Research Of Detecting The Nonstationary T-wave Alternans Based On Wavelet And L1 Trend Filter

As shown by clinical research, malignant heart diseases such as sudden cardiac death have been critical diseases that threaten human health in our modern society. Through years of clinical practice shows that ECG (electrocardiograph) analysis is an important means for clinical diagnosis of heart disease patients. T-wave alternans (T wave alternans, TWA) are related to ventricular arrhythmia and sudden cardiac death and other cardiac diseases. T wave is a weak signal, that is difficult to detect TWA with eyes, and the non-stationary of T wave alternans causes the detection difficult, so we need to use specific algorithms to detect TWA phenomenon. TWA has become a non-invasive and independence predictor of cardiovascular diseases, so research on TWA also has a high clinical value.This paper mainly focuses on TWA detection, specifically the work is divided into the following sections:(1) Preprocessing algorithm of ECG signal.After studying the noise properties of ECG signal, based on wavelet denoising theory, a novel threshold function is constructed by both soft and hard thresholds. This algorithm is a good solution to soft and hard threshold method defect.This improved wavelet threshold denoising method can effectively improves the effect of denoising and well preserve the details provided a good platform for further feature point detection and TWA detection.(2) Identification algorithm of features of R wave points. After ECG signal reprocessing, through the R wave points recognition algorithm, this paper presents the threshold algorithm based on wavelet to detect and track the R wave points. The proposed algorithm is very good to detect and track the R wave points and solves the existing problem of the R-wave detection.Used typical datas of TWAdb database as a verificationfor this algorithm and the results showed that this algorithm detected the R-wave points with high accuracy.(3) TWA detection algorithm. This paper presents a strong robustness and better time resolution algorithm-L1 trend estimation algorithm. In the aspect of robustness, time resolution and accuracy, the algorithm for the detection of non-stationary TWA are better than Modified Moving Average Method and Spectrum Method.Through TWA detection By TWAdb database, the results showed that this algorithm can be achieved manually labeling standards, and has a high robustness, and the ability able to meet the current clinical diagnostic prediction.