Advanced Methods and Tools for ECG Data Analysis - Part 1 ppt

[...]...P1: Shashi August 24, 2006 11 :32 Chan-Horizon Azuaje˙Book x Contents 11 .9 Duration Modeling for Robust Segmentations 11 .10 Conclusions References 312 316 316 CHAPTER 12 Supervised Learning Methods for ECG Classification/Neural Networks and SVM Approaches 319 12 .1 Introduction 12 .2 Generation of Features 12 .2 .1 Hermite Basis Function Expansion 12 .2.2 HOS Features of the ECG 12 .3 Supervised... Classifiers 12 .3 .1 Multilayer Perceptron 12 .3.2 Hybrid Fuzzy Network 12 .3.3 TSK Neuro-Fuzzy Network 12 .3.4 Support Vector Machine Classifiers 12 .4 Integration of Multiple Classifiers 12 .5 Results of Numerical Experiments 12 .6 Conclusions Acknowledgments References 319 320 3 21 323 324 325 326 327 328 330 3 31 336 336 336 CHAPTER 13 An Introduction to Unsupervised Learning for ECG Classification 339 13 .1 Introduction... Introduction 13 .2 Basic Concepts and Methodologies 13 .3 Unsupervised Learning Techniques and Their Applications in ECG Classification 13 .3 .1 Hierarchical Clustering 13 .3.2 k-Means Clustering 13 .3.3 SOM 13 .3.4 Application of Unsupervised Learning in ECG Classification 13 .3.5 Advances in Clustering-Based Techniques 13 .3.6 Evaluation of Unsupervised Classification Models: Cluster Validity and Significance 13 .4... methodologies, and Chapter 8 presents a comparative study of ECG derived respiration techniques Chapter 9 presents advanced techniques for extracting relevant features from the ECG, and Chapter 10 uses these techniques to describe a robust ST-analyzer Chapter 11 presents a wavelet and hidden Markov model–based procedure for robust QT -analysis Chapter 12 describes techniques for supervised classification and hybrid... multidimensional signal analysis is well known to the ECG signal analyst, and parallel analysis of the ECG (or ECG- derived parameters) almost always enhances an algorithm’s performance For instance, blood pressure waves contain information that is highly correlated with the ECG, and analysis of these changes can help reduce false arrhythmia alarms The ECG is also highly correlated with respiration and can be used... presents an overview of simple, practical ECG and beat-tobeat models, together with methods for applying these models to ECG analysis Chapter 5 describes a unified framework for linear filtering techniques including wavelets, principal component analysis, neural networks, and independent component analysis Chapter 6 discusses methods and pitfalls of nonlinear ECG analysis, with a practical emphasis on... Models: Cluster Validity and Significance 13 .4 GSOM-Based Approaches to ECG Cluster Discovery and Visualization 13 .4 .1 The GSOM 13 .4.2 Application of GSOM-Based Techniques to Support ECG Classification 13 .5 Final Remarks References 339 339 354 359 362 About the Authors 367 Index 3 71 3 41 342 343 343 346 347 350 352 352 P1: Shashi August 24, 2006 11 :32 Chan-Horizon xii Azuaje˙Book Preface • • • • • • • rate... 1. 5), and the ECG represents the difference between each of these electrodes (V1–6) and the central terminal [as in (1. 3)] 10 The His bundle is a collection of heart muscle cells specialized for electrical conduction that transmits electrical impulses from the AV node, between the atria and the ventricles to the Purkinje fibers, which innervate the ventricles P1: Shashi August 24, 2006 11 :34 Chan-Horizon... cardiac filling pressure, and the lungs themselves, during the respiratory cycle Figure 1. 10 Sinus tachycardia (From: [2] c 2004 MIT OCW Reprinted with permission.) Figure 1. 11 Sinus bradycardia (From: [2] c 2004 MIT OCW Reprinted with permission.) Figure 1. 12 Sinus arrhythmia (From: [2] c 2004 MIT OCW Reprinted with permission.) P1: Shashi August 24, 2006 11 :34 Chan-Horizon 16 Azuaje˙Book The Physiological... very important, particularly if the algorithm’s output is fed to another data fusion algorithm The inverse problem It is well known that no unique solution exists for the inverse problem in ECG mapping Attempts to reconstruct the dipole moment in the ECG have had some degree of success, and recent models for the ECG have proved useful in this respect P1: Shashi August 24, 2006 11 :32 Chan-Horizon Preface . Segmentation and Feature Extraction 2 91 11. 1 Introduction 2 91 11. 2 The Electrocardiogram 292 11 .2 .1 The ECG Waveform 292 11 .2.2 ECG Interval Analysis 292 11 .2.3 Manual ECG Interval Analysis 293 11 .3 Automated. 305 11 .7.3 Types of Model Segmentations 307 11 .7.4 Performance Evaluation 307 11 .8 Wavelet Encoding of the ECG 311 11 .8 .1 Wavelet Transforms 311 11 .8.2 HMMs with Wavelet-Encoded ECG 312 P1: Shashi August. Processes and Markov Models 297 11 .6.3 Hidden Markov Models 298 11 .6.4 Inference in HMMs 3 01 11. 6.5 Learning in HMMs 302 11 .7 Hidden Markov Models for ECG Segmentation 304 11 .7 .1 Overview 305 11 .7.2 ECG