HEALTHCARE TECHNOLOGIES SERIES 50 Explainable Artificial Intelligence in Medical Decision Support Systems IETBookSeriesone–HealthTechnologies BookSeriesEditor:ProfessorJoelJ.P.C.Rodrigues,CollegeofComputerScienceand Technology,ChinaUniversityofPetroleum(EastChina),Qingdao,China;SenacFacultyof Ceara´,Fortaleza-CE,BrazilandInstitutodeTelecomunicac¸o˜es,Portugal BookSeriesAdvisor:ProfessorPranjalChandra,SchoolofBiochemicalEngineering,Indian InstituteofTechnology(BHU),Varanasi,India Whilethedemographicshiftsinpopulationsdisplaysignificantsocio-economicchallenges,they triggeropportunitiesforinnovationsine-Health,m-Health,precisionandpersonalized medicine,robotics,sensing,theInternetofthings,cloudcomputing,bigdata,softwaredefined networks,andnetworkfunctionvirtualization.Theirintegrationishoweverassociatedwith manytechnological,ethical,legal,social,andsecurityissues.Thisbookseriesaimsto disseminaterecentadvancesfore-healthtechnologiestoimprovehealthcareandpeople’s wellbeing. Couldyoubeournextauthor? Topicsconsideredincludeintelligente-Healthsystems,electronichealthrecords,ICT-enabled personalhealthsystems,mobileandcloudcomputingfore-Health,healthmonitoring, precisionandpersonalizedhealth,roboticsfore-Health,securityandprivacyine-Health, ambientassistedliving,telemedicine,bigdataandIoTfore-Health,andmore. Proposalsforcoherentlyintegratedinternationalmulti-authorededitedorco-authored handbooksandresearchmonographswillbeconsideredforthisbookseries.Eachproposalwill bereviewedbythebookSeriesEditorwithadditionalexternalreviewsfromindependent reviewers. Todownloadourproposalformorfindoutmoreinformationaboutpublishingwithus,please visithttps://www.theiet.org/publishing/publishing-with-iet-books/. PleaseemailyourcompletedbookproposalfortheIETBookSeriesone-HealthTechnologies to:[email protected][email protected]. Explainable Artificial Intelligence in Medical Decision Support Systems Edited by Agbotiname Lucky Imoize, Jude Hemanth, Dinh-Thuan Do and Samarendra Nath Sur The Institution of Engineering andTechnology PublishedbyTheInstitutionofEngineeringandTechnology,London,UnitedKingdom TheInstitutionofEngineeringandTechnologyisregisteredasaCharityinEngland& Wales(no.211014)andScotland(no.SC038698). †TheInstitutionofEngineeringandTechnology2022 Firstpublished2022 ThispublicationiscopyrightundertheBerneConventionandtheUniversalCopyright Convention.Allrightsreserved.Apartfromanyfairdealingforthepurposesofresearch orprivatestudy,orcriticismorreview,aspermittedundertheCopyright,Designsand PatentsAct1988,thispublicationmaybereproduced,storedortransmitted,inany formorbyanymeans,onlywiththepriorpermissioninwritingofthepublishers,orin thecaseofreprographicreproductioninaccordancewiththetermsoflicencesissued bytheCopyrightLicensingAgency.Enquiriesconcerningreproductionoutsidethose termsshouldbesenttothepublisherattheundermentionedaddress: TheInstitutionofEngineeringandTechnology FuturesPlace KingsWay,Stevenage HertfordshireSG12UA,UnitedKingdom www.theiet.org Whiletheauthorsandpublisherbelievethattheinformationandguidancegiveninthis workarecorrect,allpartiesmustrelyupontheirownskillandjudgementwhenmaking useofthem.Neithertheauthornorpublisherassumesanyliabilitytoanyoneforany lossordamagecausedbyanyerrororomissioninthework,whethersuchanerroror omissionistheresultofnegligenceoranyothercause.Anyandallsuchliabilityis disclaimed. Themoralrightsoftheauthortobeidentifiedasauthorofthisworkhavebeen assertedbyhiminaccordancewiththeCopyright,DesignsandPatentsAct1988. BritishLibraryCataloguinginPublicationData AcataloguerecordforthisproductisavailablefromtheBritishLibrary ISBN978-1-83953-620-5(hardback) ISBN978-1-83953-621-2(PDF) TypesetinIndiabyMPSLimited PrintedintheUKbyCPIGroup(UK)Ltd,Croydon CoverImage:JorgGreuel/PhotodiscviaGettyImages Contents AbouttheEditors xv Preface xix Acknowledgments xxi 1 Explainable artificial intelligence (XAI)inmedical decision systems (MDSSs):healthcare systems perspective 1 Oluwafisayo Babatope Ayoade, Tinuke OmolewaOladele, Agbotiname Lucky Imoize, Joseph Bamidele Awotunde, Adetoye Jerome Adeloye, Segun Omotayo Olorunyomi and Ayorinde Oladele Idowu 1.1 Introduction 2 1.2 Overview of HMDSSs 4 1.2.1 MDSSsin healthcare system 5 1.2.2 Basisof HMDSS 7 1.2.3 Characterizing and categorizing HMDSS 8 1.3 Case studyof XAIenabled with MDSSsinvarious infectious diseases 13 1.3.1 SCD 13 1.3.2 Diabetes mellitus (DM) 20 1.3.3 Hypertensive retinopathy (HR) 22 1.3.4 Carcinoma 24 1.3.5 COVID-19pandemic 26 1.4 XAIresearch trends and open issues 29 1.4.1 XAIperspective in healthcare 30 1.5 Conclusion and future directions 31 Acknowledgment 31 References 32 2 Explainable artificial intelligence (XAI)inmedical decision support systems(MDSS):applicability, prospects,legal implications, and 45 challenges JosephBamideleAwotunde,EmmanuelAbidemiAdeniyi,SundayAdeola Ajagbe,AgbotinameLuckyImoize,OlukayodeAyodeleOkiandSanjayMisra 2.1 Introduction 46 2.1.1 Chapter organization 48 2.2 MDSSoverview in healthcare systems 48 2.2.1 Importance and prospects of MDSSs 50 2.2.2 The challenges and pitfalls of MDSS 54 vi XAI in MDSS 2.3 AIin MDSS 57 2.3.1 The basis of AI inhealthcare systems 59 2.3.2 The role of AIin MDSS 60 2.3.3 Related workof AIin MDSS 63 2.3.4 AI weakness in healthcare system 65 2.4 XAI 66 2.4.1 The basis of XAI 67 2.4.2 The role of XAIin MDSSs 69 2.5 Ethical effects and implications 72 2.5.1 XAIweaknesses in medicine 75 2.6 Conclusion and future directions 75 Acknowledgment 76 References 76 3 Explainable Artificial Intelligence-based framework for medical decision supportsystems 91 Joseph Bamidele Awotunde, Oluwafisayo Babatope Ayoade, Panigrahi Ranjit, Amik Gargand Akash KumarBhoi 3.1 Introduction 92 3.1.1 Key contributions of the chapter 94 3.1.2 Chapter organization 94 3.2 Applicability of XAIin MDSSs 94 3.3 The challenges in the applicability of XAIin MDSSs 97 3.4 The proposed DeepSHAPenabled withDNNframework 100 3.4.1 The pre-processing stage 100 3.4.2 The hyper-parameters and DNN 101 3.4.3 The Shapley additive explainable (SHAP) 101 3.5 Experimental design forcancer prediction 102 3.5.1 The Wisconsin breast cancer (WBCD) dataset 102 3.5.2 The performance evaluation metrics 103 3.6 Experimental results 104 3.6.1 Thecomparisonoftheproposedmodelwithexistingmethods 105 3.6.2 The local explanation results 106 3.7 The future research direction of XAIin healthcare systems 106 3.8 Conclusion and future scopes 107 References 108 4 Prototype interface for detecting mental fatigue withEEG andXAI frameworks inIndustry4.0 117 Mart´ınMontes Rivera, Luciano Martinez, Alberto OchoaZezzatti, Alan Navarro,Jesu´sRodarte and Ne´storLo´pez 4.1 Introduction 118 4.1.1 Measurement of mental fatigue 119 4.1.2 EEG in mental fatigue 119 4.1.3 Acquisition with brain–machine interface (BCI) 121 4.1.4 EEGNET 122 Contents vii 4.2 Related work 123 4.3 Materials and methods 124 4.3.1 Selectionofcomputerequipmentformentalfatiguedetection 125 4.3.2 Generation of the dataset for training 126 4.3.3 Training of EEGNet 127 4.3.4 Graphical interface and control of communication with the trained model 128 4.4 Results and discussions 128 4.4.1 Results 128 4.4.2 Discussionsof results 129 4.5 Conclusions 134 References 134 5 XAIfor medical image segmentationinmedical decision support 137 systems AbasiamaGodwinAkpan,FlaviousBobuinNkubli,VictoriaNnaemekaEzeano, AnayoChristianOkwor,MabelChikodiliUgwujaandUdemeOffiong 5.1 Introduction 138 5.1.1 Contributions of the current study 139 5.1.2 Chapter organization 139 5.2 Related work 139 5.2.1 Concept of XAI 139 5.2.2 Framework forXAI 140 5.2.3 Explainabilityin healthcare 140 5.2.4 DLconcept and applications 142 5.2.5 Computer vision tasks 142 5.2.6 Convolutional neural networks(CNNs) 144 5.2.7 Medical image segmentation 145 5.2.8 Medical image segmentation techniques 146 5.2.9 Medical imaging modality 146 5.2.10 Summary of related works 149 5.3 Analysis of the proposed system 152 5.3.1 Analysis of algorithm forproposed system 154 5.3.2 Advantages of the hybrid system 158 5.3.3 Disadvantages of the system 158 5.3.4 Justification of the system 160 5.4 Conclusion 160 References 161 6 XAIrobot-assistedsurgeries infuture medical decisionsupport 167 systems Aishat Titilola Rufai, Kenechi Franklin Dukor, Opeyemi Michael Ageh and Agbotiname Lucky Imoize 6.1 Introduction 168 6.2 Related work 169 6.2.1 Current applications of AIin the healthcare systems 169 6.2.2 Limitations of AI in the medical field 171 viii XAI inMDSS 6.2.3 XAI 172 6.2.4 XAIinhealthcare 172 6.2.5 Howexplainability works—bridging the AI gap 174 6.2.6 Benefits of XAIfor the medical field 175 6.3 Medical robots 176 6.3.1 History of robotic surgery 178 6.3.2 Current and future useof medical robotsand devices 179 6.3.3 Robotic surgery and AI 179 6.3.4 Current application of AIin robotic surgery 180 6.3.5 Current application of AIin emerging robotic systems 181 6.3.6 XAIrobot-assisted surgeries for MDSS 183 6.3.7 Current limitations of XAIand robotic surgeryfor MDSS 183 6.4 Explanation methods 184 6.4.1 Explanationmethods in robotics 185 6.4.2 SHAPs 186 6.4.3 Layer-wise relevance propagation 186 6.4.4 LIMEs 187 6.5 Conclusion 187 Acknowledgment 188 References 188 7 Prediction of erythemato squamous-disease usingensemble learning framework 197 Efosa Charles Igodan, Olumide Olayinka Obe,Aderonke Favour-Bethy Thompsonand Otasowie Owolafe 7.1 Introduction 197 7.2 Related literature review 200 7.3 Materials and methods 200 7.3.1 Data collection 203 7.3.2 Dataset analysis 203 7.3.3 Feature selection 203 7.3.4 Multi-filter-basedfeature selection approach 204 7.3.5 Multi-embedded-based feature selection approach 205 7.3.6 Anensemble multi-feature selection(EMFME-FS)approach 208 7.3.7 Machine learning classifiers 209 7.3.8 Ensemble methods 210 7.4 Experimental results and discussion 213 7.5 Conclusion 222 References 223 8 Security-based explainable artificial intelligence (XAI)inhealthcare system 229 Hu¨seyin Gu¨ru¨ler, Naveed Islamand Alloud Din 8.1 Introduction 230 8.1.1 XAI 232 8.1.2 Model-based explanation 232 Contents ix 8.1.3 Post-hoc XAI 232 8.1.4 Model-specific explanation 233 8.1.5 Model-agnostic explanation 233 8.1.6 Global explanation 233 8.1.7 Local explanation 233 8.2 Literature review 234 8.2.1 XAIand AI 234 8.2.2 Explanation meaningfulnessand veracity 235 8.2.3 ML in healthcare 236 8.2.4 Intelligibilityand explainable systems research in HCI 237 8.3 Methodology 238 8.3.1 Explainable video action recognition system 238 8.3.2 TL 238 8.3.3 Model architecture 239 8.3.4 Freeze model 240 8.3.5 Fine-tune model 241 8.3.6 Pre-trained model followed by anew classifier 241 8.3.7 Pre-trained CNNsimplementation 243 8.4 Experimental result 243 8.4.1 Human action dataset 243 8.4.2 ResNet50 visual explanations 245 8.4.3 VGG16visual explanations 245 8.4.4 VGG19visual explanations 246 8.4.5 Final discussion 250 8.5 Conclusion and future scope 252 Acknowledgment 252 References 253 9 Explainable dimensionality reduction model withdeep learning for diagnosinghypertensive retinopathy 259 Micheal OlaoluArowolo,HadassahOluwadamilola Olumuyiwa,Ruth Omorinsola Adesina, Royal Afonime, Mobayonle Ayodeji Ajayi and Paul Adeoye Omosebi 9.1 Introduction 260 9.2 Overview and related works 262 9.2.1 Hypertension 262 9.2.2 Machine learning 262 9.2.3 Related works 264 9.3 Materials and methods 271 9.3.1 Data description 271 9.3.2 Data preprocessing 271 9.4 Results and discussions 273 9.4.1 Importing the dataset 273 9.4.2 Resizing and convertingthe images to array 275 9.4.3 Data splitting 275