Medical Data Analysis and Processing using Explainable Artificial Intelligence

著者名：Jena, Om Prakash (EDT)/Panda, Mrutyunjaya (EDT)/Kose, Utku (EDT)

CRC Press（2023/11/06発売）

• 言語：ENG
• ISBN：9781032191133
• eISBN：9781000983654

Description

The text presents concepts of explainable artificial intelligence (XAI) in solving real world biomedical and healthcare problems. It will serve as an ideal reference text for graduate students and academic researchers in diverse fields of engineering including electrical, electronics and communication, computer, and biomedical

• Presents explainable artificial intelligence (XAI) based machine analytics and deep learning in medical science
• Discusses explainable artificial intelligence (XA)I with the Internet of Medical Things (IoMT) for healthcare applications
• Covers algorithms, tools, and frameworks for explainable artificial intelligence on medical data
• Explores the concepts of natural language processing and explainable artificial intelligence (XAI) on medical data processing
• Discusses machine learning and deep learning scalability models in healthcare systems

This text focuses on data driven analysis and processing of advanced methods and techniques with the help of explainable artificial intelligence (XAI) algorithms. It covers machine learning, Internet of Things (IoT), and deep learning algorithms based on XAI techniques for medical data analysis and processing. The text will present different dimensions of XAI based computational intelligence applications. It will serve as an ideal reference text for graduate students and academic researchers in the fields of electrical engineering, electronics and communication engineering, computer engineering, and biomedical engineering.

Table of Contents

Chapter 1 Explainable AI (XAI): Concepts and Theory

Tanvir Habib Sardar, Sunanda Das, Bishwajeet Kumar Pandey

Abstract

1. Introduction
2. Formal Definitions of Explainable Artificial Intelligence
3. The Working Mechanism of Explainable Artificial Intelligence: How Explainable Artificial Intelligence Generates Explanations
4. How Humans Reason (with Errors)
5. How Explainable Artificial Intelligence Support Reason and Solve Human Error Issue
6. Applications and Impact Areas of Explainable Artificial Intelligence

1. Threat Detection
2. Object Detection
3. Adversarial ML Prevention
4. Open Source Intelligence (OSI)
5. Automated Medical Diagnosis
6. Autonomous Vehicles

7. Benefits of Explainable Artificial Intelligence
8. Research Challenges of Explainable Artificial Intelligence
9. Use Cases of Explainable Artificial Intelligence
10. Limitations of Explainable Artificial Intelligence
11. Conclusion

References

Chapter 2: Utilizing Explainable Artificial Intelligence to Address Deep Learning in Biomedical Domain

Priyanka Sharma

Abstract

2.1 Introduction: Background and Driving Forces

2.2 XAI Taxonomy

2.3 Review of State of Art

2.3.1 Methods focused on features

2.3.2 Global methods

2.3.3 Concept Methods

2.3.4 Surrogate Methods

2.3.5 Local, Pixel-based Techniques

2.3.6 Human Centered Methods

2.4 Deep Learning –Reshaping Healthcare

2.4.1 Deep Learning Methods

2.4.1.1 Multi-layer Perceptron or Deep Feed Forward Neural Network

2.4.1.2 Restricted Boltzmann Machine

2.4.1.3 Deep Belief Network

2.4.1.4 Autoencoder

2.4.1.5 Convolutional Neural Network

2.4.1.6 Recurrent Neural Network

2.4.1.7 Long Short- Term Memory (LSTM) and Gated Recurrent Unit (GRU)

2.4.2 Deep Learning Applications in Healthcare

2.5 Results

2.6 Benefits and Drawbacks of XAI Methods

2.7 Conclusion

Chapter 3 Explainable Fuzzy Decision Tree for Medical Data Classification

Authors: Swathi Jamjala Narayanan, Boominathan Perumal, Sangeetha Saman

Abstract

3.1. Introduction

3.2. Literature survey

3.3. Fuzzy classification problem

3.4. Induction of fuzzy decision tree

3.4.1 Fuzzy c-means clustering (FCM)

3.4.2 Cluster validity indices and Optimality Condition

3.4.2.1 Separation and Compactness (SC)

3.4.2.2 Compact Overlap (CO)

3.4.2.3 Fukuyama and Sugeno (FS)

3.4.2.4 Xie and Beni (XB)

3.4.2.5 Partition entropy

3.4.2.6 Fuzzy hyper volume (FHV)

3.4.2.7 PBMF

3.4.2.8 Partition coefficient

3.4.3 Basics of developing Fuzzy ID3

3.5. Case Study: Explainable FDT for HCV Medical Data

3.6. Conclusion and Future work

Chapter 4 Statistical Algorithm for Change Point Detection in Multivariate Time Series of Medicine Data Based on Principles of Explainable Artificial Intelligence

D. Klyushin, A. Urazovskyi

Abstract

4.1 Introduction

4.2 Detection of change points in multivariate time series

4.3 Petuninʼs ellipses and ellipsoids

4.4 Numerical experiments

4.4.1 Almost non-overlapped uniform distributions with different locations

4.4.2 Uniform distributions with different locations that initially
are strongly overlapped, then slightly overlapped,
and finally are not overlapped

4.4.3 Almost non-overlapped normal distributions with different locations

4.4.4 Normal distributions with the same location and scales that are gradually
begin to differ

4.4.5　Normal distributions with the same locations and strongly different scales

4.4.6. Exponential distributions with different parameters

4.4.7 Gamma-distributions with the same location and different scales

4.4.8　Gamma-distributions with different locations and the same scale

4.4.9　Gumbel distributions with different locations and the same scale

4.4.10　Gumbel distributions with the same location and different scales

4.4.11 Rayleigh　distributions with different scales

4.4.12　Laplace　distributions with different means and the same variance

4.4.13 Laplace　distributions with the same location and different scales

4.4.14 Logistic distributions with different locations and the same scale

4.4.15 Logistic distributions with the same location and different scales

4.4.16 Conclusion on numerical experiments

4.5 Quasi-real experiments

4.5.1 Simulation of tachycardia

4.5.2 Simulation of coronavirus pneumonia

4.5.3 Simulation of cancer lung

4.5.4 Simulation of physical activity

4.5.5 Simulation of stress/panic attack

4.5.6 Conclusion on quasi-real experiments

4.6 Conclusion

4.7 References

Chapter 5 XAI and Machine learning for Cyber security: A Systematic Review

Gousia Habib*, Shaima Qureshi

5.1. Introduction to Explainable AI (XAI).

5.2 Principles followed by XAI Algorithm.

5.3 Types of Explainability.

5.4 Some Critical Applications of Explainability

　

5.5 Related Work.

5.6 Historical Origins of the Need for Explainable AI.

5.7 Taxonomy of MAP of explainability approaches.

5.8 Challenges posed by XAI.

5.8.1 A Black Box Attack on XAI in Cybersecurity.

5.8.2 Manipulation of Adversarial Models to Deceive Neural Network Interpretations.

5.8.3 Geometry is responsible for the manipulation of explanations.

5.8.4 Saliency Method's Unreliability.

5.8.5 Misleading Black Box Explanations are used to manipulate user trust.

　

5.9 Various Suggested Solutions for XAI security Challenges

5.9.1 Addressing Manipulation of User Trust through Misleading Black Box Explanations.

5.9.2 Improved Interpretability of Deep Learning.

5.9.3 Heat-map explanations Defense against adversarial cyber-attacks.

5.9.4 Curvature minimization.

5.9.5 Weight decay.

5.9.6 Smoothing activation functions.

5.10 Conclusion

References

Chapter 6 Classification and regression tree (CART) modelling approach to predict the number of lymph node dissection among endometrial cancer patients

Prafulla Kumar Swain, Manas Ranjan Tripathy, Pravat Kumar Sarangi, Smruti Sudha Pattnaik

Abstract

6.1 Introduction

6.2 Data source

6.3 Methods used

6.3.1 Regression Tree

6.3.2 Optimal threshold value (cut off point)

6.3.3 Regression tree algorithm

6.3.4 Optimal threshold value (cut off point)

6.3.5 Validation of models

6.4 Applications to EC Data

6.5 Discussion

6.6 Conclusion

References

Chapter 7: Automated Brain Tumor Analysis using Deep Learning based Framework

Amiya Halder, Rudrajit Choudhuri, Apurba Sarkar

7.1 Introduction

7.2 Related Works

7.3 Background

7.3.1 Autoencoder

7.3.2 Convolutional Autoencoder

7.3.3 Pre-trained Deep Classification Architectures

7.4 Proposed Methodology

7.4.1 Image Denoising

7.4.2 Tumor Detection and Tumor Grade Classification

7.4.2.1 Data Acquisition

7.4.2.2 Experimental Setup: Fine Tuning the Architectures

7.4.3 Model Training

7.5 Result Analysis

7.5.1 Evaluation Metrics

7.5.2 Performance Evaluation

7.6 Conclusion

Chapter 8 A Robust Framework for Prediction of Diabetes Mellitus using Machine Learning

Sarthak Singh, Rohan Singh, Arkaprovo Ghosal, Tanmaya Mahapatra

Abstract

8.1 introduction

8.2 Background

8.3 Related Work

8.4 Conceptual Approach

8.5 Evaluation

8.6 Discussion

8.7 Conclusion

References

Chapter 9 Effective Feature Extraction for Early Recognition and Classification of Triple Modality Breast Cancer Images Using Logistic Regression Algorithm

Manjula Devarakonda Venkata , Sumalatha Lingamgunta

Abstract

9.1 Introduction

9.2 Symptoms of Breast Cancer

9.3 Need for Early detection

9.4 Datasets used

9.5 Classification of Medical features from three modalities using Logistic

Regression Algorithm

9.5.1 Pre processing

9.6 Results

9.6.1 Pre-processed US images

9.6.2 Preprocessed Mammogram Image

9.6.3 Pre processed MRI images

9.7 Conclusion

References

Chapter 10: Machine Learning and Deep Learning Models Used to Detect Diabetic Retinopathy and Its Stages

S. Karthika, M. Durgadevi

　Abstract

1. Introduction

1. Conventional ML & DL Algorithms

1. ML - Support Vector Machine (ML- SVM)
2. ML - K_Nearest Neighbors (ML- KNN)
3. ML - Random Forest (ML- RF)
4. ML - Neural Networks (ML- NN)
5. Deep Learning (DL)
6. DL - Classic Neural Networks

1. DL - Convolutional Neural Networks (DL- CNN)
2. DL - LSTMNs (Long Short-Term Memory Networks)
3. DL - Recurrent Neural Networks (DL- RNN)
4. DL - Generative Adversarial Networks (DL - GAN)
5. DL -Reinforcement Learning

1. Retinal Image Datasets used in DR detection

1. DR Process Detection

1. Non-Proliferative Diabetic Retinopathy
2. Proliferative diabetic retinopathy

1. Techniques for Detecting Microaneurysms (MA)
2. Techniques for Detecting Hemorrhage (HEM)
3. Techniques for Detecting Exudate (EX)
4. Techniques for Detecting Macular Edema

1. Diabetic Retinopathy Lesion Segmentation
2. Performance Metrics

1. True Positive
2. True Negative
3. False Positive
4. False Negative

1. Conclusion

　

Chapter 11 Clinical Natural Language Processing Systems for Information Retrieval from Unstructured Medical Narratives

S. Lourdumarie Sophie, S. Siva Sathya

11.1 Introduction

11.2 Components of NLP

11.2.1 Natural Language Understanding (NLU)

11.2.2 Natural Language Generation (NLG)

11.3 Stages of NLP

11.3.1 Phonological Analysis

11.3.2 Morphological and Lexical Analysis

11.3.3 Syntactic Analysis

11.3.4 Semantic Analysis

11.3.5 Discourse Integration

11.3.6 Pragmatic Analysis

11.4 Applications & Techniques

11.4.1 Optical Character Recognition (OCR)

11.4.2 Named Entity Recognition (NER)

11.4.3 Question Answering

11.4.4 Chatbots

11.4.5 Machine Translation

11.4.6 Sentiment Analysis

11.4.7 Topic Modelling

11.4.8 Automatic Text Summarization (ATS)

11.4.9 Co-reference Resolution

11.4.10 Disease Prediction

11.4.11 Text Classification

11.4.12 Cognitive Assistant (CA)

11.4.13 Automatic Speech Recognition (ASR)

11.5 NLP Systems in Health Care

11.6 Conclusion

References