Advanced Biosignal Processing (eBook)

Preface 5
Contents 10
Contributors 12
to 1 Biosignals: Acquisition and General Properties 16
1.1 Biopotential Recording 16
1.1.1 Biopotentials Recording Electrodes 17
1.1.1.1 Equivalent Circuit 17
1.1.1.2 Other Electrodes 17
1.1.2 Power Supply Artifact Rejection 18
1.1.2.1 Displacement Currents 18
1.1.2.2 Common Mode Rejection 20
1.1.2.3 Magnetic Induction 20
1.1.2.4 The Right Leg Driver 21
1.1.3 Safety 22
1.1.3.1 Current Limitation 22
1.1.3.2 Galvanic Isolation 22
1.1.4 To Conclude this Section… 22
1.2 General Properties of Some Common Biosignals 23
1.2.1 The Electrocardiogram (ECG) 23
1.2.2 The Electroencephalogram (EEG) 24
1.2.3 Evoked Potentials (EP) 25
1.2.4 The Electromyogram (EMG) 26
1.3 Some Comments 27
1.4 Conclusion 27
to 2 Extraction of ECG Characteristics Using Source Separation Techniques: Exploiting Statistical Independence and Beyond 29
2.1 Introduction 29
2.2 Approaches to Signal Extraction in the ECG 34
2.3 Blind Source Separation 35
2.3.1 Signal Model and Assumptions 35
2.3.2 Why Blind? 39
2.3.3 Achieving the Separation 40
2.4 Second-Order Approaches: Principal Component Analysis 41
2.4.1 Principle 41
2.4.2 Covariance Matrix Diagonalization 41
2.4.3 Limitations of Second-Order Statistics 42
2.5 Second-Order Approaches Exploiting Spectral Diversity 43
2.6 Higher-Order Approaches: Independent Component Analysis 45
2.6.1 Contrast Functions, Independence and Non-Gaussianity 45
2.6.2 Source Separation or Source Extraction? 47
2.6.3 Optimal Step-Size Iterative Search 49
2.7 Exploiting Prior Information 49
2.7.1 Source Statistical Characterization 50
2.7.1.1 Combining Non-Gaussianity and Spectral Features 51
2.7.1.2 Extraction of Sources with Known Kurtosis Sign 51
2.7.2 Reference Signal 53
2.7.3 Spatial Reference 55
2.8 Conclusions and Outlook 56
to 3 ECG Processing for Exercise Test 62
3.1 Introduction 62
3.2 ECG Acquisition During Exercise Test 64
3.3 Interval Estimation and Analysis 64
3.3.1 Heart Rate Variability Analysis 65
3.3.2 PR Interval Estimation 71
3.4 Shape Variations Estimation and Analysis 73
3.4.1 Curve Registration of P-waves During Exercise 74
3.4.1.1 Self-Modelling Registration (SMR) 75
3.4.1.2 Application to Real P-waves 76
3.4.2 Simulation Study 76
3.5 Signal Averaging with Exercise 78
3.5.1 Equal Shape Signals 78
3.5.2 Non Equal Shape Signals 80
to 4 Statistical Models Based ECG Classification 83
4.1 Introduction 83
4.2 Hidden Markov Models 85
4.2.1 Overview 85
4.2.2 Heart Beat Modeling 87
4.2.3 Beat Classification 88
4.2.4 HMM for Beat Segmentation and Classification 90
4.2.4.1 Parameter Extraction 90
4.2.4.2 Training HMMs 91
4.2.4.3 Classifying Patterns 92
4.2.5 HMM Adaptation Adaptation 93
4.2.6 Discussion 94
4.3 Hidden Markov Trees 95
4.3.1 Overview 95
4.3.2 Electrocardiogram Delineation by HMT 96
4.3.2.1 Model Training training 97
4.3.2.2 Scale Segmentation segmentation 98
4.3.2.3 Inter-Scale Fusion 99
4.3.2.4 Discussion 101
4.3.3 Association of HMM and HMT 102
4.4 Conclusions 103
to 5 Heart Rate Variability Time-Frequency Analysis for Newborn Seizure Detection 106
5.1 Identification of Newborn Seizures Using EEG, ECG and HRV Signals 106
5.1.1 Origin of Seizures 107
5.1.2 Seizure Manifestation 107
5.1.3 The Need for Early Seizure Detection 107
5.1.4 Seizure Monitoring Through EEG 108
5.1.5 Limitations of EEG-Based Seizure Identification 108
5.1.6 ECG-Based Seizure Identification 108
5.1.7 Combining EEG and ECG for Robust Seizure Identification 109
5.1.8 The Need for Time-Frequency Signal Processing 109
5.2 Time-Frequency Signal Analysis 110
5.2.1 Addressing the Non-Stationarity Issue 110
5.2.2 Formulation of TFDs 110
5.2.3 Trade-Off Resolution Versus Cross-Terms Trade-Off 113
5.2.4 The Signal Instantaneous Frequency (IF) 113
5.2.5 Multi-Component IF Estimation 114
5.2.6 TFD as a Density Function 115
5.3 ECG Pre-Processing and HRV Extraction 115
5.3.1 Data Acquisition 116
5.3.2 Extracting HRV from ECG 116
5.3.2.1 The ECG and QRS Wave Detection 116
5.3.2.2 Removal of Outliers 118
5.3.2.3 Quantification of HRV 119
5.3.2.4 Detrending 120
5.4 HRV Time-Frequency Feature Extraction 120
5.4.1 HRV Spectral Components 120
5.4.2 Selection of a TFD for HRV Analysis 121
5.4.2.1 Performance Comparison of Relevant TFDs for HRV Analysis 121
5.4.2.2 The Choice of the Modified B Distribution for HRV Analysis 123
5.4.3 Feature Selection in the Time-Frequency Domain 124
5.5 Performance Evaluation and Discussion of the Results 124
5.5.1 Performance of the Classifier 124
5.6 Interpretation and Discussion of the Results 128
5.7 Conclusions and Perspectives 129
to 6 Adaptive Tracking of EEG Frequency Components 133
6.1 Motivation 133
6.1.1 Oscillatory Activity as a Key Neuronal Mechanism 133
6.1.2 Exploring the Oscillatory Content of EEG 135
6.1.2.1 Time-Frequency Analysis 135
6.1.2.2 Filter Bank Decomposition 137
6.1.2.3 Empirical Mode Decomposition 137
6.2 Adaptive Frequency Tracking 138
6.2.1 Single Frequency Tracking 138
6.2.2 A Multiple Frequency Tracking Solution 142
6.2.2.1 Structure of the Filter-Bank 142
6.2.2.2 Example of Tracking Ability 144
6.2.3 An Extension to the Multi-Signal Case 145
6.3 Tracking EEG Oscillations 146
6.3.1 Tracking of a Single EEG Oscillation 146
6.3.2 Tracking of Multiple Neuronal Oscillations 149
6.4 Discussion 150
to 7 From EEG Signals to Brain Connectivity: Methods and Applications in Epilepsy 155
7.1 Introduction 155
7.2 State of the Art 157
7.2.1 Linear and Nonlinear Regression Based Methods 158
7.2.2 Phase Synchronization Based Methods 159
7.2.3 Generalized Synchronization Based Methods 160
7.2.4 Frequency-Dependence of Brain Connectivity Measures 161
7.2.5 Performance Evaluation of Brain Connectivity Measures 162
7.3 Model-Based Comparison of Methods Aimed at Characterizing the Connectivity Between Brain Structures 162
7.3.1 General Model of Interdependence Between Two Time-Series 162
7.3.2 Physiologically-Relevant Model of Depth-EEG Signals 164
7.4 Comparison Criteria 166
7.5 Results 167
7.6 Discussion 169
to 8 Neural Network Approaches for EEG Classification 175
8.1 Introduction 175
8.2 Feature Extraction Algorithms 178
8.3 ANN Based Classification Algorithms 181
8.4 EEG Data Sets 188
8.5 Performance Measures 189
8.6 Conclusion 190
to 9 Analysis of Event-Related Potentials Using Wavelet Networks 193
9.1 Introduction 193
9.1.1 Time-Frequency Analysis of Event-Related Potentials 194
9.2 Wavelet Networks 196
9.2.1 Topology 196
9.2.2 Training Algorithms 197
9.3 Examples of an ERP Study on Children with Attention-Deficit/Hyperactivity Disorder 200
9.3.1 Analysis of Averaged ERPs 201
9.3.2 ERP Single-Trial Analysis 201
9.3.2.1 Amplitude vs. Phase Effect 203
9.3.2.2 Time-On-Task Analysis 203
9.4 Discussion and Further Perspectives 207
to 10 Detection of Evoked Potentials 210
10.1 Evoked Potentials 210
10.1.1 Transient Evoked Potentials, tEP 210
10.1.2 Steady-State Evoked Potentials, ssEP 212
10.1.3 Advanced Evoked Potentials, aEP 212
10.2 Basic Approaches for EP Detection 213
10.2.1 Preliminary Considerations 213
10.2.2 Correlation Detector 215
10.2.3 Energy Detector 216
10.3 Methods for Signal Processing of EP 217
10.3.1 Prewhitening and Detection 217
10.3.2 Enhancement of the SNR 224
to 11 Visual Evoked Potential Analysis Using Adaptive Chirplet Transform 230
11.1 Introduction 230
11.2 Non-windowed ACT Method and Application 232
11.2.1 From "Wavelet" to "Chirplet" 233
11.2.2 Gaussian Chirplet Transform and Adaptive Analysis 235
11.2.3 VEP Analysis with Non-windowed ACT 239
11.3 Windowed ACT Method and Applications 244
11.3.1 Optimal Window Length 244
11.3.2 VEP Analysis with Non-windowed ACT 246
11.4 Summary and Applications of ACT to Other Bio-signals 249
to 12 Uterine EMG Analysis: Time-Frequency Based Techniques for Preterm Birth Detection 254
12.1 Introduction 254
12.2 Event Detection in Uterine EMG 257
12.2.1 Wavelet Packet Transform (WPT) 257
12.2.2 Best Basis Selection 258
12.2.2.1 Introduction 258
12.2.2.2 Criterion of the Best Basis Selection 258
12.2.2.3 WPC Distribution 258
12.2.2.4 Distribution of the Estimated Kullback-Leibler Distance 260
12.2.2.5 Wavelet Packet Selection for Detection Purposes 261
12.2.3 Detection from the Selected Wavelet Packets 262
12.2.3.1 Detection Algorithm 262
12.2.3.2 Change Time Fusion 264
12.2.4 Results on Detection 264
12.2.4.1 Data Description 264
12.2.4.2 Results 264
12.3 Classification of Detected Events 266
12.3.1 Selection of the Best Basis for Classification 266
12.3.2 Classification Algorithms 267
12.3.2.1 K-Nearest Neighbour 267
12.3.2.2 Mahalanobis Distance Based Classification 267
12.3.2.3 Neural Network - Feed Forward Neural Network 268
12.3.2.4 Support Vector Machines 268
12.3.3 Classification of Uterine EMG Events 268
12.4 Classification of Contractions 270
12.4.1 Wavelet Networks (WAVNET) 270
12.4.2 Classifications of Contractions 271
12.4.2.1 Populations 271
12.4.2.2 Results 272
12.5 Discussion and Conclusion 272
to 13 Pattern Classification Techniques for EMG Signal Decomposition 276
13.1 Introduction 276
13.2 EMG Decomposition Process 277
13.3 Supervised Classification of MUPs 279
13.4 Single Classifier Approaches 280
13.4.1 Certainty Classifier 281
13.4.2 Fuzzy k-NN Classifier 282
13.4.3 Matched Template Filter Classifier 284
13.5 Multiple Classifier Approaches 286
13.5.1 Decision Aggregation Module 286
13.5.2 One-Stage Classifier Fusion 287
13.5.2.1 Majority Voting Aggregation 287
13.5.2.2 Average Rule Aggregation 287
13.5.2.3 Fuzzy Integral Aggregation 288
13.5.3 Diversity-Based One-Stage Classifier Fusion 288
13.5.3.1 Assessing Base Classifiers Agreement 288
13.5.4 Hybrid Classifier Fusion 290
13.5.5 Diversity-Based Hybrid Classifier Fusion 291
13.6 Results and Comparative Study 291
13.7 Conclusion 295
to 14 Parametric Modeling of Some Biosignals Using Optimization Metaheuristics 299
14.1 Introduction 299
14.2 Modeling the Dynamics (Specific Case) 300
14.2.1 Defining a Parametric Model 300
14.2.2 Optimization Using Genetic Algorithms 303
14.2.2.1 What are Genetic Algorithms? 303
14.2.2.2 An Example on Simulated BAEPs 304
14.3 Shape Modeling Using Parametric Fitting 306
14.3.1 Defining a Parametric Model 307
14.3.2 Optimization Using PSO 308
14.3.2.1 What is PSO? 308
14.3.2.2 An Example on Event-Related Potential (ERP) 309
14.3.2.3 An Example on ECG Modeling (Fitting) 311
14.4 Conclusion 312
to 15 Nonlinear Analysis of Physiological Time Series 314
15.1 Introduction 314
15.2 Background Concepts 315
15.3 Nonlinear Time Series Analysis 317
15.3.1 Quantification of Fractal Fluctuations 319
15.3.1.1 Detrended Fluctuation Analysis 319
15.3.1.2 Fano/Allan Factor Analysis 319
15.3.1.3 Multifractal Analysis 321
15.3.2 Quantification of Degree of Chaos/Determinism 322
15.3.2.1 Phase Space Reconstruction 322
15.3.2.2 Characterization of the Reconstructed Attractor 324
15.3.3 Quantification of Roughness, Irregularity, Information Content 326
15.3.3.1 Fractal Dimension 326
15.3.3.2 Approximate Entropy 327
15.3.3.3 Multiscale Entropy 327
15.3.3.4 Symbolic Dynamics 328
15.3.4 Methodological Issues 329
15.3.4.1 Surrogate Data Analysis 329
15.3.4.2 Practical Limitations 330
15.4 Quantifying Health and Disease with Nonlinear Metrics 330
15.4.1 Heart Rate Analysis 330
15.4.2 Respiration Pattern Analysis 330
15.4.3 EEG Analysis 331
15.4.4 Human Movement Analysis 331
15.4.4.1 Gait Pattern 331
15.4.4.2 Postural Control 331
15.4.4.3 Daily-Life Physical Activity Pattern 332
15.4.4.4 Movement Irregularity/Roughness 334
15.5 Discussion and Conclusion 334
to 16 Biomedical Data Processing Using HHT: A Review 341
16.1 Introduction 341
16.2 Empirical Mode Decomposition 343
16.3 Cardiorespiratory Synchronization 345
16.4 Human Ventricular Fibrillation 352
16.5 Conclusions 355
to 17 Introduction to Multimodal Compression of Biomedical Data 359
17.1 Introduction 359
17.2 Benefits of Multimodal Recording Analysis: An Example from Sleep Medicine 360
17.2.1 Sleep Recordings 360
17.2.2 Different Levels of "Multimodality" in Sleep Recordings 363
17.2.3 Utility of Multimodal Recording in Sleep Medicine 363
17.2.4 Need for Signal Compression 363
17.3 Biomedical Data Compression 364
17.3.1 Generalities on Data Compression 365
17.3.1.1 Lossless Compression 365
17.3.1.2 Lossy Compression 365
17.3.2 Medical Image Compression 366
17.3.3 Biosignal Compression 368
17.3.3.1 EEG Compression 368
17.3.3.2 ECG Compression 369
17.3.3.3 EMG Compression 370
17.4 Multimodal Compression 370
17.4.1 Joint Image-Biosignal Compression 371
17.4.1.1 Extension to Video-Biosignal Compression 375
17.5 Conclusion 376
Index 43