Research Of T Waves Alternans Detection Based On Bayes Estimation And Nonparametric Test

Sudden cardiac death causes 540000 deaths in China each year and most of the cause of death is malignant ventricular arrhythmia. A large number of clinical trials and research reports have shown that T Wave alternating (TWA) has a close relationship with the ventricular arrhythmias and sudden cardiac death. TWA has become a non-invasive, independent, and statistically significant indicator of malignant ventricular arrhythmias and sudden cardiac death. But due to weak TWA phenomena and the complexity of the noise in ECG, improve the accuracy of the TWA testing method and robustness, becomes the key to promote the TWA widely clinical application.This paper introduces the general system model of the TWA test, mainly including the preprocessing module, T wave extraction and alignment module, and the TWA testing module. In this paper, firstly research present situation and progress in the study of TWA, and use classic pretreatment method to preprocess the ecg signal. We mainly focus on the T wave extraction and alignment, and TWA detection method these two aspects and proposed new algorithms. The following is the main work in this paper:(1) T waves extraction and alignment. Firstly, classic gaussian electrical model is used for describing each cardiac cycle of ECG waveform. Then use particle swarm optimization algorithm to fit every cardiac cycle waveform. Thus we can get phase, amplitude and length parameters of T wave in a cardiac cycle. By using these parameters, T wave can be accurately extracted. After extraction the length of each T wave is different, this makes TWA dectection unconveniently, so we use the cubic spline interpolation method to align the T wave matrix.(2) Qualitative and quantitative analysis of TWA. Firstly we described various testing methods which proposed by others. Then we put forward a new joint TWA testing and quantitative analysis method. First, use the method of nonparametric test to judge whether there is a TWA. If there is TWA phenomenon, then use bayesian estimation method to estimate TWA alternating amplitude. These nonparametric method and parameter combination method can greatly improve the detection accuracy.Compared with the traditional methods, the TWA testing methods proposed in this paper have certain features and benefits. T wave extraction method in this paper ues the T wave shape information, but common extraction method not use. Then we can get a more accurate description of the T wave. Also particle swarm optimization is used to fit the curve, so we can get a fast convergence rate and a high fitting accuracy. For the Signal model is not comprehensive, use nonparametric test method can reduce the false detection. The bayesian estimation method import a priori information about the noise, these can improve the accuracy of TWA estimate.