Research On Key Techniques Of T-wave Alternans Detection In Electrocardiogram

Sudden cardiac death has become the second deadly disease after tumor. It is caused mostly by malignant ventricular arrhythmia. A large number of clinical trials and research literature suggests that TWA(T Wave Alternans, TWA) has a close relationship with ventricular arrhythmia, which can predicts the occurrence of malignant ventricular arrhythmia and SCD, and it is one of the most important index having independent and statistical significance. At present, the research of TWA phenomenon detection algorithm has caused wide public concern. It is expected to become a superior, noninvasive assessment of the risk of sudden cardiac death. But due to transient phenomena of TWA, and the complexity of the noise in ECG,it is difficult to extract TWA index of diagnosis value, which has become a key problem affecting the TWA for predicting sudden death in clinical application. Based on the general framework of TWA detection, the TWA detection scheme can be divided into three stage:preprocessing, T wave extraction and alignment, TWA analysis. Researches are done in all phases of the TWA detection algorithm, concrete research content is as follows:(1) This dissertation proposes an adaptive denoising algorithm based on the combination of mathematical morphology method and EMD method.The main work of preprocessing stage:denoise and keep the T wave information for the TWA detection. Based on the mathematical morphology method, it extracts peak valley of ECG signal waveform after the EMD decomposition, so as to retain most of the electrocardiogram signal characteristic waveform information. The peak valley waveform contains only high frequency noise waveform and a small number of feature information. For the peak valley waveform, it eliminate the noise by designing adaptive threshold method to calculate the threshold value of peak valley information processing, so as to keep the T wave in electrocardiogram (ecg) in maximum level. While aiming at filtering out baseline wander in electrocardiogram (ecg), the traditional method median filtering method filter out the baseline drift easily having the phenomenon of "steps", by conbining the denoising method proposed before with the median filtering method, it proposes a method not only filtering out the baseline wander but also solving the "steps" problem in median filtering method. In this paper, the method proposed can not only filter out the baseline noise but also not destroy useful features of ecg waveform. Simulation shows that this method can effectively filter out random noise and baseline wander noise which influence TWA detection mostly.(2) It proposes a method of extraction and align T waves based on the combination of particle swarm method and ECG signal model.The main work of T wave extraction phase:For the specific purpose of TWA detection, it not only requires accurate extraction of T wave, but also need ensures that it is consistent alignmeng of T wave on the load-point and width in different cardiacs. Based on the model of gaussian kernel for ECG waveform, the P wave, Q wave, R wave, S wave and T wave of electrocardiogram (ecg) are presented by a three parameters of gaussian function for time, amplitude and width. Using waveform curve generating from the ecg waveform model to fitting the realistic waveform of ECG signal, it can extract the position of the T wave and load-point correctly. It transforms the T wave extraction problem into the the multi-peak gaussian curve fitting problem. Based on the fitting idea, we do the optimal extraction of T wave using particle swarm optimization algorithm. Simulation shows that the proposed method can extract and align of T wave accurately.In the stage of TWA analysis, it researches the electrocardiogram TWA calculation method of quantitative and qualitative, in view of hot and difficult problems from the present TWA detection algorithm research, This article studies the TWA analysis method respectively from the building and solving dynamic TWA space equation, convex optimization and nonparametric statistical method.(3) It proposes a non-gaussian TWA analysis method based on particle filter.TWA phenomenon in electrocardiogram has the property of nonstationarity, non-linearity and non-gaussian, so how to improve the robustness of TWA detection method is a difficult problem of current research TWA analysis method. Due to the effect of patients with pathological conditions, TWA phenomenons in ECG tend to be nonstaionary and non-gaussian. This paper establish a nonlinear, nonstaionary and non-gaussian state space equation, based on this state space equation, this paper use particle filter method to estimate TWA values. Simulations show that this method of TWA has good analytical performance in estimate nonstaionary and non-gaussian TWA phenomenon.(4) It proposes a nonstationary TWA analysis method based on convex optimization analysis method.Because the ECG denoising method for TWA detection specific purpose, it must retain the maximum information of T wave in ecg in order to not affect the accuracy of the subsequent TWA detection. Thus the ecg data may contain some noise in the TWA analysis phase of TWA detection method. Studying nonstationary TWA analysis method of noise resistance has always been the hot spot of the current TWA analysis research. From the perspective of convex optimization, this paper analysis TWA phenomenon using L1 trend filtering to estimate nonstationary TWA. The simulation shows that the method can estimate nonstationary TWA accurately and has good antinoise performance.(5) It proposes a TWA analysis method based on the combining of wavelet analysis and the nonparametric Bootstrap method.The use of statistical methods to detect TWA has been one of the important trends of TWA analysis research. Based on actual TWA experience in probability and statistics model, it put forward a TWA detection method based on the nonparametric statistical method Bootstrap. Firstly, using time-frequency analysis, ECG data can be divided into relatively stationary segments of electrocardiogram (ecg). For each segment of small sample data, this paper uses the Bootstrap resampling method to qualitative and quantitative estimate TWA index, so that it has higher robustness analysis TWA when the actual TWA probability model is difficult to determine. Simulation shows that the method has good robustness in the premise of TW probability model is difficult to accurately determine.It studys the key technical problems around TWA detection algorithm. For the specific problem of TWA detection, it proposes reliable ECG denoising algorithm and T wave extraction algorithm. Based on the current hot and difficult problems in the research of TWA analysis method, methods are respectively solved through three angles: spaced solving, convex optimization and nonparametric statistic. Simulation results show that the research results of this paper has good performance in improving the robustness of TWA detection algorithm and. It can be used to promote the TWA index noninvasive predicting sudden cardiac death technology and provide a robust TWA index quantitative and qualitative estimation method.