Advances in Non-Invasive Biomedical Signal Sensing and Processing with Machine Learning

Description

This book presents the modern technological advancements and revolutions in the biomedical sector. Progress in the contemporary sensing, Internet of Things (IoT) and machine learning algorithms and architectures have introduced new approaches in the mobile healthcare. A continuous observation of patients with critical health situation is required. It allows monitoring of their health status during daily life activities such as during sports, walking and sleeping. It is realizable by intelligently hybridizing the modern IoT framework, wireless biomedical implants and cloud computing. Such solutions are currently under development and in testing phases by healthcare and governmental institutions, research laboratories and biomedical companies. The biomedical signals such as electrocardiogram (ECG), electroencephalogram (EEG), Electromyography (EMG), phonocardiogram (PCG), Chronic Obstructive Pulmonary (COP), Electrooculography (EoG), photoplethysmography (PPG), and image modalities such as positron emission tomography (PET), magnetic resonance imaging (MRI) and computerized tomography (CT) are non-invasively acquired, measured, and processed via the biomedical sensors and gadgets. These signals and images represent the activities and conditions of human cardiovascular, neural, vision and cerebral systems. Multi-channel sensing of these signals and images with an appropriate granularity is required for an effective monitoring and diagnosis. It renders a big volume of data and its analysis is not feasible manually. Therefore, automated healthcare systems are in the process of evolution. These systems are mainly based on biomedical signal and image acquisition and sensing, preconditioning, features extraction and classification stages. The contemporary biomedical signal sensing, preconditioning, features extraction and intelligent machine and deep learning-based classification algorithms are described. Each chapter starts with the importance, problem statement and motivation. A self-sufficient description is provided. Therefore, each chapter can be read independently. To the best of the editors’ knowledge, this book is a comprehensive compilation on advances in non-invasive biomedical signal sensing and processing with machine and deep learning. We believe that theories, algorithms, realizations, applications, approaches, and challenges, which are presented in this book will have their impact and contribution in the design and development of modern and effective healthcare systems. Foreword Preface Acknowledgements Contents Contributors Chapter 1: Introduction to Non-Invasive Biomedical Signals for Healthcare 1.1 Introduction to Biomedical Signals 1.2 Invasive and Non-Invasive Procedures 1.3 Non-Invasive Biomedical Signals 1.3.1 Electroencephalography (EEG) 1.3.2 Magnetoencephalography (MEG) 1.3.3 Electromyography (EMG) 1.3.4 Electrocardiography (ECG) 1.3.5 Electrooculography (EOG) 1.3.6 Phonocardiogram (PCG) 1.3.7 Photoplethysmography (PPG) 1.3.8 Magnetic Resonance Imaging (MRI) 1.4 Biomedical Signal Processing 1.4.1 Signal Acquisition 1.4.2 Signal Visualization and Annotation 1.4.3 Artifacts Removal and Preprocessing 1.4.4 Feature Extraction 1.5 Machine Learning in Biomedical Signal Analysis 1.6 Brain-Computer Interface BCI) 1.7 Neurofeedback & Biofeedback Systems 1.8 Conclusion 1.9 Teaching Assignments References Chapter 2: Signal Acquisition Preprocessing and Feature Extraction Techniques for Biomedical Signals 2.1 Introduction 2.2 The Biomedical Signal Acquisition and Processing 2.2.1 The Analog to Digital Conversion 2.2.2 The Digital Filtering 2.2.3 The Windowing 2.3 The Features Extraction Techniques 2.3.1 The Spectral Analysis 2.3.1.1 The Fourier Transform (FT) 2.3.1.2 The Parametric Model Based Methods 2.3.1.3 The Subspace Based Methods 2.3.2 The Time-Frequency Analysis 2.3.2.1 The Short-Time Fourier Transform 2.3.2.2 The Wavelet Transform 2.3.2.3 The Empirical Wavelet Analysis 2.3.2.4 The Empirical Mode Decomposition 2.4 Conclusion 2.5 Assignments for Readers References Chapter 3: The Role of EEG as Neuro-Markers for Patients with Depression: A Systematic Review 3.1 Introduction 3.2 Brain Structure and Depression 3.2.1 Brain Structure 3.2.2 Depression Types 3.2.2.1 Major Depression Disorder (MDD) 3.2.2.2 Premenstrual Dysphoric Disorder (PMDD( 3.2.2.3 Psychotic Depression 3.2.2.4 Postpartum Depression (PPD) 3.2.2.5 Persistent Depressive Disorder (PDD) 3.2.2.6 Seasonal Affective Disorder (SAD) 3.2.3 Effect of Depression on the Brain 3.3 Depression Diagnosis 3.3.1 Biomarkers 3.3.2 Psychological Assessments 3.3.3 Physiological Measurements 3.4 EEG-Based Depression Recognition Neuromarker 3.4.1 EEG and the Brain 3.4.2 Experimental EEG Protocol for Recognizing Depression 3.4.3 EEG Publicly Available Dataset for Depression Diagnosis 3.4.4 Function of EEG in Depression Detection and Classification 3.4.4.1 EEG Signal Acquisition Stage 3.4.4.2 Preprocessing Stage 3.4.4.3 Features Extraction Stage 3.4.4.3.1 Linear Spectral Features 3.4.4.3.2 Nonlinear Features 3.4.4.4 Dimensionality Reduction Stage 3.4.4.5 Depression Classification Techniques 3.5 Discussion 3.6 Conclusion References Chapter 4: Brain-Computer Interface (BCI) Based on the EEG Signal Decomposition Butterfly Optimization and Machine Learning 4.1 Introduction 4.2 The Evolution of BCI 4.3 Studies on the BCI 4.4 Methodology 4.4.1 Brain Waves Acquisition 4.4.2 Analog Amplifier and Filter 4.4.3 Analog to Digital (A/D) Conversion 4.4.4 Signal Conditioning 4.4.5 Features Extraction 4.4.6 Dimension Reduction 4.4.7 Classification 4.4.7.1 Support Vector Machine Classifier (SVM) 4.4.7.2 k-Nearest Neighbors (k-NN) 4.4.7.3 Artificial Neural Network (ANN) 4.4.8 Evaluation Measures 4.4.8.1 Accuracy 4.4.8.2 Precision 4.4.8.3 Recall 4.4.8.4 Specificity 4.4.8.5 F-Measure 4.4.8.6 Kappa 4.5 Results and Discussion 4.6 Conclusion 4.7 Assignments for Readers References Chapter 5: Advances in the Analysis of Electrocardiogram in Context of Mass Screening: Technological Trends and Application of AI Anomaly Detection 5.1 Introduction 5.2 Evolution of Views on the Role of the Electrocardiogram in Assessing the Risk of Major Adverse Cardiovascular Events 5.3 The Systems of Electrocardiographic Leads, Electrocardiogram with Limited Number of Leads for Heart Disease Screening 5.4 The Generations of ECG Analysis, some Modern Approaches Based on Mathematical Transformation of ECG Signal 5.5 Anomaly Detection in ECG Using Machine Learning Approach 5.6 Isolation Forest Anomaly Detection for Quantifying the Deviation of Signal Averaged ECG from Population Norm 5.6.1 Isolation Forest Unsupervised Anomaly Detection 5.6.2 Subjects Data 5.6.3 Quantification of the Distance to the Norm 5.6.4 Experiment Results 5.7 Teaching Assignment 5.8 Conclusion References Chapter 6: Application of Wavelet Decomposition and Ma-Chine Learning for the sEMG Signal Based Ges-Ture Recognition 6.1 Introduction 6.2 Literature Review 6.2.1 Background 6.2.2 Preprocessing for sEMG Based Gesture Recognition 6.2.3 Feature Selection Techniques for sEMG Based Gesture Recognition 6.2.4 Machine Learning and Deep Learning Techniques for sEMG Based Gesture Recognition 6.3 Methodology 6.3.1 Dataset 6.3.2 Machine Learning Algorithms 6.3.2.1 Support Vector Machine Classifier (SVM) 6.3.2.2 K-Nearest Neighbor (KNN) 6.3.3 Evaluation Measures 6.3.3.1 Accuracy 6.3.3.2 Precision 6.3.3.3 Specificity 6.3.3.4 Recall 6.3.3.5 F-Score 6.3.3.6 Kappa Statistics 6.4 Results and Discussion 6.5 Conclusion 6.6 Assignments for Readers References Chapter 7: Review of EEG Signals Classification Using Machine Learning and Deep-Learning Techniques 7.1 Introduction 7.2 Signal Pre-Processing 7.3 Features Extraction 7.3.1 Fast Fourier Transform (FFT) 7.3.2 Short-Time Fourier Transform (STFT) 7.3.3 Continuous Wavelet Transform 7.3.4 Discrete Wavelet Transform 7.3.5 Wavelet Packet Decomposition (WPD) 7.4 Features Selection 7.5 Machine Learning Techniques 7.6 Deep Learning Techniques 7.7 Case Studies 7.7.1 Epilepsy Detection 7.7.1.1 Dataset 7.7.1.2 Methodology 7.7.2 Schizophrenia Detection 7.7.2.1 Dataset 7.7.2.2 Methodology 7.8 Discussion 7.9 Conclusion 7.10 Assignments References Chapter 8: Biomedical Signal Processing and Artificial Intelligence in EOG Signals 8.1 Introduction 8.1.1 EOG Fundamentals 8.1.2 EOG Signal Measurement 8.2 EOG Signal Denoising 8.3 Compression 8.4 EOG Feature Processing 8.4.1 Feature Extraction 8.4.1.1 Time-Domain Features 8.4.1.2 Frequency Domain Features 8.4.1.3 Time-Frequency Features 8.4.1.4 Non-Linear Features 8.4.2 Feature Selection 8.4.3 Feature Normalization 8.5 Classification 8.5.1 Machine Learning Techniques 8.5.2 Deep Learning Techniques 8.6 Decision-Making 8.6.1 Intelligent Decision Support Systems 8.6.2 Learning Approaches 8.7 Discussion 8.8 Conclusions References Chapter 9: Peak Spectrogram and Convolutional Neural Network-Based Segmentation and Classification for Phonocardiogram Signals 9.1 Introduction 9.1.1 Auscultation 9.1.2 Phonocardiogram Signal 9.1.3 PCG Signal Acquisition 9.2 Related Work 9.2.1 Segmentation 9.2.2 Extracted Features and Classifiers 9.2.3 Unsegmented PCG Classification 9.3 Quality Assessment and Pre-processing of PCG Signals 9.3.1 Evaluation Criteria 9.3.2 Filtering and Spike Removal 9.4 Single and Multi-Level Threshold-Based Peak Detection Methods 9.5 Segmentation Methods of PCG Signals 9.5.1 Segmentation Based on Statistical Features and Support Vector Machine 9.5.2 Segmentation Based on Spectrograms and Convolutional Neural Network 9.6 Post-processing and Classification of PCG Signals 9.6.1 Post-processing and PCG Labeling 9.6.2 PCG Classification 9.7 Experimentation on the PhysioNet2016 Challenge Dataset 9.7.1 Dataset 9.7.2 Results of Pre-processing 9.7.3 Results of Segmentation 9.7.4 Results of Post-processing 9.7.5 Results of PCG Segmentation 9.8 Comparison Analysis and Discussions 9.9 Conclusions References Chapter 10: Eczema Skin Lesions Segmentation Using Deep Neural Network (U-Net) 10.1 Introduction 10.1.1 Eczema Area and Severity Index Measurement 10.1.2 Segmentation 10.2 Deep Learning Approach in Segmentation 10.2.1 Neural Network 10.2.2 Convolutional Neural Network (CNN) 10.2.3 Region-Based CNN (R-CNN) 10.2.4 Fully Convolutional Network (FCN) 10.2.5 Summary of Lesion Segmentation Literature 10.3 Methodology 10.3.1 Image Acquisition and Ground Truth Preparation 10.3.2 Image Pre-processing 10.3.2.1 Data Structure Preparation for Supervised Learning Methods 10.3.2.2 Adaptive Light Compensation (ALC) 10.3.2.3 Color Model Conversion 10.3.3 Data Augmentation 10.3.4 Image Segmentation 10.3.4.1 U-Net Architecture 10.3.4.2 U-Net Implementation 10.3.5 Image Post Processing 10.3.6 Segmentation Performance Analysis 10.4 Results and Discussion 10.4.1 Image Pre-processing for Ground Truth Images 10.4.2 Image Segmentation 10.4.2.1 Color Channels 10.4.2.2 Adaptive Light Compensation Technique (ALC) 10.4.3 Post-processing 10.4.4 Analysis of the Effect of Varying Kernel Number in Convolution Layer 10.4.5 Analysis of the Effect of Varying Steps per Epoch 10.4.6 Analysis of the Effect of Varying Number of Epochs 10.4.7 Comparison of Machine Learning and Deep Learning Methods 10.5 Conclusions References Chapter 11: Biomedical Signal Processing for Automated Detection of Sleep Arousals Based on Multi-Physiological Signals with Ensemble Learning Methods 11.1 Introduction 11.2 Polysomnography 11.2.1 EEG 11.2.1.1 Special Patterns in EEG 11.3 Sleep Stage 11.4 Methodology 11.4.1 Ensemble Learning 11.4.1.1 Bootstrap Aggregation (Bagging) 11.4.1.1.1 Random Forest (RF) 11.4.1.2 Boosting 11.4.1.2.1 Gradient Boosting Decision Tree (GBDT) 11.4.1.2.2 Light Gradient Boosting Machine (LightGBM) 11.4.2 Evaluating Performance 11.4.3 Data Description 11.4.4 Pre-Processing 11.4.4.1 EEG and EOG Signals 11.4.4.2 EMG Signal 11.4.4.3 ECG Signal 11.4.4.4 Airflow Signal 11.4.4.5 Signal Segmentation 11.4.4.6 Labeling Epochs 11.4.5 Feature Extraction 11.4.5.1 Features Extracted from EEG Signals 11.4.5.1.1 Frequency Features 11.4.5.1.2 Time-Frequency Features 11.4.5.1.3 Nonlinear Features 11.4.5.2 Features Extracted from EMG 11.4.5.3 Features Extracted from SaO2 11.4.5.4 Features Extracted from Airflow 11.4.5.5 Features Extracted from ECG 11.4.6 Data Balancing 11.4.7 Feature Selection 11.5 Classification Result 11.6 Conclusion References Chapter 12: Deep Learning Assisted Biofeedback 12.1 Introduction 12.2 Current Biofeedback and Neurofeedback Devices and Practice 12.3 Deep Learning Models for Electroencephalography Analysis 12.4 Deep Learning Assisted Biofeedback (DLAB) 12.4.1 PP-net: EEG Online Preprocessing 12.4.2 Sel-net: Classifying “Targeting” Signals for Feedback 12.4.3 Control-net: Extracting and Classifying for Feedback 12.4.4 FB-net: EEG Online Feedback 12.4.5 Config-net: Predictive Maintenance and Feedback Modulation 12.4.6 iClean-net: Cleaning Performance Perturbances 12.4.7 PN-net: Feedback Quality Control System and Interactive Database 12.4.8 Assess-net: Feedback Modulation Control Database 12.4.9 E-net 12.5 Discussion 12.6 Conclusions References Chapter 13: Estimations of Emotional Synchronization Indices for Brain Regions Using Electroencephalogram Signal Analysis 13.1 Introduction 13.2 Related Works 13.3 Materials and Methods 13.3.1 Subjects and Experimental Procedure 13.3.2 Preprocessing Stage 13.3.2.1 Conventional Filtering 13.3.2.2 Empirical Mode Decomposition with Wavelet (EMD − WT) Hybrid Denoising Technique 13.3.3 Features Extraction Stage 13.3.3.1 Linear Features 13.3.3.2 Nonlinear Features 13.3.4 Features Selection Using Statistical Analysis 13.3.5 Emotion Classification Stage 13.4 Results and Discussions 13.4.1 Results of Preprocessing Stage 13.4.2 Results of Features Extraction Stage 13.4.2.1 Results of Linear Features 13.4.2.2 Results of Nonlinear Features 13.4.3 Results of Features Selection and Emotion Classification Stages 13.4.3.1 Classification Results of Linear Features 13.4.3.2 Classification Results of Nonlinear Features 13.5 Conclusion References Chapter 14: Recognition Enhancement of Dementia Patients’ Working Memory Using Entropy-Based Features and Local Tangent Space Alignment Algorithm 14.1 Introduction 14.2 Related Works 14.3 Methods and Materials 14.3.1 Participants and EEG Recording 14.3.2 Preprocessing Stage 14.3.2.1 Conventional Filters 14.3.2.2 AICA–WT Technique Methodology 14.3.3 Features Extraction 14.3.3.1 Fuzzy Entropy (FuzzEn) 14.3.3.2 Fluctuation-Based Dispersion Entropy (FDispEn) 14.3.3.3 Bubble Entropy (BubbEn) 14.3.4 Statistical Analysis 14.3.5 Preliminary Feature Processing Prior Classification 14.3.6 Local Tangent Space Alignment (LTSA) 14.3.7 Dementia Classification Techniques 14.4 Results and Discussion 14.4.1 Results of Preprocessing Stage 14.4.2 Results of Dementia Recognition by Statistical Analysis 14.4.3 Results of Dementia Recognition by Classification and Performance Measure 14.5 Conclusion References