Study On The Detection Algorithm Of T-wave Alternans In ECG Signals

Cardiovascular disease has become the top killer that threatens human health all around the world, as one of this kind of disease, sudden cardiac death (SCD) should account for about half of such deaths. As a result, how to timely analyze and effectively diagnose SCD has attracted more and more attention from the scholars. T-wave alternans(TWA) is a non-stationary phenomenon, as a noninvasive detection method using ECG, it provides a statistically useful and independent indicator to prognose the risk of ventricular arrhythmias and SCD. Therefore, it is necessary and of great importance to study and work out effective and accurate TWA automatic detection algorithm.Based on the summary of current situations of ECG automatically analysis technologies and TWA detection algorithms, an algorithm with spectral method (SM) using multi-pretreatment is designed to quantify TWA. By the means of window optimizing, beat rejection and alignment, data matrix detrending filtering, the algorithm of the dissertation has improved the ordinary methods. And from the input of original digital ECG signal, after the whole processing flow of the algorithm, the decision whether TWA was present with a simple index is given.The algorithm is then tested and assessed with ECGs in the PhysioNet TWA database, and clinical data combined with the research of our team. The testing results show that this improved algorithm is effective to obtain enhanced and higher sensitivity results in comparison of ordinary methods. It is also showed that the algorithm is effective in the clinical diagnose and provides a new reference for doctors.The dissertation includes five chapters:Chapter 1 elaborates the background and signification of the study, introduces the relevant knowledge of TWA, and way to assess the algorithm, including the PhysioNet TWA database and dynamic electrocardiogram analysis system. Chapter 2 describes the research methods and status of ECG automatically analysis technologies and TWA detection algorithms from three aspects:signal pretreatment, waveform detection and feature parameter extraction, TWA automatic diagnosis.Chapter 3 focuses on the study and design of the TWA automatic detection algorithm of this dissertation, the system is summarized synoptically and dissected into three stages:preprocessing of original ECG signal, acquisition of data for further analysis, and final TWA decision, the whole algorithm is explained in detail compared with ordinary methods.Chapter 4 introduces the algorithm implementation and evaluation, then the testing results are analyzed.Chapter 5 is a summary and a prospect of the entire dissertation.