Explainable Artificial Intelligence in Medical Decision Support Systems

Cover
Contents
About the editors
Preface
Acknowledgments
1 Explainable artificial intelligence (XAI) in medical decision systems (MDSSs): healthcare systems perspective
Abstract
1.1 Introduction
1.2 Overview of HMDSSs
1.2.1 MDSSs in healthcare system
1.2.2 Basis of HMDSS
1.2.3 Characterizing and categorizing HMDSS
1.3 Case study of XAI enabled with MDSSs in various infectious diseases
1.3.1 SCD
1.3.2 Diabetes mellitus (DM)
1.3.3 Hypertensive retinopathy (HR)
1.3.4 Carcinoma
1.3.5 COVID-19 pandemic
1.4 XAI research trends and open issues
1.4.1 XAI perspective in healthcare
1.5 Conclusion and future directions
Acknowledgment
References
2 Explainable artificial intelligence (XAI) in medical decision support systems (MDSS): applicability, prospects, legal implications, and challenges
Abstract
2.1 Introduction
2.1.1 Chapter organization
2.2 MDSS overview in healthcare systems
2.2.1 Importance and prospects of MDSSs
2.2.2 The challenges and pitfalls of MDSS
2.3 AI in MDSS
2.3.1 The basis of AI in healthcare systems
2.3.2 The role of AI in MDSS
2.3.3 Related work of AI in MDSS
2.3.4 AI weakness in healthcare system
2.4 XAI
2.4.1 The basis of XAI
2.4.2 The role of XAI in MDSSs
2.5 Ethical effects and implications
2.5.1 XAI weaknesses in medicine
2.6 Conclusion and future directions
Acknowledgment
References
3 Explainable Artificial Intelligence-based framework for medical decision support systems
Abstract
3.1 Introduction
3.1.1 Key contributions of the chapter
3.1.2 Chapter organization
3.2 Applicability of XAI in MDSSs
3.3 The challenges in the applicability of XAI in MDSSs
3.4 The proposed DeepSHAP enabled with DNN framework
3.4.1 The pre-processing stage
3.4.2 The hyper-parameters and DNN
3.4.3 The Shapley additive explainable (SHAP)
3.5 Experimental design for cancer prediction
3.5.1 The Wisconsin breast cancer (WBCD) dataset
3.5.2 The performance evaluation metrics
3.6 Experimental results
3.6.1 The comparison of the proposed model with existing methods
3.6.2 The local explanation results
3.7 The future research direction of XAI in healthcare systems
3.8 Conclusion and future scopes
References
4 Prototype interface for detecting mental fatigue with EEG and XAI frameworks in Industry 4.0
Abstract
4.1 Introduction
4.1.1 Measurement of mental fatigue
4.1.2 EEG in mental fatigue
4.1.3 Acquisition with brain–machine interface (BCI)
4.1.4 EEGNET
4.2 Related work
4.3 Materials and methods
4.3.1 Selection of computer equipment for mental fatigue detection
4.3.2 Generation of the dataset for training
4.3.3 Training of EEGNet
4.3.4 Graphical interface and control of communication with the trained model
4.4 Results and discussions
4.4.1 Results
4.4.2 Discussions of results
4.5 Conclusions
References
5 XAI for medical image segmentation in medical decision support systems
Abstract
5.1 Introduction
5.1.1 Contributions of the current study
5.1.2 Chapter organization
5.2 Related work
5.2.1 Concept of XAI
5.2.2 Framework for XAI
5.2.3 Explainability in healthcare
5.2.4 DL concept and applications
5.2.5 Computer vision tasks
5.2.6 Convolutional neural networks (CNNs)
5.2.7 Medical image segmentation
5.2.8 Medical image segmentation techniques
5.2.9 Medical imaging modality
5.2.10 Summary of related works
5.3 Analysis of the proposed system
5.3.1 Analysis of algorithm for proposed system
5.3.2 Advantages of the hybrid system
5.3.3 Disadvantages of the system
5.3.4 Justification of the system
5.4 Conclusion
References
6 XAI robot-assisted surgeries in future medical decision support systems
Abstract
6.1 Introduction
6.2 Related work
6.2.1 Current applications of AI in the healthcare systems
6.2.2 Limitations of AI in the medical field
6.2.3 XAI
6.2.4 XAI in healthcare
6.2.5 How explainability works—bridging the AI gap
6.2.6 Benefits of XAI for the medical field
6.3 Medical robots
6.3.1 History of robotic surgery
6.3.2 Current and future use of medical robots and devices
6.3.3 Robotic surgery and AI
6.3.4 Current application of AI in robotic surgery
6.3.5 Current application of AI in emerging robotic systems
6.3.6 XAI robot-assisted surgeries for MDSS
6.3.7 Current limitations of XAI and robotic surgery for MDSS
6.4 Explanation methods
6.4.1 Explanation methods in robotics
6.4.2 SHAPs
6.4.3 Layer-wise relevance propagation
6.4.4 LIMEs
6.5 Conclusion
Acknowledgment
References
7 Prediction of erythemato squamous-disease using ensemble learning framework
Abstract
7.1 Introduction
7.2 Related literature review
7.3 Materials and methods
7.3.1 Data collection
7.3.2 Dataset analysis
7.3.3 Feature selection
7.3.4 Multi-filter-based feature selection approach
7.3.5 Multi-embedded-based feature selection approach
7.3.6 An ensemble multi-feature selection (EMFME-FS) approach
7.3.7 Machine learning classifiers
7.3.8 Ensemble methods
7.4 Experimental results and discussion
7.5 Conclusion
References
8 Security-based explainable artificial intelligence (XAI) in healthcare system
Abstract
8.1 Introduction
8.1.1 XAI
8.1.2 Model-based explanation
8.1.3 Post-hoc XAI
8.1.4 Model-specific explanation
8.1.5 Model-agnostic explanation
8.1.6 Global explanation
8.1.7 Local explanation
8.2 Literature review
8.2.1 XAI and AI
8.2.2 Explanation meaningfulness and veracity
8.2.3 ML in healthcare
8.2.4 Intelligibility and explainable systems research in HCI
8.3 Methodology
8.3.1 Explainable video action recognition system
8.3.2 TL
8.3.3 Model architecture
8.3.4 Freeze model
8.3.5 Fine-tune model
8.3.6 Pre-trained model followed by a new classifier
8.3.7 Pre-trained CNNs implementation
8.4 Experimental result
8.4.1 Human action dataset
8.4.2 ResNet50 visual explanations
8.4.3 VGG16 visual explanations
8.4.4 VGG19 visual explanations
8.4.5 Final discussion
8.5 Conclusion and future scope
Acknowledgment
References
9 Explainable dimensionality reduction model with deep learning for diagnosing hypertensive retinopathy
Abstract
9.1 Introduction
9.2 Overview and related works
9.2.1 Hypertension
9.2.2 Machine learning
9.2.3 Related works
9.3 Materials and methods
9.3.1 Data description
9.3.2 Data preprocessing:
9.4 Results and discussions
9.4.1 Importing the dataset
9.4.2 Resizing and converting the images to array
9.4.3 Data splitting
9.4.4 Pre-processing the data with LDA
9.4.5 Training the ANN model with and without LDA
9.4.6 Plotting the scattered plot and confusion matrix for the ANN model with and without LDA
9.4.7 Comparison with previous works
9.5 Conclusions
References
10 Understanding cancer patients with diagnostically influential factors using high-dimensional data embedding
Abstract
10.1 Introduction
10.2 Literature review
10.3 Dimensionality reduction methods
10.3.1 Projection
10.3.2 Manifold learning
10.3.3 PCA
10.3.4 t-SNE
10.3.5 SDD
10.4 Methodology
10.4.1 Procedure
10.4.2 Data used
10.4.3 Performance assessment metrics
10.5 Experiments
10.6 Discussion of results
10.7 Concluding remarks and future work
References
Appendix
11 Explainable neural networks in diabetes mellitus prediction
Abstract
11.1 Introduction
11.2 Related work
11.3 Methodology
11.3.1 Key implementation requirements and strategies for xDNNs
11.3.2 DNN architecture
11.3.3 Activation function
11.3.4 Procedures for xDNN model implementation
11.3.5 Model parameters and hyper-parameters
11.3.6 Evaluation and explainability metrics
11.4 Results and discussion
11.4.1 Results for DNN models
11.4.2 Results for neural network models
11.5 Conclusion and future scope
Acknowledgment
References
12 A KNN and ANN model for predicting heart diseases
Abstract
12.1 Introduction
12.2 Overview of the literature
12.2.1 Heart diseases
12.2.2 Machine learning
12.2.3 Related work
12.3 Materials and methods
12.3.1 Standard scalar
12.3.2 ANNs
12.3.3 K-nearest neighbor
12.3.4 Performance metrics
12.4 Results and discussions
12.4.1 Comparison with previous work
12.5 Conclusions
References
13 Artificial Intelligence-enabled Internet of Medical Things for COVID-19 pandemic data management
Abstract
13.1 Introduction
13.2 Related work
13.3 IoMT for COVID-19 pandemic data management
13.3.1 Architecture of IoMT
13.3.2 Applications of the IoMT in COVID-19 data management
13.4 Reducing the workload of the medical industry
13.4.1 Applications of AI-enabled IoMT
13.4.2 Applications of AI-enabled IoMT for drug repurposing
13.5 Privacy-aware energy-efficient framework using AIoMT for COVID-19
13.6 Open research issues
13.6.1 Security and privacy
13.6.2 Energy efficiency
13.6.3 Integration of emotion-aware abilities
13.6.4 Interoperability
13.6.5 AI in IoMT
13.6.6 Ethical issues
13.7 Conclusion
Acknowledgment
References
14 A deep neural network for the identification of lead molecules in antibiotics discovery
Abstract
14.1 Introduction
14.1.1 DNN and its architecture
14.1.2 Lead identification techniques
14.2 Literature review
14.3 Materials and methods
14.3.1 Dataset preparation and preprocessing
14.3.2 Model development
14.3.3 Model evaluation
14.4 Results and discussion
14.5 Conclusion
References
15 Statistical test with differential privacy for medical decision support systems
Abstract
15.1 Introduction
15.2 Related work
15.2.1 Chi-squared hypothesis test
15.2.2 Privacy model
15.2.3 ε-Differentially private Chi-squared test
15.2.4 Adversarial model
15.2.5 Other privacy models
15.3 Proposed algorithm
15.3.1 Outline
15.3.2 Global sensitivity of Chi-squared value
15.3.3 Differentially private hypothesis testing
15.3.4 Complexity analysis
15.4 Evaluation
15.4.1 Significance results
15.4.2 Power results
15.4.3 Results of real datasets
15.5 Discussion
15.6 Conclusion
Acknowledgment
References
16 Automated decision support system for diagnosing sleep diseases using machine intelligence techniques
Abstract
16.1 Introduction
16.2 Related work
16.3 Experimental dataset
16.4 Proposed automatic sleep stage detection method
16.5 Classification
16.5.1 SVM
16.5.2 Random Forest (RF)
16.5.3 Gradient boosting decision tree (GBDT)
16.5.4 eXtreme gradient boosting (XGBoost)
16.5.5 Stacking ensembling learning
16.6 Experimental discussion
16.6.1 Feature screening results
16.6.2 Sleep staging performance with ISRUC-Sleep subgroup-I dataset
16.6.3 Automated decision on sleep staging using the ensemble learning stacking algorithm
16.7 Conclusion
References
17 XAI methods for precision medicine in medical decision support systems
Abstract
17.1 Introduction
17.1.1 Contributions of the current study
17.1.2 Chapter organization
17.2 Related works
17.2.1 Measurement of XAI in precision medicine
17.2.2 Concept of explainability and interpretability
17.3 Explainable models in MDSS: opportunities and challenges
17.4 Conclusion
References
18 The psychology of explanation in medical decision support systems
Abstract
18.1 Introduction
18.1.1 Categories of AI
18.1.2 Artificial narrow intelligence
18.1.3 Artificial broad intelligence
18.1.4 Artificial general intelligence
18.1.5 Artificial super-intelligence
18.2 Recent development of XAI in MDSS
18.2.1 AI in clinical practice
18.2.2 AI in biomedical research
18.2.3 AI for public and global health
18.2.4 AI in healthcare administration
18.3 Potential benefits of XAI in MDSS
18.3.1 Radiology
18.3.2 Early diagnosis
18.3.3 Emergency medicine
18.3.4 Risk prediction
18.3.5 Chatbots
18.3.6 Virtual nursing assistance
18.3.7 Precision medicine
18.3.8 Administrative workflow assistance
18.4 Key challenges of XAI in MDSS
18.4.1 Patient harm due to AI errors
18.4.2 Misuse of medical AI tools
18.4.3 Risk of bias in medical AI
18.4.4 Lack of transparency
18.4.5 Privacy and security issues
18.5 The future of XAI in MDSS
18.6 The research trend of XAI in MDSS
18.7 The future directions and recommendations
18.8 Conclusions and future scope
Acknowledgment
References
Index
Back Cover