Biomedical Signal and Image Processing with Artificial Intelligence

Preface
Contents
About the Authors
Voice Privacy in Biometrics
	1 Introduction
		1.1 Motivation for Voice Privacy
		1.2 De-identification vs. Anonymization
	2 Voice Privacy and Attacker's Perspective
		2.1 Target Selection by Attacker and Voice Privacy System
		2.2 Enrolled Users with Malicious Intent
	3 Voice Privacy Using Linear Prediction Model
		3.1 Speech Production Model
		3.2 Energy Losses
		3.3 Linear Prediction (LP) Model
		3.4 Experimental Setup
			3.4.1 Baseline System
			3.4.2 Corpora Used
			3.4.3 Proposed Voice Privacy System
			3.4.4 Experimental Results
			3.4.5 Gender-Based Analysis
	4 Technological Challenges in Voice Privacy
		4.1 Evaluation Metrics for Speech Quality
		4.2 Machines vs. Human Perception of Speech
		4.3 Robustness vs. Vulnerability
	5 Voice Privacy and Cryptography
		5.1 Public-Key Encryption
		5.2 Limitations of Cryptographic Approaches for Voice Privacy
	6 Summary and Conclusions
	References
Histopathology Whole Slide Image Analysis for Breast Cancer Detection
	1 Introduction
		1.1 Lymph Node Biopsy and the Need for Computer Aided Diagnosis
		1.2 Whole Slide Imaging, and Weakly Supervised Image Classification
	2 Literature Survey
		2.1 Weakly Supervised Learning
			2.1.1 Tellez et al. 3
			2.1.2 Courtiol et al. 6
			2.1.3 Compression Analysis of the Above Weakly Supervised Methods
		2.2 Fully Supervised Learning
		2.3 Contribution of This Work
	3 Dataset
	4 Methodology
		4.1 Histogram-Based Selection of Patches from WSI
		4.2 Extraction of Feature Embeddings Using a Pretrained Network
			4.2.1 Compression of the WSI Images Using the Embeddings
		4.3 Classification Using the CNN
			4.3.1 Network Architecture
			4.3.2 Cosine Loss
			4.3.3 Learning Rate Schedule
			4.3.4 Classification Decision for the Whole Slide Image
			4.3.5 Study: Analysis of Random Crops
	5 Results
		5.1 Experimental Details
		5.2 Accuracy with Varying Threshold T
		5.3 ROC-AUC Results
		5.4 Comparison
	6 Conclusion
	References
Lung Classification for COVID-19
	1 Introduction
	2 Materials and Methods
		2.1 Data Collection
			2.1.1 Description of the COVID-19 Dataset
		2.2 Datasets Summary
	3 Methods
		3.1 Normal–Abnormal Lung Classification
		3.2 COVID–Non-COVID Classification
	4 Results and Discussion
		4.1 Normal–Abnormal Lung Classification
		4.2 COVID-Non-COVID Lung Classification
		4.3 Testing the Cloud-Based COVID Lung Classification System
	5 Conclusion
	References
GRU-Based Parameter-Efficient Epileptic Seizure Detection
	1 Introduction
	2 Related Work
	3 Proposed Method
		3.1 Recurrent Neural Networks
		3.2 Gated Recurrent Units
			3.2.1 Reset Gate
			3.2.2 Update Gate
		3.3 Model Architecture
		3.4 Loss Function
	4 Experiments and Results
		4.1 Dataset and the Training Details
		4.2 Hardware Implementation
		4.3 Evaluation Metrics: Performance of the Proposed Architecture
	5 Conclusion
	References
An Object Aware Hybrid U-Net for Breast Tumour Annotation
	1 Introduction
	2 Related Work
	3 Methodology
		3.1 Overview
		3.2 Modified U-Net Architecture
		3.3 Active Contour
		3.4 Automatic Multiple Initialization
		3.5 U-Net Training with Structured SVM Loss
			3.5.1 Structured SVM Loss
	4 Experimental Setup and Results
		4.1 Dataset Preparation and Usage
		4.2 Experiments
	5 Discussion
	6 Conclusion
	References
VLSI Implementation of sEMG Based Classification for Muscle Activity Control
	1 Introduction
	2 EMG Data Acquisition
		2.1 Needle Electrodes
		2.2 Fine Wire Electrodes
		2.3 Surface EMG Electrode
	3 EMG Based Pattern Recognition for Upper Limb Prosthesis
	4 Machine Learning Models for Prosthetic Control Using EMG Signal
		4.1 Time Domain Based Feature Extraction
			4.1.1 Mean Absolute Value (MAV)
			4.1.2 Variance (VAR)
			4.1.3 Standard Deviation (SD)
			4.1.4 Mean absolute deviation (MAD)
	5 Pattern Recognition Through Classification with EMG
		5.1 Linear Discriminant Analysis
		5.2 Quadratic Discriminant Analysis
		5.3 Mathematical Approach for LDA and QDA
	6 VLSI Implementation of a Classification Algorithm for EMG
	7 Conclusion
	References
Content-Based Image Retrieval Techniques and Their Applications in Medical Science
	1 Introduction
	2 Classification of CBIR Techniques
		2.1 Feature-Based Technique
			2.1.1 Color Features for Image Retrieval
			2.1.2 Shape Features for Image Retrieval
			2.1.3 Texture-Feature-Based Image Retrieval
		2.2 Machine-Learning-Based Technique
			2.2.1 Support Vector Machine Agrawal2011, Alrahhal2019, Narasimha2018, Vani2019
			2.2.2 K-Means Cluster Chang2012, Jain2018, Serrano-Talamantes2013
			2.2.3 Machine-Learning-Based Image Retrieval System
		2.3 Deep-Learning-Based Technique
			2.3.1 Transfer-Learning-Based Technique
			2.3.2 Deep-Learning-Based Image Retrieval System
		2.4 Application of CBIR System
			2.4.1 Content-Based Medical Image Retrieval (CBMIR)
	3 Conclusion and Future Scope
	4 Copyright Statement
		4.1 Copyright
	References
Data Analytics on Medical Images with Deep Learning Approach
	1 Introduction
	2 Healthcare Data Sources and Analytics
	3 Background and Motivation
	4 Methodology
		4.1 Problem Formulation
		4.2 Contextual Selection of Medical Images
	5 Experimentation
		5.1 Dataset
		5.2 Model Selection and Training parameters
		5.3 Evaluation
	6 Conclusion
	References
Analysis and Classification Dysarthric Speech
	1 Introduction
	2 Types of Dysarthria
		2.1 Spastic Dysarthria
		2.2 Flaccid Dysarthria
		2.3 Ataxic Dysarthria
		2.4 Hypokinetic Dysarthria
		2.5 Hyperkinetic Dysarthria
		2.6 Mixed Dysarthria
	3 Analysis of Dysarthric Speech
		3.1 Time-Domain Analysis
			3.1.1 Fundamental Frequency (F0)
			3.1.2 Teager Energy Operator (TEO)
		3.2 Linear Prediction (LP) Residual
		3.3 Time-Frequency Analysis
	4 Datasets on Dysarthric Speech
		4.1 TORGO Database
		4.2 Universal Access (UA) Corpus
		4.3 HomeService Corpus
	5 Classification of Dysarthric and Normal Speech
		5.1 Experimental Setup
		5.2 Dataset Used for This Study
		5.3 Results and Analysis
	6 Conclusion
	References
Skin Cancer Detection and Classification Using DWT-GLCM with Probabilistic Neural Networks
	1 Introduction
	2 Literature Survey
	3 Proposed Method
		3.1 Database Training and Testing
		3.2 Preprocessing
		3.3 Image Segmentation
		3.4 Feature Extraction
		3.5 Classification
	4 Experimentation Environment
		4.1 Dataset
		4.2 Performance Metrics
	5 Conclusion
	References
Manufacturing of Medical Devices Using Artificial Intelligence-Based Troubleshooters
	1 Review of Literature
	2 Introduction
	3 Method
		3.1 AI Agent
		3.2 Human Cognition and the AI Agent
			3.2.1 Capture the Image and Identify the Problem
			3.2.2 Search Solution from Database and Provide It to the Operators
			3.2.3 Learn New Solutions
	4 Results
	5 Conclusion
		5.1 Future Prospects
	References
Enhanced Hierarchical Prediction for Lossless Medical Image Compression in the Field of Telemedicine Application
	1 Introduction
	2 Hierarchical Prediction
		2.1 Proposed Method for Hierarchical Prediction
		2.2 Result and Discussion of Hierarchical Prediction
	3 Proposed Method for Modified Hierarchical Prediction
		3.1 Result and Discussion of Modified Hierarchical Prediction
	4 Comparative Result Discussion on Hierarchical Prediction with Modified Hierarchical Prediction
	5 Conclusion
	Appendices
		Appendix A: List of Abbreviations
		Appendix B: Formula of Performance Parameters
	References
		Websites
LBP-Based CAD System Designs for Breast Tumor Characterization
	1 Introduction
	2 Methodology
		2.1 Dataset Description and Bifurcation
		2.2 Despeckling
		2.3 Segmentation
		2.4 Combined Feature Set Generation
			2.4.1 LBP-Based Texture Feature Extraction
			2.4.2 Correlation-Based Feature Selection
			2.4.3 Morphological Feature Extraction
			2.4.4 Serial Feature Fusion
		2.5 Optimal Feature Set Generation
			2.5.1 ANFC-LH Algorithm-Based Feature Selection
			2.5.2 GA-SVM Algorithm-Based Feature Selection
			2.5.3 PCA-SVM Algorithm-Based Dimensionality Reduction
			2.5.4 SAE-SM Algorithm-Based Dimensionality Reduction
		2.6 Classification
			2.6.1 Adaptive Neuro-fuzzy Classifier
			2.6.2 Support Vector Machine Classifier
			2.6.3 Softmax Classifier
	3 Results and Discussion
		3.1 Experiments and Classification Results
			3.1.1 Experiment 1: LBP-Based CAD System Using the ANFC-LH Algorithm
			3.1.2 Experiment 2: LBP-Based CAD System Using the GA-SVM Algorithm
			3.1.3 Experiment 3: LBP-Based CAD System Using the PCA-SVM Algorithm
			3.1.4 Experiment 4: LBP-Based CAD System Using the SAE-SM Algorithm
		3.2 Discussion
	4 Conclusion
	References
Detection of Fetal Abnormality Using ANN Techniques
	1 Introduction
	2 Artificial Neural Network
	3 Experimental Setup for Fetal Abnormality Detection
		3.1 Description of Data
	4 Results Using Back–Propagation Algorithm
	5 Results Using LMBP, SCG, and BR Algorithm
	6 Comparision of LMBP, SCG, and BR Algorithms
	7 Conclusion
	References
Machine Learning and Deep Learning-Based Framework for Detection and Classification of Diabetic Retinopathy
	1 Introduction
		1.1 Symptoms of DR
		1.2 Types and Levels of DR
		1.3 DR-Computer-Aided Diagnosis (CAD)'s Perspective
	2 Detection of DR Using Image Processing
	3 Related Work
	4 Comparative Analysis
	5 Experiments and Results
		5.1 Datasets
		5.2 Performance Evaluation
	6 Conclusion and Future Scope
	References
Applications of Artificial Intelligence in Medical Images Analysis
	1 Introduction
		1.1 Medical Images
		1.2 Artificial Intelligence
	2 Medical Imaging
		2.1 Imaging Modalities
			2.1.1 Conventional X-Ray
			2.1.2 Fluoroscopy
			2.1.3 Angiography
			2.1.4 Computed Tomography (CT)
			2.1.5 Magnetic Resonance Imaging (MRI)
			2.1.6 Mammography
			2.1.7 Ultrasound
			2.1.8 Nuclear Medicine
			2.1.9 Endoscopy
			2.1.10 Digital Pathology
		2.2 Image Formats
	3 Image Processing Methods
		3.1 Smoothing
		3.2 Sharpening
		3.3 Morphological Transformations
			3.3.1 Erosion
			3.3.2 Dilation
			3.3.3 Opening
			3.3.4 Closing
	4 Medical Image Analyses
		4.1 Radiography Image
		4.2 Ultrasound Image
		4.3 Endoscopy Image
		4.4 Digital Histopathology Image
	5 Conclusions
	References
Intelligent Image Segmentation Methods Using Deep Convolutional Neural Network
	1 Introduction
	2 Image Segmentation
		2.1 Image Database Domain Types
		2.2 Purpose of Image Segmentation
		2.3 Operations involved in Image Segmentation
		2.4 Performance Metrics for Segmentation Models
			2.4.1 Accuracy
			2.4.2 Pixel-Wise Accuracy (PA)
			2.4.3 Mean Pixel Accuracy (MPA)
			2.4.4 Precision
			2.4.5 Recall
			2.4.6 Specificity
			2.4.7 Intersection over Union (IoU) and Dice Coefficient (Dice)
	3 Types of Image Segmentation
		3.1 Segmentation Concepts
		3.2 Semantic Segmentation
			3.2.1 Fully Convolutional Networks
			3.2.2 SegNet
			3.2.3 UNet
			3.2.4 DeepLab
		3.3 Instance Segmentation
			3.3.1 DeepMask
			3.3.2 Mask R-CNN
	4 Research Challenges in Image Segmentation
		4.1 Complexity and Computation of DL Segmentation Models
		4.2 Variability in Object Appearance
		4.3 Requirement for Huge and Highly Defined Labeled Dataset
		4.4 Overfitting Issue
		4.5 Class Imbalance
		4.6 Issues with Real-Time Segmentation
	5 Conclusion
	References
Artificial Intelligence Assisted Cardiac Signal Analysis for Heart Disease Prediction
	1 Introduction
		1.1 Diagnosis of Heart Diseases
			1.1.1 Diagnosis Through ECG Cardiac Signal
			1.1.2 Diagnosis Through SCG and BCG Cardiac Signal
		1.2 Early Detection Through Smart IoT Devices
		1.3 Role of AI in Cardiac Abnormality Detection
	2 Cardiac Signal Data Acquisition
		2.1 Data Acquisition Through Wearable Gadgets
		2.2 Data Acquisition Through Body Sensors
		2.3 Data Acquisition in Laboratory
		2.4 Open Source Data
	3 Cardiac Signal Data Analysis
		3.1 ECG Signal Data Analysis
			3.1.1 Mathematical Model Based Prediction
			3.1.2 AI-Based Abnormality Prediction
		3.2 SCG Signal Data Analysis
			3.2.1 Mathematical Model Based Prediction
			3.2.2 AI-Based Abnormality Prediction
		3.3 Combined Analysis of Multichannel ECG and SCG Signals
			3.3.1 Mathematical Model Based Prediction
			3.3.2 AI-Based Abnormality Prediction
	4 Comparison of Related Literature
	5 Conclusions
	References
Early Lung Cancer Detection by Using Artificial Intelligence System
	1 Chapter 1: Introduction
	2 Chapter 2: Cell Detection and Extraction
		2.1 Rule-Based Algorithm
			2.1.1 Experiments
		2.2 Bayesian Classification
			2.2.1 Experiments
		2.3 Bayesian Classifier vs. Rule-Based Algorithm
	3 Chapter 3: Image Segmentation
		3.1 Mean Shift Procedure
		3.2 Experiments
	4 Chapter 4: Feature Extraction
	5 Chapter 5: Classification
		5.1 Rule-Based Method
		5.2 Artificial Neural Network
		5.3 Support Vector Machine
		5.4 Experiments
		5.5 Comparing Rule-Based, ANN, and SVM Classifiers
	6 Chapter 6: Performance Evaluation with Previous CAD Systems
	7 Chapter 7: Conclusions and Future Works
	References
An Optimal Model Selection for COVID 19 Disease Classification
	1 Introduction
	2 Related Works
	3 Dataset
	4 Optimization Algorithms
		4.1 Stochastic Gradient Descent (SGD) bottou2010large
		4.2 Adam
	5 Deep Learning Architectures
		5.1 CNN
		5.2 Residual Neural Network (ResNet)
		5.3 Densely Neural Network (DenseNet)
	6 Transfer Learning
	7 Learning Rate Scheduler
	8 Results
		8.1 Experimental Setup
		8.2 Training
		8.3 Results
	9 Discussion
	10 Conclusion
	References
Index