed. As such, whenever the patient provides more clinical data after he/she has completed a coaching plan, that information will be verified once again to understand if the patient health condition was deteriorated and if any achieved result has been compromised (For example, if the patient completed a smoking cessation coaching plan successfully and then started smoking again). As a result, the healthcare professional will be notified so that he/she can set a new coaching plan for that patient.

3 Conclusions and Future Work The increasing number of people suffering from CORD has led to an overload of healthcare resources to monitor and support patients in the management of their disease. Traditional methods of aiding these patients are no longer cost-effective nor adequate more so when new treatments combining technological developments become more relevant and allow patients to better self-monitor and self-manage their health

A Definition of a Coaching Plan to Guide Patients

63

condition. In this context, mobile coaching technologies can exploit the different features and embedded sensors available on the smartphone and are now being considered as an alternative option to directly monitor patients with CORD. The solution proposed by the PHE system brings further advantages by providing a healthcare solution that does not require any additional external devices other than the smartphone itself and that is therefore more friendly and appellative cost wise to the patient and that can be easily integrated in his/her daily life. In this work we have presented the overall architecture of the coaching module which is integrated in the PHE system and that is composed, among several components, of a coaching plan which is used to guide patients with CORD to adopt better and healthier behaviors. We have provided a conceptual definition of the different phases necessary for this component to operate correctly and explained how it can automatically adapt itself to the user preferences and interactions with the PHE system. As future work we intend to integrate the defined coaching plan in the developed protype for the PHE system and study its effectiveness and usability in a real case scenario. After that, and as we collect more data from the interactions between the patient and the PHE system, we will be able to apply more intelligent mechanisms (predictive analytics) to enhance the interactions and recommendations provided to the user and predict whether a certain interaction or recommendation is adequate at a given moment in time or not. Acknowledgments. The work presented in this paper has been developed under the EUREKA ITEA3 Project PHE (PHE-16040), and by National Funds through FCT (Fundação para a Ciência e a Tecnologia) under the projects UID/EEA/00760/2019 and UID/CEC/00319/2019 and by NORTE-01-0247-FEDER-033275 (AIRDOC - “Aplicação móvel Inteligente para suporte individualizado e monitorização da função e sons Respiratórios de Doentes Obstrutivos Crónicos”) by NORTE 2020 (Programa Operacional Regional do Norte).

References 1. GOLD: Pocket Guide to COPD Diagnosis, Management and Prevention. A guide for Health Care Professionals. 2019 Report (2019) 2. Naghavi, M., Abajobir, A.A., Abbafati, C., Abbas, K.M., Abd-Allah, F., Abera, S.F., Aboyans, V., Adetokunboh, O., Afshin, A., Agrawal, A.: Global, regional, and national agesex specific mortality for 264 causes of death, 1980–2016: a systematic analysis for the Global Burden of Disease Study 2016. Lancet 390, 1151–1210 (2017) 3. European Respiratory Society: The global impact of respiratory disease. Forum of International Respiratory Societies (2017) 4. Gibson, G.J., Loddenkemper, R., Lundbäck, B., Sibille, Y.: Respiratory health and disease in Europe: the new European Lung White Book. European Respiratory Society (2013) 5. Gobbi, C., Hsuan, J.: Collaborative purchasing of complex technologies in healthcare: implications for alignment strategies. Int. J. Oper. Prod. Manag. 35, 430–455 (2015) 6. Steinhubl, S.R., Muse, E.D., Topol, E.J.: Can mobile health technologies transform health care? JAMA 310, 2395–2396 (2013)

64

D. Martinho et al.

7. Luxton, D.D., McCann, R.A., Bush, N.E., Mishkind, M.C., Reger, G.M.: mHealth for mental health: integrating smartphone technology in behavioral healthcare. Prof. Psychol. Res. Pract. 42, 505 (2011) 8. Almeida, A., Amaral, R., Sá-Sousa, A., Martins, C., Jacinto, T., Pereira, M., Pinho, B., Rodrigues, P.P., Freitas, A., Marreiros, G.: FRASIS-Monitorização da função respiratória na asma utilizando os sensores integrados do smartphone. Revista Portuguesa de Imunoalergologia 26, 273–283 (2018) 9. Deterding, S., Sicart, M., Nacke, L., O’Hara, K., Dixon, D.: Gamification. Using gamedesign elements in non-gaming contexts. In: Extended Abstracts on Human Factors in Computing Systems, CHI 2011, pp. 2425–2428. ACM (2011) 10. Tinschert, P., Jakob, R., Barata, F., Kramer, J.-N., Kowatsch, T.: The potential of mobile apps for improving asthma self-management: a review of publicly available and welladopted asthma apps. JMIR mHealth uHealth 5, e113 (2017) 11. Bashshur, R.L., Shannon, G.W., Smith, B.R., Alverson, D.C., Antoniotti, N., Barsan, W.G., Bashshur, N., Brown, E.M., Coye, M.J., Doarn, C.R.: The empirical foundations of telemedicine interventions for chronic disease management. Telemed. e-Health 20, 769–800 (2014) 12. Watson, H.A., Tribe, R.M., Shennan, A.H.: The role of medical smartphone apps in clinical decision-support: a literature review. Artif. Intell. Med. 101707 (2019) 13. Vieira, A., Martinho, D., Martins, C., Almeida, A., Marreiros, G.: Defining an architecture for a coaching module to support self-monitoring of chronic obstructive respiratory diseases. Stud. Health Technol. Inform. 262, 130–133 (2019) 14. Cabrita, M., op den Akker, H., Achterkamp, R., Hermens, H.J., Vollenbroek-Hutten, M.M.: Automated personalized goal-setting in an activity coaching application. In: SENSORNETS, pp. 389–396 (2014)

Reviewing Data Analytics Techniques in Breast Cancer Treatment Mahmoud Ezzat1 and Ali Idri1,2(&) 1

Complex Systems Engineering and Human Systems, Mohammed VI Polytechnic University, Benguerir, Morocco [email protected] 2 Software Project Management Research Team, ENSIAS, Mohammed V University in Rabat, Rabat, Morocco [email protected]

Abstract. Data mining (DM) or Data Analytics is the process of extracting new valuable information from large quantities of data; it is reshaping many industries including the medical one. Its contribution to medicine is very important particularly in oncology. Breast cancer is the most common type of cancer in the world and it occurs almost entirely in women, but men can get attacked too. Researchers over the world are trying every day to improve, prevention, detection and treatment of Breast Cancer (BC) in order to provide more effective treatments to patients. In this vein, the present paper carried out a systematic map of the use of data mining technique in breast cancer treatment. The aim was to analyse and synthetize studies on DM applied to breast cancer treatment. In this regard, 44 relevant articles published between 1991 and 2019 were selected and classified according to three criteria: year and channel of publication, research type through DM contribution in BC treatment and DM techniques. Of course, there are not many articles for treatment, because the researchers have been interested in the diagnosis with the different classification techniques, and it may be because of the importance of early diagnosis to avoid danger. Results show that papers were published in different channels (especially journals or conferences), researchers follow the DM pipeline to deal with a BC treatment, the challenge is to reduce the number of non-classified patients, and affect them in the most appropriate group to follow the suitable treatment, and classification was the most used task of DM applied to BC treatment. Keywords: Data mining
Knowledge data discovery treatment
Medical informatics


Breast cancer

1 Introduction Breast cancer is the most common cancer and cause of death among women every year [1]; it often causes confusion as to the adequate treatment to be adopted in different cases. The field of BC treatment using DM techniques has known an important progress and researchers has become more interested about the topic, since the medical decision-makers require to be supported by DM techniques. Every day, the occurrence of BC is increasing [2], and researchers should be aware of that, so studies in this sense, © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2020 Á. Rocha et al. (Eds.): WorldCIST 2020, AISC 1161, pp. 65–75, 2020. https://doi.org/10.1007/978-3-030-45697-9_7

66

M. Ezzat and A. Idri

especially in the treatment task should be richer. As the treatments for BC are improving [3], patients could live longer with even the most advanced BC. Nowadays since it is possible to access BC medical data, and given the powerful DM techniques supporting the decision making in treatment, we could establish a strong strategy to deal with BC treatment using either precision medicine and/or DM tools to draw the roadmap through a robust decision making framework [3]. The most types of treatments in BC are surgery, radiotherapy, chemotherapy, hormone therapy and biological therapy [4, 5]. The appeal to DM in medical field is increasing, in particular in BC [4]. This is because DM is providing a variety of techniques and tools dealing with complex problems [6]. In fact, DM could be defined as the process of browsing data to extract useful knowledge. DM could behave under two faces, either machine learning using artificial intelligence techniques, or statisticalbased techniques. BC treatment got advantage of the variety of DM objectives (classification, regression, clustering and association) to provide useful solution to oncologists [6]. However, according to the best of authors’ knowledge, no systematic mapping study (SMS) was carried out to synthesize and summarize the findings of primary studies dealing with the use of DM techniques for BC treatment, which motivates the present study. Thus, the present study conduct a systematic map on primary studies published in SpringerLink, PubMed, ACM, Google scholar, Science Direct and IEEExplore between the period of 1991 to 2019. A set of 44 papers were selected, synthesized, and classified according to: year and channel of publication, research type, and DM techniques used. The paper is composed of 5 sections. Section 2 shows the methodology followed to carry out the present SMS. Section 3 presents the results of research questions. Section 4 discusses the results. Finally, conclusion and future work are presented in Sect. 5.

2 Research Methodology The goal of a SMS is to build a classification scheme to structure a field of interest [4]. Whilst SMS involves a horizontal approach to the published studies, Systematic Literature Review (SLR) discusses and analyses the processes and outcomes of previous works vertically. The SMS process can be summed up in five steps: Defining the research question, conducting a search, screening the papers, assigning keywords to each paper by using the abstract and data extracting and mapping results. 2.1

Research Questions and Search String

The aim of this SMS is to establish a broad idea of the published studies on the use of DM techniques to deal with BC treatments. For a broad perspective of the topic, we translated the global goal to three research questions, shown on Table 1, with their rationales. After defining the research questions, a search string is needed to find the most relevant papers for the analysis. Therefore, we targeted the candidate papers through looking forward six digital libraries: IEEExplore, PubMed, SpringerLink, Science Direct, ACM and Google Scholar. We elaborated a search string to excel the operation

Reviewing Data Analytics Techniques in Breast Cancer Treatment

67

Table 1. Research question Research question ID RQ1

RQ2

RQ3

Content

Rationale

What are the publications sources and in which years were the selected studies related to data-mining application for Breast cancer treatment published? How far data mining has contributed in making decisions on breast cancer treatment? What are the most common DM techniques and methods to deal with BC treatment?

To indicate whether there are specific publication channels and when effort regarding this research area was made To discover the type of contribution of DM to the field of BC treatment To identify the most common DM techniques investigated in BC treatment

of search, and this was done by gathering terms and synonyms figured in the RQs. We associated OR the alternatives and AND to link most present terms. The resulted search string was: (Breast OR “Mammary gland” OR “Chemotherapy” OR “mammography”) AND (cancer* OR tumor OR malignancy OR masses). AND (treatment OR cure OR medication OR Prognosis) OR (Identification OR Analysis OR monitoring) AND (data mining* OR machine learning* OR analytics* OR categorization* OR intelligent OR classificat* OR cluster* OR associat* OR predict*) AND (model* OR algorithm* OR technique* OR rule* OR method* OR tool* OR framework* OR recommend). 2.2

Study Selection

The selection process allows us to filter the most relevant papers for our SMS knowing their titles, abstract and keywords. In this context, we evaluated candidate papers in terms of several inclusion and exclusion criteria (IC/EC) to select the relevant ones in order to answer the RQs. We note that we applied OR for IC, AND for EC. Hereafter the ICs and ECs we used in this study: • IC1: Papers proposing new or using existing machine learning techniques for BC treatment. • IC2: Papers presenting an overview on the use of machine learning techniques for BC treatment. • IC3: Papers providing empirical/theoretical comparisons of machine learning techniques for BC treatment. • IC4: Papers published between 1991 and February 2019. • EC1: Papers written in other languages than English. • EC2: Papers dealing with other types of cancer than BC. • EC3: Papers dealing with BC diagnosis, screening or prognosis.

68

M. Ezzat and A. Idri

• EC4: Duplicated papers. • EC5: Short papers (with only 2–3 pages). • EC6: Presentations or posters. 2.3

Data Extraction

A data extraction strategy was established by filling a form through which the most relevant information was selected to answer the RQ: • RQ1: requires the publication source, channel, date, author and abstract relatively to each of the selected studies. • RQ2: The selected papers can be classified into the following types according to the topic they introduced [4]: ✓ Evaluation Research (ER): Evaluation of a DM technique to deal with BC treatment was adopted. ✓ Solution Proposal (SP): Proposition of a new DM approach for the treatment of BC. ✓ Experience Papers (EP): The researchers report their result when experiencing a DM tool or technique applied to BC treatment. ✓ Review: Works mapping the present situation of BC treatment with DM. • RQ3: Requires identification of DM techniques used in previous BC treatment literature. Data extraction bias: We all know the sensitivity of this task of data extraction, we must be very attentive during this step. To avoid bad data extraction, we use an excel file to evaluate carefully the selected papers able to feed our research questions.

3 Results This section reports the results relatively to the research questions of the Table 1. To do so, we present first an overview of the selection process, then the outcome of RQs 1–3. 3.1

Selection Process

Figure 1 shows that 300 candidate papers were found using the search string applied to the six digital libraries. After filtering using the ECs, we retained 139 papers, then we discarded 59 studies using the ICs. After that, eliminated duplicated studies to finally select 44 papers to answer the RQs 1–3.

Reviewing Data Analytics Techniques in Breast Cancer Treatment

69

Fig. 1. Selection process.

3.2

RQ1: What Are Publications Sources and in Which Years Were the Selected Studies Related to Data-Mining Application for Breast Cancer Treatment Published?

Table 2 shows that the 44 selected papers were published in different channels (especially journals or conferences). 44.72% of studies were published in journals, 38.64% were found in conferences, while 28.36% had books or reports as a source. Table 2. Publications sources Source Conference ACM-BCB: ACM International Conference on Bioinformatics, Computational Biology, and Health Informatics Annual International Conference of the IEEE Engineering in Medicine and Biology – Proceedings International Conference on Information and Knowledge Management, Proceedings Other conferences Total Conferences Journals Breast Cancer Research and Treatment Journal of Medical Systems Journal of Biomedical Informatics Other journals Total Journals Other sources

# of papers

(%)

2

4.55

1

2.27

1

2.27

13 17

29.55 38.64

3 3 2 13 21 6

6.38 6.38 4.26 27.68 44.72 16.64

70

M. Ezzat and A. Idri

We observe from Table 2 that Breast Cancer Research and Treatment (6.38%), Journal of Medical Systems (6.38%) and Journal of Biomedical Informatics (4.26%) are the most targeted journals, and the ACM-BCB conference has published only two papers (4.55%). We observe from Fig. 2 that researchers finally tend to focus on BC treatment. This could explain the inefficient work regarding BC treatment in the past. However, the number of studies showed an increasing rate from the year 2016 to 2019; 40% of studies done in 2016 were published in conferences, in 2017, 66.6% were published in a conference, and during 2019, 100% of studies were published in journals.

Fig. 2. Distribution of papers over years and sources.

3.3

RQ2: How Far Data Mining Has Contributed in Making Decisions on Breast Cancer Treatment?

The selected papers could be divided into four research types: Evaluation Research (ER) [8], Solution Proposal (SP) [7], Experience Papers or empirical evaluation (EP) [9] and Reviews [10]: 46.51% of selected papers were SP, proposing new DM tools or techniques dealing with BC treatment, 14% are ER, 16% were considered as R, and 23.26% of the papers were classified as Experience [11]. Note that the Experience studies are in general difficult to carry out due to the difficulty of getting data of patients, moreover it needs a medical expertise to evaluate and validate the outcome. The nature of the treatment task requires the experience, so we need solution proposal, experience papers which will be evaluated in ER papers, and we need studies that will gather all this in reviews.

Reviewing Data Analytics Techniques in Breast Cancer Treatment

71

5 4 3 2 1 0 1991 2001 2005 2007 2008 2010 2011 2012 2013 2014 2015 2016 2017 2019 SP

ER

EP

R

Fig. 3. Types of contributions over years

3.4

RQ3: What Are the Most Commons Techniques and Methods to Deal with BC Treatment?

Figure 4 shows the distributions of DM techniques used according to each DM objective. The four objectives were investigated with: Association: 21.4%, Classification: 52.3%, Clustering: 9.5%, and Prediction: 16.6%. We observe that DT [12–15] is the most used DM technique for classification with a percentage of 50%, followed by Fuzzy logic based models (18%), then SVM [16] (18.1%) and association rules [17] (18.1%), after that we found Neural networks [18] GA [18] and BN [19] with a percentage of 4.5% for each. For the clustering objective, K-means [20] (75%) is the most frequent followed by association rules with a percentage of 25%, As for prediction Neural networks and Decision trees are the most used with a percentage of 43% each, followed by association rules (14.3%). Association rules are the most present when it comes to association with a percentage of (55.6), then fuzzy logic-based models and Apriori with a percentage of 22.2% each.

Fig. 4. Distribution of DM techniques per objective (ARM: Association rules mining. FM: Fuzzy methods. BN: Bayesian network. DT: Decision tree. GA: Genetic algorithm. NN: Neural networks. SVM: Support vector machine).

72

M. Ezzat and A. Idri

4 Discussion This section discusses the results of this systematic map of data analytics in BC treatment. It also analyses the results obtained for each RQ. 4.1

RQ1: What Are Publication Sources, and in Which Years Were the Selected Studies Related to Data-Mining Application for Breast Cancer Treatment Published?

This study selected 44 relevant articles dealing with DM techniques for BC treatment. The variety of sources could be explained by the variety of DM techniques and objectives to solve real world problems. These sources have relationship with computer science, data analytics applied to BC treatment. Even though the number of studies in BC treatment is still low, it has been increased in the last years. Therefore, we conclude that the coming years will come up with very interesting outcomes in treatment of BC using DM techniques; based on the result of Fig. 3 where SP is the highest type of studies. 4.2

RQ2: How Far Data Mining Has Contributed in Making Decisions on Breast Cancer Treatment?

Figure 3 shows that the number of SP is the highest, which can be explained by the fact that the DM based solutions could bring an interesting push for BC treatment studies. SP has reached its top in 2017, whereas experience papers start to show up remarkably in 2017; the evaluation researches are not present due to the lack of empirical evaluations to assess the treatment task. This shows that the use of DM techniques in BC treatment is still not mature. EP studies are presents and this is very important in the medical context, because it gives more credibility to any study guiding to the suitable treatment. 4.3

RQ3: What Are the Most Common DM Techniques to Deal with BC Treatment?

Several DM techniques were evaluated, and Decision tree is the most frequent DM technique used when it comes to the classification objective which is the most recurrent task in BC treatment. Moreover, the association rules are the best choice for researchers to deal with association task; we could also note that the famous k-means still the most powerful technique for clustering issues, whereas the neural networks still very accurate for prediction problems. We could explain that, by the fact that decision trees are faster and easier to use for classification, especially with the large choice of libraries offering the possibility to get advantage of this technique by reusing the functionalities with a large choice of languages. In addition, decision trees can be easily interpreted by the oncologist while taking the decision, without any background of data mining. As for association, the association rules, still the most preferred technique among researchers for the association task, they are easily understood for oncologists who can therefore trust them

Reviewing Data Analytics Techniques in Breast Cancer Treatment

73

when deciding on the BC treatments. For clustering, k-means is still the most powerful and popular clustering techniques, because it is easy to use, and can provide accurate clustering [20]. For the prediction, neural networks were widely used due to their robustness to model complex relationships and their flexibility to be adapted to more complex situations [6–8]. Since there is no DM technique that can outperform all the others in all contexts, many selected studies combined more than two DM techniques to deal with a specific DM objective for BC treatment [21]. Also, combining more than two techniques allows avoiding limitations and consolidating advantages of the used techniques [22].

5 Conclusion and Future Work This study carried out a systematic map of data analytics in breast cancer treatment. It summarized and analyzed 44 selected papers published between 1991 and 2019 according to three RQs. The findings per RQ are: (RQ1) The use of DM among researchers has increased during the last years; the number of publications has remarkably increased since 2016; Most of the papers (44.72%) were published in journals and (38.64%) were found in conferences. (RQ2) This SMS found out that the contribution of DM in BC treatment is very low but has increased recently. Therefore, researchers should devote more effort to the treatment task. Most of the selected papers were captured as SP and EP. (RQ3) Classification is the most frequent objective in DM, because the problem is a classification one. For classification, the Decision tree gained more interest during the years followed by fuzzy methods and SVM. As for future work we aim to: (1) Get advantage of this SMS outcome to perform a systematic literature review about the ML techniques investigated in BC treatment. (2) Implement a solution of BC treatment by evaluating the different ML techniques.

References 1. Soria, D., Garibaldi, J.M., Green, A.R., Powe, D.G., Nolan, C.C., Lemetre, C., Ball, G.R., Ellis, I.O.: A quantifier-based fuzzy classification system for breast cancer patients. Artif. Intell. Med. 58, 175–184 (2013). https://doi.org/10.1016/j.artmed.2013.04.006 2. Umesh, D.R., Ramachandra, B.: Association rule mining-based predicting breast cancer recurrence on SEER breast cancer data. In: 2015 International Conference on Emerging Research in Electronics, Computer Science and Technology, ICERECT 2015, pp. 376–380 (2016). https://doi.org/10.1109/ERECT.2015.7499044 3. Alford, S.H., Michal, O.-F., Ya’ara, G.: Harvesting population data to aid treatment decisions in heavily pre-treated advanced breast cancer. Breast 36, S76 (2017). https://doi. org/10.1016/s0960-9776(17)30764-6 4. Idri, A., Chlioui, I., Ouassif, B.E.: A systematic map of data analytics in breast cancer. In: Proceedings of the Australasian Computer Science Week Multiconference on - ACSW 2018, pp. 1–10. ACM Press, Brisband (2018) 5. Breast Cancer (female) - Treatment - NHS Choices. http://www.nhs.uk/Conditions/Cancerof-the-breast

74

M. Ezzat and A. Idri

6. Khrouch, S., Ezziyyani, M., Ezziyyani, M.: Decision System for the Selection of the Best Therapeutic Protocol for Breast Cancer Based on Advanced Data Mining: A Survey. Springer, Cham (2019) 7. Fan, Q., Zhu, C.J., Xiao, J.Y., Wang, B.H., Yin, L., Xu, X.L., Rong, F.: An application of Apriori Algorithm in SEER breast cancer data. In: Proceedings - International Conference on Artificial Intelligence and Computer Intelligence, AICI 2010, vol. 3, pp. 114–116 (2010). https://doi.org/10.1109/AICI.2010.263 8. Tran, W.T., Jerzak, K., Lu, F.-I., Klein, J., Tabbarah, S., Lagree, A., Wu, T., RosadoMendez, I., Law, E., Saednia, K., Sadeghi-Naini, A.: Personalized breast cancer treatments using artificial intelligence in radiomics and pathomics. J. Med. Imaging Radiat. Sci. 50, 1– 10 (2019). https://doi.org/10.1016/j.jmir.2019.07.010 9. Shen, S., Wang, Y., Zheng, G., Jia, D., Lu, A., Jiang, M.: Exploring rules of traditional Chinese medicine external therapy and food therapy in treatment of mammary gland hyperplasia with text mining. In: Proceedings - 2014 IEEE International Conference on Bioinformatics and Biomedicine, IEEE BIBM 2014, pp. 158–159 (2014). https://doi.org/10. 1109/BIBM.2014.6999347 10. Ondrouskova, E., Sommerova, L., Nenutil, R., Coufal, O., Bouchal, P., Vojtesek, B., Hrstka, R.: AGR2 associates with HER2 expression predicting poor outcome in subset of estrogen receptor negative breast cancer patients. Exp. Mol. Pathol. 102, 280–283 (2017). https://doi. org/10.1016/j.yexmp.2017.02.016 11. Oskouei, R.J., Kor, N.M., Maleki, S.A.: Data mining and medical world: breast cancers’ diagnosis, treatment, prognosis and challenges. Am. J. Cancer Res. 7, 610–627 (2017) 12. Razavi, A.R., Gill, H., Ahlfeldt, H., Shahsavar, N.: Predicting metastasis in breast cancer: comparing a decision tree with domain experts. J. Med. Syst. 31, 263–273 (2007). https:// doi.org/10.1007/s10916-007-9064-1 13. Chao, C.M., Yu, Y.W., Cheng, B.W., Kuo, Y.L.: Construction the model on the breast cancer survival analysis use support vector machine, logistic regression and decision tree. J. Med. Syst. 38, 1–7 (2014). https://doi.org/10.1007/s10916-014-0106-1 14. Kuo, W.J., Chang, R.F., Chen, D.R., Lee, C.C.: Data mining with decision trees for diagnosis of breast tumor in medical ultrasonic images. Breast Cancer Res. Treat. 66, 51–57 (2001). https://doi.org/10.1023/A:1010676701382 15. Takada, M., Sugimoto, M., Ohno, S., Kuroi, K., Sato, N., Bando, H., Masuda, N., Iwata, H., Kondo, M., Sasano, H., Chow, L.W.C., Inamoto, T., Naito, Y., Tomita, M., Toi, M.: Predictions of the pathological response to neoadjuvant chemotherapy in patients with primary breast cancer using a data mining technique. Breast Cancer Res. Treat. 134, 661– 670 (2012). https://doi.org/10.1007/s10549-012-2109-2 16. Coelho, D., Sael, L.: Breast and prostate cancer expression similarity analysis by iterative SVM based ensemble gene selection. In: Proceedings of International Conference on Information and Knowledge Management, pp. 23–26 (2013). https://doi.org/10.1145/ 2512089.2512099 17. He, Y., Zheng, X., Sit, C., Loo, W.T.Y., Wang, Z.Y., Xie, T., Jia, B., Ye, Q., Tsui, K., Chow, L.W.C., Chen, J.: Using association rules mining to explore pattern of Chinese medicinal formulae (prescription) in treating and preventing breast cancer recurrence and metastasis. J. Transl. Med. 10(Suppl 1), 1–8 (2012). https://doi.org/10.1186/1479-5876-10s1-s12 18. Hasan, M., Büyüktahtakın, E., Elamin, E.: A multi-criteria ranking algorithm (MCRA) for determining breast cancer therapy. Omega U. K. 82, 83–101 (2019). https://doi.org/10.1016/ j.omega.2017.12.005

Reviewing Data Analytics Techniques in Breast Cancer Treatment

75

19. Turki, T., Wei, Z.: Learning approaches to improve prediction of drug sensitivity in breast cancer patients. In: Proceedings of Annual International Conferences of the IEEE Engineering in Medicine and Biology Society, EMBS, October 2016, pp. 3314–3320 (2016). https://doi.org/10.1109/EMBC.2016.7591437 20. Radha, R., Rajendiran, P.: Using K-means clustering technique to study of breast cancer. In: Proceedings - 2014 World Congress on Computing and Communication Technologies, WCCCT 2014, pp. 211–214 (2014). https://doi.org/10.1109/WCCCT.2014.64 21. Fahrudin, T.M., Syarif, I., Barakbah, A.R.: Feature selection algorithm using information gain based clustering for supporting the treatment process of breast cancer. In: 2016 International Conference on Informatics and Computing, ICIC 2016, pp. 6–11 (2017). https://doi.org/10.1109/IAC.2016.7905680 22. Çakır, A., Demirel, B.: A software tool for determination of breast cancer treatment methods using data mining approach. J. Med. Syst. 35, 1503–1511 (2011). https://doi.org/10.1007/ s10916-009-9427-x

Enabling Smart Homes Through Health Informatics and Internet of Things for Enhanced Living Environments Gonçalo Marques1,2(&) 1

and Rui Pitarma2,3

Instituto de Telecomunicações, Universidade da Beira Interior, 6201-001 Covilhã, Portugal [email protected] 2 Polytechnic Institute of Guarda, 6300-559 Guarda, Portugal [email protected] 3 CISE - Electromechatronic Systems Research Centre, Universidade da Beira Interior, 6201-001 Covilhã, Portugal

Abstract. As people spend most of their time inside buildings, indoor environment quality must be monitored in real-time for enhanced living environments and occupational health. Indoor environmental quality assessment is based on the satisfaction of the thermal, sound, light and air quality conditions. The indoor quality patterns can be directly used to promote health and wellbeing. With the proliferation of the Internet of Things related technologies, smart homes must incorporate monitoring solutions for data acquisition, transmission, and microsensors for several real-time monitoring activities. This paper presents a low-cost and scalable multi-sensor smart home solution based on Internet of Things for enhanced indoor quality considering acoustic, thermal and luminous comfort. The proposed system incorporates three sensor modules for data collection and use Wi-Fi communication technology for Internet access. The system has been developed using open-source and mobile computing technologies for real-time data visualization and analytics. The acquisition modules incorporate light intensity and colour temperature, particulate matter, formaldehyde, relative humidity, ambient temperature and sound sensor capabilities. The results have successfully validated the scalability, reliability and easy installation of the proposed system. Keywords: Ambient assisted living
Enhanced living environments
Health informatics
Indoor environmental quality
Internet of Things
Smart home

1 Introduction The proliferation and continuous technological improvements in numerous fields of computer science contribute every day to the decrease the cost of the smart homes design. The smart home concept aims to address numerous applications and present an effective and efficient method for the digitalization of people’s daily routine activities and to promote health and well-being [1]. Smart homes incorporate an ecosystem of medical systems, which include medical sensors, microcontrollers, wireless © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2020 Á. Rocha et al. (Eds.): WorldCIST 2020, AISC 1161, pp. 76–85, 2020. https://doi.org/10.1007/978-3-030-45697-9_8

Enabling Smart Homes Through Health Informatics and IoT for ELE

77

communication technologies, and open-source software platforms for data visualization and analytics. Therefore, the smart homes present a relevant potential to address several healthcare issues through the incorporation of mobile computing technologies and medical systems. The Internet of Things (IoT) is the concept that involves the ubiquitous presence of a diversity of cyber-physical systems that support sensing and communication capabilities [2]. Ambient Assisted Living (AAL) is a multi-disciplinary domain which is related to new methods for personalized healthcare systems using microcontrollers, sensors, actuators, computer networks, open-source frameworks and mobile computing technologies to design enhanced living environments (ELE) [3, 4]. IoT provides several benefits to smart homes, healthcare and AAL [5]. Smart homes are typically designed to support older adults in order to integrate healthcare systems and real-time monitoring features for enhanced occupational health and well-being. The ELE is a concept closely related to the AAL field. However, ELE are more associated with information and communications technologies than AAL [6]. The smart home concept can be directly associated with the ELE research field as the smart homes incorporate algorithms, platforms, and systems to maintain an independent and autonomous living of older adults for as long as possible. A smart home incorporates a set of hardware and software systems that deliver a wide range of services to improve health and well-being for all the individuals in general and older adults in particular. People typically spend most of their time inside buildings. Therefore, indoor environmental quality (IEQ) must be monitored in real-time for enhanced occupational health and well-being. IEQ assessment is based on the satisfaction of the thermal comfort, sound, light and air quality conditions [7]. Thermal comfort is a primary concern of the occupants and is usually achieved with temperature ranges of 17–30 °C and depends on as physical factors such as humidity and air temperature but also by individual considerations [8]. Therefore, thermal comfort is not easy to measure and study. Acoustic comfort also has a direct impact on people’s health and well-being. The noise effects on health are related to annoyance, sleep and cognitive performance for both adults and children but can also be associated with raised blood pressure [9]. Noise pollution is a risk factor for people who have pregnancy-related hypertension and preeclampsia [10]. Moreover, noise exposure is also associated with cardiovascular disease [11], psychiatric problems and anti-social behavior. People’s health and wellbeing directly depend on their sleep quality which is affected by sound levels [12]. Therefore the developed countries have designed policies and laws for noise regulation [13]. The World Health Organization states that noise exposure is increasing in Europe [14]. Considering the noise pollution effect on health is particularly relevant to monitor the indoor living environments for ELE and occupational health [15]. It is also pertinent to develop more aggressive policies for sound level supervision [16]. Furthermore, most people are concerned about the health problems related to noise pollution and acknowledge the critical need to design efficient and effective mechanisms for noise assessment and control [17].

78

G. Marques and R. Pitarma

The Environmental Protection Agency ranked indoor air quality (IAQ) on the top five environmental risks to public health [18]. Therefore, IAQ monitoring must be a requisite for all buildings. Reduced air quality levels are associated with numerous health effects such as headaches, dizziness, restlessness, difficulty breathing, increase heart rate, elevated blood pressure, coma and asphyxia [19–21]. The indoor light levels are also related to people’s health, well-being [22, 23] and daylight exposure in buildings is also related to energy costs [24]. Luminous comfort corresponds to the individual’s satisfaction regarding the environmental light levels, and thermal comfort depends on physical parameters, which can be measured, such as light intensity and color but also in personalized conditions. People’s attention on this topic has been increased as now is perceived that light levels are directly related to people’s psychological health, performance and productivity [25]. The IEQ assessment can perceive patterns on the indoor living quality, which can be directly used to plan interventions for ELE. Regarding the proliferation of IoT technologies, smart homes must incorporate different monitoring solutions that make use of open source technologies for data acquisition, transmission, and microsensors for several monitoring activities such as noise monitoring, activity recognition, and thermal and light comfort assessment [26–33]. Therefore, this paper presents an integrated solution for IEQ, which provides thermal, acoustic and luminous comfort supervision. This solution incorporates open-source and mobile computing technologies for data consulting and analysis. The rest of the paper is structured as follows: Sect. 2 presents the materials and methods used in the design of the proposed solution; Sect. 3 presents the results and discussion, and the conclusion is presented in Sect. 4.

2 Materials and Methods The system architecture of the proposed multi-sensor smart home solution is presented in Fig. 1. The proposed method uses a native Wi-Fi compatible microcontroller for data acquisition, process and transmission. The data collected is stored in a SQL Service

Fig. 1. System architecture.

Enabling Smart Homes Through Health Informatics and IoT for ELE

79

database using a web application program interface (API) developed in .NET. This API contains the web services to receive and manage the data collected by the microcontroller and also to provide the data output for webpage visualization and analytics features. The proposed method incorporates several sensing features such as light intensity and colour temperature, particulate matter (PM), formaldehyde, relative humidity, ambient temperature and sound level using three sensor modules. Each sensor module is connected to an ESP8266 microcontroller. The sensor selection was conducted with the primary goal of creating an ELE to promote occupational health and enhanced IEQ (Fig. 2).

Fig. 2. The proposed multi-sensor system block diagrams representing the sensor’s components.

The PMS5003ST sensor (Beijing Plantower Co., Ltd., Beijing, China) has been used for air quality and thermal comfort assessment. This sensor supports temperature, humidity, PM and formaldehyde sensing features. It is a 5 V sensor which has a 100 mA and 200 lA for active and standby current consumption and a response time lesser than 10 s. The particle counting efficiency is 98%, the PM2.5 measurement range is 0–2000 ug/m3, and the maximum error is ±10 (PM2.5 100–500 lg/m3). The temperature range is from −10 °C to 50 °C, and the maximum error is ±0.5 °C. The relative humidity range is from 0–90%, and the maximum error is ±2%. Regarding the formaldehyde sensing capabilities, the range is 0–2 mg/m3, and the maximum error is less than ±5% of the output value. The PMS5003ST is connected using the I2C interface. The acoustic comfort is monitored using the calibrated sound sensor (DFRobot, Shanghai, China), which is connected using analogue communication. This sensor has a measurement range of 30 dBA–130 dBA with a measurement error of ±1.5 dBA. The frequency response is 31.5 Hz–8.5 kHz and the response time is 125 ms.

80

G. Marques and R. Pitarma

The TCS3472 sensor (Adafruit Industries, New York, United States) has been selected to monitor the luminous comfort. This sensor can detect RGB light color temperature and intensity levels. This sensor supports high sensitivity and dynamic range, which allow a reliable lighting conditions assessment. This sensor is connected using the I2C interface. The data collected by the proposed solution not only can be used to provide a reliable IEQ assessment of the monitored space but also to support the energy management of the building using the web portal anywhere and anytime. The cost of the systems is presented in Table 1, and the total system cost is below 175 USD. Table 1. System prototype cost. Component ESP8266 PMS5003ST Sound sensor TCS3472 Prototyping case

Units 3 1 1 1 3

Price (USD) 20,97 53,65 60,30 10,19 27,00

The proposed multi-sensor system has been designed using the ESP8266, a lowcost Wi-Fi microchip developed by Espressif Systems in Shanghai, China. This microcontroller incorporates a 32-bit RISC microprocessor core based on the Tensilica Xtensa Diamond Standard 106Micro with 80 MHz clock speed and supports 32 KiB instruction RAM (Fig. 2). The modules are powered using a 230 V–5 V AC-DC 2 A power supply. This smart home system is based on Wi-Fi connectivity for Internet access to provide real-time IEQ data monitoring. Furthermore, the system supports easy Wi-Fi configuration using a Wi-Fi compatible device with a web browser. When the system is connected to a Wi-Fi network, the access credentials are saved on the hardware memory for future access. If no saved network is available, the system enters in hotspot mode, and the user can access this hotspot to configure the Wi-Fi network which the system should be connected. After initialization, the system performs data acquisition, and the data is then processed. If the defined timer is overflowed, the system performs the data transmission and sends the collected data to the database for storage. The sensing activities are performed every 15 s, but this timmer can be updated according to the user’s requirements. Figure 3 represents the flowchart of the sensor modules used in the proposed multi-sensor smart home system.

Enabling Smart Homes Through Health Informatics and IoT for ELE

81

Fig. 3. Flow diagram of the acquisition module used in the proposed multi-sensor system.

3 Results and Discussion The proposed smart home solution supports IAQ, thermal comfort, luminous comfort and acoustic comfort for enhanced occupational health and well-being. The proposed system has been tested in a laboratory of a Portuguese university (Fig. 4). The monitored room is typically occupied per 15 persons, 4 h per day, five days per week and is used for teaching activities. The laboratory is constituted by two rooms. The room has an area around 64 m2 and was monitored in real-time for two months.

Fig. 4. Installation schema of the tests conducted. R – router; 1 – luminous comfort module, 2 – IAQ and thermal comfort module, 3 – acoustic comfort module.

The tests performed have the primary goal of testing the system functional requirements of the proposed monitoring system. The data collected ensures the operability and performance of the proposed smart home system for real-time data collection and visualization.

82

G. Marques and R. Pitarma

Table 2. Comparison of the proposed systems and smart home monitoring solutions available in the literature. Microcontroller

Sensors

Connectivity IAQ Acoustic Luminous Thermal comfort comfort comfort PIC Temperature, relative nRF24L01 √   √ 24F16KA102 [34] humidity and CO2 Arduino UNO CO2 ZigBee √    [35] ZigBee √   √ Waspmote [36] CO, CO2, PM, temperature and relative humidity STM PM, temperature and IEEE √   √ 32F103RC [37] relative humidity 802.15.4 k ESP8266 Temperature, relative Wi-Fi √ √ √ √ [proposed system] humidity, noise, PM, formaldehyde, light

Several cost-effective and open-source monitoring systems are proposed by [34– 38], a summarised comparison review is presented in Table 2. From the analysis of Table 2, it is possible to conclude that the referred solutions are developed with different microcontrollers using Arduino, PIC, Waspmote and STMicroelectronics. The proposed solution uses an ESP8266 with CPU clock speed of 80 MHz, which is higher than used by the authors of [34] PIC (32 MHz), [35] Arduino (16 MHz), [36] Waspmote (14.74 MHz) and [37] STMicroelectronics (72 MHz). Regarding connectivity, all the presented methods referred to in Table 2 incorporate wireless communication technologies. The proposed system uses Wi-Fi as a standard communication methodology implemented in most buildings in developed countries. All the proposed methods presented in Table 2 support IAQ monitoring features; however, acoustic comfort and luminous comfort is not conducted in any of the studies. This smart home system proposed system in this paper provides integrated IEQ monitoring for enhanced occupational health and includes six types of sensorial capabilities, and other sensors can be added for monitoring specific parameters. Furthermore, this system provides an easy installation process that can be done by the enduser. On the one hand, the easy configuration, which avoids installation costs and is based on different modules which lead to enhanced scalability as installation can start using one unit and new modules regarding the needs of the case study. On the other hand, the proposed system is based on open-source technologies, which is particularly significant because other researchers and manufacturers can develop new compatible sensors to be integrated into this smart home solution. However, the proposed smart home solution has some limitations. The prototype appearance needs to be improved, and the proposed method needs additional experimental validation to ensure calibration and accuracy. In the future, the authors aim to test the sensors’ accuracy by study the real average output error and to calculate the real response time for each sensor. Moreover, the primary goal is to make technical improvements, including the

Enabling Smart Homes Through Health Informatics and IoT for ELE

83

development of critical alerts and notifications to notify the building manager when the thermal, acoustic and luminous comfort requirements are not meet.

4 Conclusion In this paper, a low-cost, open-source and scalable multi-sensor smart home solution based on IoT for enhanced IEQ considering acoustic, thermal and luminous comfort, is presented. The proposed method incorporates three sensor modules for data collection and uses Wi-Fi communication technology for Internet access. The data collected is available in real-time for data visualization and analytics through a web portal. This smart home solution provides easy installation and easy Wi-Fi configuration methods. Furthermore, the proposed solution was successfully tested and validated to ensure the functional architecture. The tests conducted present positive results on behalf of an essential contribution to enhanced occupational health and well-being. Furthermore, based on the data collected in the tests performed, we conclude that under certain conditions, IEQ circumstances are significantly lower than those considered healthy for people’s health and well-being. Nevertheless, the proposed needs further experimental validation to ensure calibration and accuracy.

References 1. Wilson, C., Hargreaves, T., Hauxwell-Baldwin, R.: Smart homes and their users: a systematic analysis and key challenges. Pers. Ubiquit. Comput. 19, 463–476 (2015) 2. Marques, G., Pitarma, R., Garcia, N.M., Pombo, N.: Internet of Things architectures, technologies, applications, challenges, and future directions for enhanced living environments and healthcare systems: a review. Electronics 8, 1081 (2019). https://doi.org/10.3390/ electronics8101081 3. Ganchev, I., Garcia, N.M., Dobre, C., Mavromoustakis, C.X., Goleva, R. (eds.): Enhanced Living Environments: Algorithms, Architectures, Platforms, and Systems. Springer, Cham (2019). https://doi.org/10.1007/978-3-030-10752-9 4. Marques, G., Garcia, N., Pombo, N.: A survey on IoT: architectures, elements, applications, QoS, platforms and security concepts. In: Mavromoustakis, C.X., Mastorakis, G., Dobre, C. (eds.) Advances in Mobile Cloud Computing and Big Data in the 5G Era, pp. 115–130. Springer, Cham (2017). https://doi.org/10.1007/978-3-319-45145-9_5 5. Marques, G.: Ambient assisted living and Internet of Things. In: Cardoso, P.J.S., Monteiro, J., Semião, J., Rodrigues, J.M.F. (eds.) Harnessing the Internet of Everything (IoE) for Accelerated Innovation Opportunities, pp. 100–115. IGI Global, Hershey (2019). https://doi. org/10.4018/978-1-5225-7332-6.ch005 6. Dobre, C., Mavromoustakis, C.X., Garcia, N.M., Mastorakis, G., Goleva, R.I.: Introduction to the AAL and ELE systems. In: Ambient Assisted Living and Enhanced Living Environments, pp. 1–16. Elsevier (2017). https://doi.org/10.1016/B978-0-12-805195-5. 00001-6 7. Yang, L., Yan, H., Lam, J.C.: Thermal comfort and building energy consumption implications – a review. Appl. Energy 115, 164–173 (2014). https://doi.org/10.1016/j. apenergy.2013.10.062

84

G. Marques and R. Pitarma

8. Havenith, G., Holmér, I., Parsons, K.: Personal factors in thermal comfort assessment: clothing properties and metabolic heat production. Energy Build. 34, 581–591 (2002). https://doi.org/10.1016/S0378-7788(02)00008-7 9. Stansfeld, S.A., Matheson, M.P.: Noise pollution: non-auditory effects on health. Br. Med. Bull. 68, 243–257 (2003). https://doi.org/10.1093/bmb/ldg033 10. Auger, N., Duplaix, M., Bilodeau-Bertrand, M., Lo, E., Smargiassi, A.: Environmental noise pollution and risk of preeclampsia. Environ. Pollut. 239, 599–606 (2018). https://doi.org/10. 1016/j.envpol.2018.04.060 11. Foraster, M., Eze, I.C., Schaffner, E., Vienneau, D., Héritier, H., Endes, S., Rudzik, F., Thiesse, L., Pieren, R., Schindler, C., Schmidt-Trucksäss, A., Brink, M., Cajochen, C., Marc Wunderli, J., Röösli, M., Probst-Hensch, N.: Exposure to road, railway, and aircraft noise and arterial stiffness in the SAPALDIA study: annual average noise levels and temporal noise characteristics. Environ. Health Perspect. 125, 097004 (2017). https://doi.org/10.1289/ EHP1136 12. Gupta, A., Gupta, A., Jain, K., Gupta, S.: Noise pollution and impact on children health. Indian J. Pediatr. 85, 300–306 (2018). https://doi.org/10.1007/s12098-017-2579-7 13. Zanella, A., Bui, N., Castellani, A., Vangelista, L., Zorzi, M.: Internet of Things for smart cities. IEEE Internet Things J. 1, 22–32 (2014). https://doi.org/10.1109/JIOT.2014.2306328 14. Murphy, E., King, E.A.: An assessment of residential exposure to environmental noise at a shipping port. Environ. Int. 63, 207–215 (2014). https://doi.org/10.1016/j.envint.2013.11. 001 15. Murphy, E., King, E.A.: Environmental noise and health. In: Environmental Noise Pollution, pp. 51–80. Elsevier (2014). https://doi.org/10.1016/B978-0-12-411595-8.00003-3 16. Stansfeld, S.: Noise effects on health in the context of air pollution exposure. Int. J. Environ. Res. Public Health 12, 12735–12760 (2015). https://doi.org/10.3390/ijerph121012735 17. Morillas, J.M.B., Gozalo, G.R., González, D.M., Moraga, P.A., Vílchez-Gómez, R.: Noise pollution and urban planning. Curr. Pollut. Rep. 4, 208–219 (2018). https://doi.org/10.1007/ s40726-018-0095-7 18. Seguel, J.M., Merrill, R., Seguel, D., Campagna, A.C.: Indoor air quality. Am. J. Lifestyle Med. 11(4), 284–295 (2016). https://doi.org/10.1177/1559827616653343 19. Tsai, W.-T.: Overview of green building material (GBM) policies and guidelines with relevance to indoor air quality management in Taiwan. Environments 5, 4 (2017). https://doi. org/10.3390/environments5010004 20. Singleton, R., Salkoski, A.J., Bulkow, L., Fish, C., Dobson, J., Albertson, L., Skarada, J., Ritter, T., Kovesi, T., Hennessy, T.W.: Impact of home remediation and household education on indoor air quality, respiratory visits and symptoms in Alaska native children. Int. J. Circumpolar Health 77, 1422669 (2018). https://doi.org/10.1080/22423982.2017. 1422669 21. Bruce, N., Pope, D., Rehfuess, E., Balakrishnan, K., Adair-Rohani, H., Dora, C.: WHO indoor air quality guidelines on household fuel combustion: strategy implications of new evidence on interventions and exposure–risk functions. Atmos. Environ. 106, 451–457 (2015). https://doi.org/10.1016/j.atmosenv.2014.08.064 22. Azmoon, H., Dehghan, H., Akbari, J., Souri, S.: The relationship between thermal comfort and light intensity with sleep quality and eye tiredness in shift work nurses. J. Environ. Public Health 2013, 1–5 (2013). https://doi.org/10.1155/2013/639184 23. Gropper, E.I.: Promoting health by promoting comfort. Nurs. Forum 27, 5–8 (1992). https:// doi.org/10.1111/j.1744-6198.1992.tb00905.x 24. Xue, P., Mak, C.M., Cheung, H.D.: The effects of daylighting and human behavior on luminous comfort in residential buildings: a questionnaire survey. Build. Environ. 81, 51–59 (2014). https://doi.org/10.1016/j.buildenv.2014.06.011

Enabling Smart Homes Through Health Informatics and IoT for ELE

85

25. Hwang, T., Kim, J.T.: Effects of indoor lighting on occupants’ visual comfort and eye health in a green building. Indoor Built Environ. 20, 75–90 (2011). https://doi.org/10.1177/ 1420326X10392017 26. Marques, G., Roque Ferreira, C., Pitarma, R.: A system based on the Internet of Things for real-time particle monitoring in buildings. Int. J. Environ. Res. Public Health 15, 821 (2018). https://doi.org/10.3390/ijerph15040821 27. Feria, F., Salcedo Parra, O.J., Reyes Daza, B.S.: Design of an architecture for medical applications in IoT. In: Luo, Y. (ed.) Cooperative Design, Visualization, and Engineering, pp. 263–270. Springer, Cham (2016). https://doi.org/10.1007/978-3-319-46771-9_34 28. Marques, G., Pitarma, R.: A cost-effective air quality supervision solution for enhanced living environments through the Internet of Things. Electronics 8, 170 (2019). https://doi. org/10.3390/electronics8020170 29. Marques, G., Ferreira, C.R., Pitarma, R.: Indoor air quality assessment using a CO2 monitoring system based on Internet of Things. J. Med. Syst. 43, 67 (2019). https://doi.org/ 10.1007/s10916-019-1184-x 30. Marques, G., Pitarma, R.: mHealth: indoor environmental quality measuring system for enhanced health and well-being based on Internet of Things. JSAN 8, 43 (2019). https://doi. org/10.3390/jsan8030043 31. Marques, G., Pitarma, R.: Noise monitoring for enhanced living environments based on Internet of Things. In: Rocha, Á., Adeli, H., Reis, L.P., Costanzo, S. (eds.) New Knowledge in Information Systems and Technologies, pp. 45–54. Springer, Cham (2019). https://doi. org/10.1007/978-3-030-16187-3_5 32. Marques, G., Pitarma, R.: Noise mapping through mobile crowdsourcing for enhanced living environments. In: Rodrigues, J.M.F., Cardoso, P.J.S., Monteiro, J., Lam, R., Krzhizhanovskaya, V.V., Lees, M.H., Dongarra, J.J., Sloot, P.M.A. (eds.) Computational Science – ICCS 2019, pp. 670–679. Springer, Cham (2019). https://doi.org/10.1007/978-3-030-227449_52 33. Marques, G., Pitarma, R.: Air quality through automated mobile sensing and wireless sensor networks for enhanced living environments. In: 2019 14th Iberian Conference on Information Systems and Technologies (CISTI), Coimbra, pp. 1–7. IEEE (2019). https:// doi.org/10.23919/CISTI.2019.8760641 34. Shah, J., Mishra, B.: IoT enabled environmental monitoring system for smart cities. In: 2016 International Conference on Internet of Things and Applications (IOTA), Pune, pp. 383– 388. IEEE (2016). https://doi.org/10.1109/IOTA.2016.7562757 35. Salamone, F., Belussi, L., Danza, L., Galanos, T., Ghellere, M., Meroni, I.: Design and development of a nearable wireless system to control indoor air quality and indoor lighting quality. Sensors 17, 1021 (2017). https://doi.org/10.3390/s17051021 36. Bhattacharya, S., Sridevi, S., Pitchiah, R.: Indoor air quality monitoring using wireless sensor network. Presented at the December (2012). https://doi.org/10.1109/ICSensT.2012. 6461713 37. Zheng, K., Zhao, S., Yang, Z., Xiong, X., Xiang, W.: Design and implementation of LPWAbased air quality monitoring system. IEEE Access 4, 3238–3245 (2016). https://doi.org/10. 1109/ACCESS.2016.2582153 38. Gao, Y., Dong, W., Guo, K., Liu, X., Chen, Y., Liu, X., Bu, J., Chen, C.: Mosaic: a low-cost mobile sensing system for urban air quality monitoring. In: IEEE INFOCOM 2016 - The 35th Annual IEEE International Conference on Computer Communications, San Francisco, pp. 1–9. IEEE (2016). https://doi.org/10.1109/INFOCOM.2016.7524478

MyContraception: An Evidence-Based Contraception mPHR for Better Contraceptive Fit Manal Kharbouch1, Ali Idri1,2(&), Taoufiq Rachad1, Hassan Alami3, Leanne Redman4, and Youssef Stelate1 1 Software Project Management Research Team, Department of Web and Mobile Engineering, ENSIAS, Mohamed V University in Rabat, Rabat, Morocco [email protected] 2 CSEHS, University Mohammed VI Polytechnic, Ben Guerir, Morocco 3 Faculty of Medicine, University Mohammed V, Rabat, Morocco 4 Pennington Biomedical Research Center, Baton Rouge, LA 70808, USA

Abstract. The fulfillment of unmet needs for contraception can help women reach their reproductive goals. It was proven to have a significant impact on reducing the rates of unintended pregnancies, and thereby cut the number of morbidity and mortality resulting from these pregnancies, and improving the lives of women and children in general. Therefore, there is a growing concern worldwide about contraception and women’s knowledge of making an advisedchoice about it. In this aspect, an outgrown number of apps are now available providing clinical resources, digital guides, or educational information concerning contraception whether it concerns natural contraception or modern contraception. However, vast amounts of these apps contain inaccurate sexual health facts and non-evidence based information concerning contraception. On these bases, and in respect to the needs of women to effectively prevent unintended pregnancies while conducting a stress-free healthy lifestyle, the World Health Organization (WHO) Medical Eligibility Criteria (MEC) for contraception’s recommendations, and the results and recommendations of a field study conducted in the reproductive health center Les Oranges in Rabat to collect the app’s requirements, we developed an Android app named ‘MyContraception’. Our solution is an evidence-based patient-centered contraceptive app that has been developed in an attempt to facilitate: (1) Seeking evidence-based information along with recommendations concerning the best contraceptive fit (according to one’s medical characteristics, preferences and priorities) helping users make informed decisions about their contraceptive choices. (2) Monitoring one’s own menstrual cycle, fertility window, contraceptive methods usage, and the correlation between these different elements and everyday symptoms in one app. (3) Keeping record of one’s family medical history, medical appointments, analyses, diagnoses, procedures and notes within the same app. In future work, conducting an empirical evaluation of MyContraception solution is intended, to exhaustively examine the effects of this solution in improving the quality of patient-centered contraception care. Keywords: Contraception


mPHR
MEC
WHO
Android

© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2020 Á. Rocha et al. (Eds.): WorldCIST 2020, AISC 1161, pp. 86–94, 2020. https://doi.org/10.1007/978-3-030-45697-9_9

MyContraception: An Evidence-Based Contraception mPHR

87

1 Introduction Although the majority of women seeking out contraceptive measures are most likely to be young, healthy, and with less medical challenges than women over 35 years old, teenagers, or those with intercurrent diseases [1]. Yet, health care providers often prescribe contraceptives to women of reproductive age with core medical conditions as well [2]. Despite the fact that contraceptive counseling can be a challenge overall, it can get more complicated in the presence of concomitant diseases or risk factors [3]. In this vein, women with comorbidities may not receive adequate counseling on contraceptive methods [2]. The first contraception consultation is of crucial importance that it requires a minimum recommended time of 30 min [1]. Beyond sufficient time, offering a wide range of contraceptive methods, evidence-based knowledge of the efficacy, risks, and benefits of the different methods, as well as building a respectful and confidential relationship between the doctor and the women, are key quality features of good contraceptive counseling, allowing women to make informed decisions [3]. However, many health care providers may find this protocol quite intimidating in practice. Consequently, iatrogenic unintended pregnancies are a reality. Since they result from errors or omissions that can be avoided during the consultation, especially the omission of sufficient time [1]. Moreover, obsolete clinical guidelines and lack of knowledge of new evidence can limit both the quality of contraceptive counseling and the user’s access to safe and effective contraception [3]. According to the World Health Organization (WHO), an estimate of 33 million unintended pregnancies over the world are a result of contraceptive failure or incorrect use [4]. At a worldwide level, unintended pregnancy was and still one of the most public health issues; it is considered the main sexual and reproductive health issue associated with the highest risk of morbidity and mortality for women [5]. Women with chronic conditions can have serious health consequences in the event of an unwanted pregnancy. Since pregnancy can aggravate certain diseases or associate them with harmful consequences endangering the life of the woman. In addition, drugs used to treat many chronic diseases are potentially teratogenic [2], affect the development of the embryo and fetus and when exposed to a pregnant woman may cause birth defects, fetal loss or abnormal growth and development [6]. With the fast pace of medical advancement in the reproductive health sector, especially contraception, quick, reliable, and accurate access to evidence-based information is mandatory for health care providers to provide quality care to women based on the most current available evidence [7]. In compliance with the expansion of technology, the number of web and mobile applications (apps) available now to assist clinicians in providing care for women is increasing. It is also becoming increasingly common for women to use technology in the form of websites and apps to monitor and track their cycles for fertility purposes and to inquire about contraception [8]. However, only a few are reliable and exhaustive source of information [9]. In this light and taking advantage of new technologies, we have developed an evidence-based Mobile Personal Health Record (mPHR) to provide interactive, individually tailored information and decision support for contraceptive use. The app is meant to prepare women for their

88

M. Kharbouch et al.

contraception consultations with health care providers and perform as a clinician extender to support the delivery of evidence-based contraception awareness and enhance the overall quality of patient-centered contraception care. The rest of the paper is organized as follows: Sect. 2 explains the Fertility Awareness-Based (FAB) contraception and the WHO’s Medical Eligibility Criteria (MEC) for contraceptive use. MyContraception solution, its purpose, and specifications are detailed in Sect. 3. While Sect. 4 combines the development tools and the implementation of MyContraception solution previously presented in Sect. 3. Finally, Sect. 5 highlights this work’s conclusions and its future perspectives.

2 Theoretical Contraception Aspects 2.1

Fertility Awareness Method

The Fertility Awareness Method (FAM) supported is a form of natural birth control to prevent unwanted pregnancy. FAM-based apps are designed around a statistical algorithm that provides a ‘safe’ or ‘unsafe’ result to the user regarding the risk of pregnancy [10]. The algorithm takes into account the day of ovulation, the yellow phase, the follicular phase, the duration of the cycle, as well as the average temperature between the different phases, and set the safe/unsafe periods. Some apps, support adding luteinizing hormone (LH) test results as well for better accuracy [11]. However, Fertility awareness methods are commonly misperceived as traditional methods and thus are often left out of family planning programming [12]. 2.2

Modern Contraceptive Methods

Modern contraceptive methods are techniques and technologies designed to overcome biology and allow complete sexual freedom while reducing the risk of pregnancy [13]. According to this definition, various products, and medical approaches are defined as modern contraceptives: short-acting contraceptives like pills, injectables and condoms; Long-Acting Reversible Contraceptives (LARC) such as implants and Intrauterine Devices and systems (IUDs); and permanent contraceptive methods so-called sterilization to name few. However, some beliefs and the fear of the side effects of modern contraceptive methods push women to resort to less effective traditional methods. Given this, On the one hand, women with chronic conditions may not be able to safely use traditional contraceptive methods, as the risks associated with pregnancy may be too high. On the other hand, health care providers are often less comfortable prescribing contraception to patients with concomitant conditions, while contraception is often safer than pregnancy for these women [14]. In this regard, the Centers for Disease Control and Prevention (CDC) has developed a MEC for contraception based on the WHO guidelines. Thus, evidence-based recommendations for safe and effective contraceptive methods for women with different medical characteristics and conditions are provided [15].

MyContraception: An Evidence-Based Contraception mPHR

2.3

89

Medical Eligibility Criteria for Contraception

Since 1996, in collaboration with the US Centers for Disease Control and Prevention (CDC), the WHO has been publishing an evidence-based manual referred to as “Medical Eligibility Criteria for contraceptive use (MEC)” [16]. This manual involving a set of medical criteria for the selection of effective contraceptive methods. It is a fourlevel risk classification of various contraceptive methods, not only in certain physiological situations such as postpartum and breastfeeding but also in the presence of concomitant diseases and risk factors. Apps that do implement the WHO’s MEC for contraception are decision aids that add up the scores for each contraceptive method to suggest best-fit choices. As there are no perfect choices when it comes to contraceptive use, these decision aids weigh up different factors concerning the user’s profile and medical history, in order to propose evidence-based suggestions concerning modern contraceptives.

3 MyContraception Solution 3.1

Purpose

The use of contraception has become commonplace in modern society that nearly all women are using contraception at some point in their lifetime [17]. Thus, when seeking contraception, women need a justified, individualized contraceptive counseling in which every decision about a contraceptive method, the advantages and drawbacks are weighed and discussed individually [3]. Moreover, in order to achieve an optimal contraceptive effect and a better adherence rate, Women should be involved in a shared decision-making process [3]. In this respect, the main purpose of MyContraception solution consists of giving women the control and ability to make an informed choice over contraception and to organize and inform many other aspects of their contraception use. All in a convenient, easy and discreet way. The fact of the matter is that these characteristics were recognized to be valued by women when comes to their body decisions according to previous research in the field of health apps [18]. 3.2

Requirements Specification

During the app development process, developers focused on creating a patient-centered contraceptive application, where the entire content of the application is adapted to the contraceptive method chosen from the Selected Practice Recommendations (SPR). Whole in compliance with the following functional requirements. • Download mobile application: A user should be able to download the mobile application through an application repository. The app should be free to download. • Update mobile application: The user should be able to download a new/updated version or release of the app. • User registration: Given that the user has downloaded the app, then she should be able to register by providing login credentials (email, password).

90

M. Kharbouch et al.

• Login: Given that a user has registered, then the user should be able to log in to the mobile application. The log in information will be stored on the phone and in the future, the user should be logged in automatically. • Retrieve password: A user should be able to retrieve her password by email. • Consult ‘About Contraception’: The user should be able to consult the ‘About Contraception’ section to learn more about contraceptive methods, eligibility criteria, efficiency, risks and more. • Enter Menstrual Cycle Information: The user should be able to enter the date of her last menstrual cycle, its length, and duration of period among other information. • Monitor Menstrual Period: The user should be able to track and predict her period, ovulation and know about chances of falling pregnant on a specific day. • Take ‘Eligibility Test’: The user should take an Eligibility Test based on WHO’s MEC for contraception to obtain a list of her best-suited contraceptive methods. • ‘Eligibility Test’ Result: Once the eligibility status identified, the user should obtain information about her recommended contraceptive methods. • Chose a Contraceptive Method: The user should be able to choose one of her recommended contraceptive methods upon which the app will be adapted. • View contraception history: The user should be able to visualize the dated list of her past contraceptive methods. • Receive reminders: The user should be reminded of her ovulation period, to take her pill, schedule a medical checkup… based on her current contraceptive method. • Receive notifications: The user should be notified when it is her predicted first/last day of the period, when her menstrual cycle is abnormal and when she needs to enter some information (symptoms, mood, weight, temperature…). • Change reminders settings: The user should be able to choose how and when she would like to receive reminders based on her current contraceptive method. • Change notification settings: The user should be able to choose how and when she would like to receive notifications. • Archive Medical Notice: The user should be able to scan or upload pictures of her medical notice from her gallery to her medical notice archive on the app and add notes on them. • Archive Medical analysis: The user should be able to scan or upload pictures of her medical analysis from her gallery to her medical analysis archive on the app and add notes on them. • Consult Medical Notice Archive: The user should be able to consult her medical notice archive on the app. • Consult Medical analysis Archive: The user should be able to consult her medical analysis archive on the app. Previous studies had implemented ISO/IEC 25010 standard [19] to health-related software products. Ouhbi et.al had applied this standard on Mobile Personal Health Record (mPHR) [20], while Idri et.al had conducted an evaluation of free mobile personal health records for pregnancy monitoring based on the aforementioned standard [21], and a quality evaluation of gamified blood donation apps using the same standard [22]. Likewise, a set of non-functional requirements was deemed to be improving the software

MyContraception: An Evidence-Based Contraception mPHR

91

quality of MyContraception solution. These requirements are quality characteristics of the ISO/IEC 25010 quality model, and are described as following: • Functional suitability: MyContraception solution should meet users’ needs (stated and/or implied) through well-integrated functions and suitable resulted in content with the needed degree of precision. • Performance efficiency: MyContraception solution should have a short response time to enhance User Experience (UX). • Usability: MyContraception solution should help users to achieve specified goals with effectiveness, efficiency, and satisfaction. • Reliability: MyContraception solution should remain operational and accessible in a specific manner under the possible circumstances (with/without internet connection). • Security: MyContraception solution should secure encrypted communication, protection, and security of users’ accounts and sensitive information. • Maintainability: MyContraception solution should have a readable and composed of discrete components code to easily implement new functions and to avoid introducing defects or degrading existing product quality. 3.3

Integration of Theoretical Contraception Aspects

The app, although a recent development, is based on medical protocols and the WHO guidelines. In this regard, the first step lied in collecting reference-based information about fertility and contraception. The WHO Medical Eligibility Criteria for Contraceptive Use [16] and the WHO Selected Practice Recommendations for Contraceptive Use [23] were the main sources for the scientific basis of this app. Second, the results and recommendations of a field study conducted in the reproductive health center Les Oranges in Rabat to collect the app’s requirements, along with the results and recommendations of an ongoing study reviewing features and functionalities of contraception mPHRs, were elaborated into a Software Requirements Specifications (SRS). The following step was the development of the interfaces. In which the users are able to log their menstrual information and visualize predictions of both their coming periods and ovulation windows, to take an eligibility test and visualize the results of the medical criteria computed, and to organize and inform many other aspects of their contraception use. All in a convenient, easy and discreet way. Subsequently, the algorithms, which track menstrual period and fertility windows and computes the WHO selected practice recommendations of each contraceptive option for all selected medical conditions automatically were implemented. Finally, the app was debugged and tested by the authors.

4 Implementation Our contraception software solution is developed using native Android while data is stored in Firebase cloud service in order to enable data backup, sharing logs, and securing access to the application for privacy concerns. In the current phase of

92

M. Kharbouch et al.

development, the application is dedicated to patients exclusively and not linked to any kind of clinician-centered application. Few user interfaces are shown in the Appendix. The Appendix is accessible at the following link: https://www.um.es/giisw/manal/ Appendix.pdf. When authenticated successfully with his email or an existing login system as depicted in Fig. 1 and Fig. 2 of the Appendix, the user logs in her menstrual information concerning her menstrual cycle as shown in Fig. 3 and Fig. 4 of the Appendix. Then the user is redirected to the home page illustrated in Fig. 5 of the Appendix. From there, the user can: (1) Monitor her period and fertility windows as in Fig. 6 of the Appendix. (2) Log her specific symptoms, mood, measurements, analysis/notices records, journaling and questions for her next obstetric appointment. See Fig. 7 and Fig. 8 of the Appendix. (3) Take an eligibility quiz to obtain her best-fitted contraceptive method based on her age, health condition, and medical history to cite few as referred to in Fig. 11, Fig. 12 and Fig. 13 of the Appendix. Once the user picks her current contraceptive method or chooses one from suggested methods according to her eligibility test results as in Fig. 14 of the Appendix, the whole application is personalized to meet her selected contraceptive method. Moreover, the user can consult her contraceptive history, medical archive, menstruation history, and past obstetric appointments as can be seen in Fig. 9 of the Appendix, consult awareness section about her contraceptive method as described in Fig. 10 of the Appendix, set a reminder for future obstetric appointments as detailed in Fig. 15 of the Appendix, and change the settings of the app. In the settings, as Fig. 16 of the Appendix shows, the user is allowed to customize the content of the app to her liking, choose the language of the app to have a fair understanding of its content, and manage how and when she would like to receive reminders/notifications. The user can log out from the app at any time and navigate smoothly between the different activities thanks to a material design-based menu.

5 Conclusion and Future Perspectives Eager to offer women a patient-centered comprehensive contraceptive counseling and to help them make informed decisions concerning contraception use. An android solution called ‘MyContraception’ was designed on the basis of the WHO’s guidelines concerning contraception and on the needs that raised in a field study results that we conducted in the reproductive health center Les Oranges in Rabat to collect the app’s requirements. The app serves as a clinician extender that compensates for the lack of time, comfort and skills that health care providers may face, and the overwhelming changes in the field of contraception [24] to help overcome barriers to strengthening sexual and reproductive health services [25–34]. In future work, it is intended to conduct an empirical evaluation with real participants to assess the effectiveness of the app in matching women with the contraceptive method best for them, the active engagement of women in monitoring their contraceptive use, and the adoption and integration of this mPHR technology into clinical practice.

MyContraception: An Evidence-Based Contraception mPHR

93

Acknowledgments. This work was conducted within the research project PEER, 7-246 supported by the US Agency for International Development. The authors would like to thank the National Academy of Science, Engineering, and Medicine, and USAID for their support.

References 1. Guillebaud, J.: Contraception Today, 9th edn. CRC Press, Boca Raton (2019) 2. Bonnema, R.A., McNamara, M.C., Spencer, A.L.: Contraception choices in women with underlying medical conditions. Am. Fam. Phys. 82, 621–628 (2010) 3. Moffat, R., Sartorius, G., Raggi, A., et al.: Consultation de contraception basée sur l’évidence. Forum Médical Suisse – Swiss Med. Forum (2019). https://doi.org/10.4414/fms. 2019.08065 4. World Health Organization (2014) WHO | Unsafe abortion: global and regional estimates of the incidence of unsafe abortion and associated mortality in 2008. WHO. https://doi.org/10. 1017/CBO9781107415324.004 5. Kassahun, E.A., Zeleke, L.B., Dessie, A.A., et al.: Factors associated with unintended pregnancy among women attending antenatal care in Maichew Town, Northern Ethiopia, 2017. BMC Res. Notes 12, 1–6 (2019). https://doi.org/10.1186/s13104-019-4419-5 6. Gweneth, L.: Pharmacovigilance in pregnancy. In: Doan, T., Renz, C., Bhattacharya, M., Lievano, F., Scarazzini, L. (eds.) Pharmacovigilance: A Practical Approach, 1st edn, p. 228. Elsevier, Amsterdam (2019) 7. Arbour, M.W., Stec, M.A.: Mobile applications for women’s health and midwifery care: a pocket reference for the 21st century. J. Midwifery Women’s Health (2018). https://doi.org/ 10.1111/jmwh.12755 8. Mendes, A.: What’s new in the world of prescribing contraception? Nurse Prescr. 16, 410– 411 (2018). https://doi.org/10.12968/npre.2018.16.9.410 9. Rousseau, F., Da Silva Godineau, S.M., De Casabianca, C., et al.: State of knowledge on smartphone applications concerning contraception: a systematic review. J. Gynecol. Obstet. Hum. Reprod. 48, 83–89 (2019). https://doi.org/10.1016/j.jogoh.2018.11.001 10. Berglund Scherwitzl, E., Lundberg, O., Kopp Kallner, H., et al.: Perfect-use and typical-use pearl index of a contraceptive mobile app. Contraception 96, 420–425 (2017). https://doi. org/10.1016/j.contraception.2017.08.014 11. Berglund Scherwitzl, E., Gemzell Danielsson, K., Sellberg, J.A., Scherwitzl, R.: Fertility awareness-based mobile application for contraception. Eur. J. Contracept. Reprod. Health Care 21, 234–241 (2016). https://doi.org/10.3109/13625187.2016.1154143 12. Malarcher, S., Spieler, J., Fabic, M.S., et al.: Fertility awareness methods: distinctive modern contraceptives. Glob. Health Sci. Pract. 4, 13–15 (2016) 13. Hubacher, D., Trussell, J.: A definition of modern contraceptive methods. Contraception 92, 420–421 (2015) 14. Chor, J., Rankin, K., Harwood, B., Handler, A.: Unintended pregnancy and postpartum contraceptive use in women with and without chronic medical disease who experienced a live birth. Contraception 84, 57–63 (2011). https://doi.org/10.1016/j.contraception.2010.11.018 15. Curtis, K.M., Tepper, N.K., Jatlaoui, T.C., et al.: U.S. medical eligibility criteria for contraceptive use, 2016. MMWR Recomm. Rep. 65, 1–104 (2016). https://doi.org/10.1089/ jwh.2011.2851 16. WHO: Medical Eligibility Criteria for Contraceptive Use, 5th edn. WHO, Geneva (2015) 17. Daniels, K., Daugherty, J., Jones, J.: Current contraceptive status among women aged 15– 44: United States, 2011–2013. NCHS Data Brief 173, 1–8 (2014)

94

M. Kharbouch et al.

18. Newman, L.: Apps for health: what does the future hold? Br. J. Midwifery 26, 561 (2018). https://doi.org/10.12968/bjom.2018.26.9.561 19. International Organization For Standardization ISO: Software Process Improvement Practice. ISO/IEC 25010:34 (2011) 20. Ouhbi, S., Idri, A., Fern, L.: Applying ISO/IEC 25010 on mobile personal health records. In: 8th International Conference on Health Informatics, pp. 405–412 (2015) 21. Idri, A., Bachiri, M., Fernández-alemán, J.L., Toval, A.: ISO/IEC 25010 based evaluation of free mobile personal health records for pregnancy monitoring. In: IEEE 41st Annual Computing Software Application Conference, pp. 262–267 (2017) 22. Idri, A., Sardi, L., Fernández-alemán, J.: Quality evaluation of gamified blood donation apps using ISO/IEC 25010 standard. In: 12th International Conference Health Informatics, pp. 607–614 (2018) 23. WHO: Selected Practice Recommendations for Contraceptive Use, 3rd edn. WHO, Geneva (2016) 24. Arbour, M.W., Stec, M.A.: Mobile applications for women’s health and midwifery care: a pocket reference for the 21st century. J. Midwifery Women’s Health 63, 330–334 (2018). https://doi.org/10.1111/jmwh.12755 25. Tebb, K.P., Trieu, S.L., Rico, R., et al.: A mobile health contraception decision support intervention for Latina adolescents: Implementation evaluation for use in school-based health centers. J. Med. Internet Res. 21 (2019). https://doi.org/10.2196/11163 26. Sardi, L., Idri, A., Readman, L.M., et al.: Mobile health applications for postnatal care: review and analysis of functionalities and technical features. Comput. Methods Programs Biomed. 184, 1–26 (2020). https://doi.org/10.1016/j.cmpb.2019.105114 27. Bachiri, M., Idri, A., Redman, L.M., et al.: A requirements catalog of mobile personal health records for prenatal care. In: Lecture Notes in Computer Science (Including Subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), pp. 483–495. Springer (2019) 28. Bachiri, M., Idri, A., Abran, A., et al.: Sizing prenatal mPHRs using COSMIC measurement method. J. Med. Syst. 43, 1–11 (2019). https://doi.org/10.1007/s10916-019-1446-7 29. Bachiri, M., Idri, A., Redman, L., et al.: COSMIC functional size measurement of mobile personal health records for pregnancy monitoring. In: Advances in Intelligent Systems and Computing, pp. 24–33. Springer (2019) 30. Bachiri, M., Idri, A., Fernández-Alemán, J.L., Toval, A.: Evaluating the privacy policies of mobile personal health records for pregnancy monitoring. J. Med. Syst. 42, 1–14 (2018). https://doi.org/10.1007/s10916-018-1002-x 31. Idri, A., Bachiri, M., Fernández-Alemán, J.L., Toval, A.: Experiment design of free pregnancy monitoring mobile personal health records quality evaluation. In: 2016 IEEE 18th International Conference on e-Health Networking, Applications and Services, Healthcom 2016. Institute of Electrical and Electronics Engineers Inc., pp. 1–6 (2016) 32. Bachiri, M., Idri, A., Fernández-Alemán, J.L., Toval, A.: Mobile personal health records for pregnancy monitoring functionalities: analysis and potential. Comput. Methods Programs Biomed. 134, 121–135 (2016) 33. Bachiri, M., Idri, A., Fernandez-Aleman, J.L., Toval, A.: A preliminary study on the evaluation of software product quality of pregnancy monitoring mPHRs. In: Proceedings of 2015 IEEE World Conference on Complex Systems, WCCS 2015. Institute of Electrical and Electronics Engineers Inc., pp. 1–6 (2016) 34. Idri, A., Bachiri, M., Fernández-Alemán, J.L.: A framework for evaluating the software product quality of pregnancy monitoring mobile personal health records. J. Med. Syst. 40, 1– 17 (2016). https://doi.org/10.1007/s10916-015-0415-z

Predictors of Acceptance and Rejection of Online Peer Support Groups as a Digital Wellbeing Tool John McAlaney1(&), Manal Aldhayan1, Mohamed Basel Almourad2, Sainabou Cham1, and Raian Ali3 1

2

3

Faculty of Science and Technology, Bournemouth University, Bournemouth, UK {jmcalaney,maldhayan,scham}@bournemouth.ac.uk College of Technological Innovation, Zayed University, Dubai, UAE [email protected] College of Science and Engineering, Hamad Bin Khalifa University, Doha, Qatar [email protected]

Abstract. Digital media usage can be problematic; exhibiting symptoms of behavioural addiction such as mood modification, tolerance, conflict, salience, withdrawal symptoms and relapse. Google Digital Wellbeing and Apple Screen Time are examples of an emerging family of tools to help people have a healthier and more conscious relationship with technology. Peer support groups is a known technique for behaviour change and relapse prevention. It can be facilitated online, especially with advanced social networking techniques. Elements of peer support groups are being already embedded in digital wellbeing tools, e.g. peer comparisons, peer commitments, collective usage limit-setting and family time. However, there is a lack of research about the factors influencing people acceptance and rejection of online peer support groups to enhance digital wellbeing. Previous work has qualitatively explored the acceptance and rejection factors to join and participate in such groups. In this paper, we quantitatively study the relationship between culture, personality, selfcontrol, gender, willingness to join the groups and perception of their usefulness, on such acceptance and rejection factors. The qualitative phase included two focus groups and 16 interviews while the quantitative phase consisted of a survey (215 participants). We found a greater number of significant models to predict rejection factors than acceptance factors, although in all cases the amount of variance explained by the models was relatively small. This demonstrates the need to design and, also, introduce such technique in a contextualised and personalised style to avoid rejection and reactance. Keywords: Online peer groups Behavioural change


Digital addiction
Digital wellbeing


© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2020 Á. Rocha et al. (Eds.): WorldCIST 2020, AISC 1161, pp. 95–107, 2020. https://doi.org/10.1007/978-3-030-45697-9_10

96

J. McAlaney et al.

1 Introduction Digital media including social networks, gaming and online shopping have various benefits and represent an integral part of modern society. Such media empower social connectedness, information exchange and freedom of information exchange introducing a new lifestyle and concepts such as digital humanity and digital citizenships. However, some compulsive and obsessive usage styles and over-reliance on digital media can lead to negative consequences such as reduced involvement in real-life communities and a lack of sleep [1]. Some usage styles can be seen as addictive meeting common criteria of behavioural addiction such as salience, conflict, mood modification, and relapse [2, 3]. There is a limited number of preventative, control and recovery mechanisms available for Digital Addiction (DA). Although the problematic relationship with technology has been recognised in a wide range of literature, DA is still not classified as a mental disorder in the latest 5th edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM 5). Recently, in 2018, the World Health Organization recognised Gaming Disorder, which represents a significant step is searching for preventative and recovery mechanisms. Most of the existing research on DA focuses on the reasons for people to become overly reliant on social media and the relationship of that with factors such as personality traits [4]. Few works have placed software design at the centre of the DA problems, both in facilitating and also in combatting DA, e.g. the digital addiction labels and the requirements engineering for digital well-being requirements in [6, 7]. With the advances in sensing and communication technology and internet connectivity, there has been a proliferation of software and smartphone applications to assist with behavioural change. It is still questionable whether these solutions are effective and whether we understand the acceptance and rejection factors from the users’ perspective. The perception of their role and trustworthiness of such proposed solutions has changed following some failures and the recognition of associated risks [8]. Linking the intention to change behaviour with the act of doing so is the main purpose of behaviour change theories [5]. Peer support groups are one of the approaches to behaviour change which can be utilised to combat addictive behaviours by providing support and helping in relapse prevention [9, 11]. Peer support groups consist of people sharing similar interests and in view of supporting and influencing each other’s behaviour towards achieving common goals [10]. Alrobai et al. [13] focused on the processes involved when running the group, e.g. the roles involved in doing so and the steps to be taken to prevent relapse. Aldhayan et al. [18], explored the acceptance and rejection factors of online peer support groups by people with DA. This exploration was meant to inform the strategies used to introduce such online peer group software, as well as the configuration and governance processes of their online platform. Hsiao Shu and Huang [17] explored the relationships between personality traits and compulsive usage of social media apps, and showed that extraversion, agreeableness, and neuroticism have significant effects on such compulsive usage. Being an online

Predictors of Acceptance and Rejection of Online Peer Support Groups

97

social technique for behaviour change itself, acceptance and rejection of peer support groups could be in turn subject to such personal and environmental factors. In this paper, we study the effect of personality traits, self-control, gender, and perception of usefulness, willingness to join and culture (comparing UK to Middle Eastern users) on the acceptance and rejection factors of online peer support groups. To achieve this target, we designed a survey around the acceptance and rejection factors reported in [18] and derived from two focus groups and 16 interviews. The survey also consisted of various demographics questions and measures for personality [20] and self-control [19]. We collected 215 completed responses. We report on the statistical analysis results and discuss their implications on the design of future online peer support groups to combat DA.

2 Research Method We adopted a mixed-methods approach which consisted of an initial qualitative phase followed by a quantitative one. The participants in both phases self-declared as experiencing problematic digital behaviour and wellbeing issues. 2.1

Qualitative Phase: Exploring Acceptance, Rejection and Governance

We conducted a focus group study of two sessions. The first session aimed at getting insights around how online peer groups are perceived by people self-declaring to have DA and what they wished to see in it. The second focus group served the purpose of identifying the design features of an online peer group platform. For this reason, mock interfaces were made available to the second session participants based on the results of the first focus group. The participants were asked about opinions regarding the mock design and to amend them if needed. The two focus group sessions were conducted with the same six university students; three male and three females, aged between 20 and 26. The participants were a social group in real life, and this was beneficial as it removed concerns regarding trust and privacy during the discussion process. We performed a thematic analysis [12] on the data collected through the sessions. This analysis revealed main factors concerning the acceptance and rejection of this approach as well as its governance styles and process. The objective of the interview stage was to explore in-depth the acceptance and rejection factors and the variability space of designing online peer groups platforms so that we can accommodate different users’ preferences and governance styles. The interview questions were based on the acceptance and rejection factors explored in the focus groups as well as five themes related to governance, including group moderation, feedback and monitoring, membership and exit protocol. We conducted 16 interviews with students who self-declared to have a wellbeing issue around their digital

98

J. McAlaney et al.

behaviour, e.g., obsessive or compulsive use. The sample consisted of 8 males and 8 females, aged between 18 and 35. Each interview lasted between 30 and 40 min. The interviews were transcribed and analysed via thematic analysis [12]. 2.2

Quantitative Phase: Confirmation, Personal and Environmental Influences

This phase was based on a survey that reflected the interview themes, i.e. the acceptance and rejection themes as well as governance themes such as moderator role, feedback, membership and exit procedure. The survey was disseminated both online and in person. A £5 incentive was offered to respondents given the lengthy nature of the survey. We collected 215 completed responses; 105 participants (49%) identified as male and 109 participants (50%) identified as female, with the remaining 1% preferring not to answer on the gender question. The participants were 17 to 55 years old. The survey started with a validation question of whether a participant has wellbeing issues as a precondition to take part. To study the effect of personal and environmental factors on the acceptance and rejection factors, the survey included questions around six factors which were gender (male/female); country/culture (UK/Middle East); perceived usefulness of peer support groups; willingness to join a peer support group; five personality traits [20] (extraversion, agreeableness, conscientiousness, neuroticism and openness); and selfcontrol [19]. We disseminated the survey mainly in the UK, the Kingdom of Saudi Arabia and Syria. We collected 104 completed surveys from KSA and Syria, and 85 from the UK. This allowed us to study statistically whether there was a difference between Middle Eastern culture (KSA and Syria) and Western Culture (UK).

3 Acceptance and Rejection Factors The factors which affect users’ acceptance and rejection of online peer support groups to combat DA are presented in Tables 1 and 2, respectively. The elaborated descriptions of themes A1 to A4 and R1 to R4 can be found in [18]. Further analysis of the data revealed another theme, which is A5. Table 1. Online peer support groups to combat digital addiction: acceptance factors Acceptance theme Sub-themes [A1] Accepting online peer groups as [A1.1] Provide awards: gamification of performance an entertainment auxiliary [A1.2] Peer comparison: to see how I and others do [A1.3] Goal achievement: rewards, information and graphs of my progress towards the goal [A2] Accepting online peer groups as [A2.1] Self-Monitoring: show actual usage and a DA awareness tool performance [A2.2] Peer comparison: benchmarking through others [A2.3] Goal achievement: awareness of how I am achieving goals (continued)

Predictors of Acceptance and Rejection of Online Peer Support Groups

99

Table 1. (continued) Acceptance theme [A3] Accepting online peer support groups as an educational tool

[A4] Accepting online peer support groups as a prevention tool

[A5] Accepting online peer support groups as a support tool

Sub-themes [A3.1] Peer learning: learning from others how to improve [A3.2] Moderator role: learning from moderator, learning from acting as moderator [A3.3] Set up goals: learning how to set up SMART goals [A4.1] Peer feedback: alert/feedback through peer feedback [A4.2] Moderator feedback: alert/feedback by a moderator [A4.3] Authority: steps and restrictions set by a moderator [A5.1] Provide advice: by experienced moderator; alternatives lifestyle [A5.2] Emotional support: when struggling to avoid relapse [A5.3] Feedback: when performing well and underperforming, sending warnings

Table 2. Online peer support groups to combat digital addiction: rejection factors Rejection theme [R1] Rejecting online peer support groups when seen as intimidation tool

[R2] Rejecting online peer support groups when seen as overly judgmental [R3] Rejecting online peer supports group when hosting unmanaged interactions [R4] Rejecting online peer groups due to unclear membership protocol

Sub-themes [R1.1] Negative feedback: dismissive feedback when failing [R1.2] Harsh penalty, e.g. banning and locking out [R2.1] Being overly judged by a moderator [R2.2] Being judged by peers, known and unknown in person [R3.1] Weak management [R3.2] Large group size [R4.1] Relatedness: group including relatives and friends [R4.2] Exit control: free and uncontrolled exit as well as conditions on exiting the group without considering others

4 Personal and Cultural Effects on Acceptance and Rejection The survey questions around acceptance and rejection can be found in Appendix A. A Likert scale indicating level of agreements was used for each of the statements under each theme. A series of linear multiple regressions using the enter method were

100

J. McAlaney et al.

conducted. In each model the predictors were gender (male/female); region (UK/Middle East); perceived usefulness of peer support groups; willingness to join a peer support group; the five personality trait scores of extraversion, agreeableness, conscientiousness, neuroticism and openness; and self-control score. For each model, the outcome measure was the individual questions used to measure attitudes relating to the acceptance and rejection factors of online peer groups, as identified within the description of each model result in the section below. Multicollinearity diagnostics were conducted prior to the analysis to determine the suitability of conducting multiple regressions. 4.1

Effects on Acceptance Factors

[A1] Accepting online peer groups as an entertainment auxiliary. Three models under this category were non-significant, which were [A1.1a] Awards when achieving behavioural targets, e.g. points, badges, etc.; [A1.1b] Awards when making progress towards the behavioural target; [A1.3] Information and graphs how I am progressing to keep me engaged. The model for [A1.2] Peer comparisons, i.e. to see how I and others are performing was significant, predicting 12% of the variance (R2 = .12, F(10,159) = 2.16, p