^{1}

^{1}

^{*}

^{1}

^{1}

Presently T-wave alternans (TWA) has become a clinical index of non-invasive diagnosis for heart sudden death prediction, and detecting T-wave alternate accurately is particularly important. This paper introduces an algorithm for detecting TWA using Poincare mapping method which is a technique for nonlinear dynamic systems to display periodic behavior. Sample series of beat to beat cycles were selected to prepare Poincare mapping method. Vector Angle Index (VAI), which is the mean of the difference between *θ*_{i}_{} (the angle between the line connecting the i point to the origin and the X axis) and 45 degrees was used to present the presence or absence of TWA. The value of 0.9 rad ≤ VAI ≤ 1.03 rad is accepted as a level determinative for presence of TWA. VAI via Poincare mapping method (PM) is used for correlation analysis with T-wave alternans voltage (V_{twa}) by way of the spectral method (SM). The cross-correlation coefficient between V_{twa} and VAI is γ = 0.8601. The algorithm can identify the absence and presence of TWA accurately and provide idea for further study of TWA-PM.

A number of studies recently indicate that ventricular arrhythmias are one of the primary causes of cardiac death, and the microvolt T-wave alternans (MTWA) is an important index for ventricular arrhythmias prediction. TWA is a phenomenon of electro cardio variation that beat to beat variation of T-wave morphology and polarity at constant heart rate is embodied in neat cardiac rhythm [

In accordance with the statistical method difference of TWA detection, the methods with pathologic significance of detecting MTWA are divided into three kinds: Short Time Fourier Transform (STFT), symbol transform and nonlinear methods [

The article presents the Poincare mapping (PM) [_{twa} (of SM) and VAI (of PM). This algorithm was applied in MIT/BIH Arrhythmia database and European ECG ST-T database, and the cross-correlation coefficient between V_{twa} and VAI is ã = 0.8601.

Firstly, the measured ECG S was filtered, and in this paper, integral coefficient was used to eliminate 50 HZ power-line interference, and then the signal was filtered using threshold denoising algorithm by applying bior 2.2 wavelet [

With the aim of detecting MTWA more accurately, this article adopts a T wave analysis window method that L = 128 ECG complex was selected for TWA measurement and there were m = 7 sampling points in every cardiac cycle [

The T-wave analysis window was divided equally into m parts in a cardiac cycle, and the sampling interval was ID between two sampling points [

(2)

In line with the (1) and (2), 7 × L selected samples of T-wave can be get across L consecutive heartbeat cycle and 7 sampling points in a cardiac cycle. At the same time, we obtain a set of signal samples, and a new sequence with subtraction between adjacent samples is formed. When the sequence of feature point was known [

Among them, and denoted the angle between the line connecting the i point to the origin and the X axis; N was the total point of Poincare map. Used for Matlab simulation, 45 degrees were turned into radian system namely for 0.7854. VAI value denoted the dispersed degree of the adjacent T-wave amplitude difference along the 45˚ line. When vector angle index is 0.9 rad ≤ VAI ≤ 1.03 rad, TWA was present in the ECG signal. When VAI < 0.9 rad or VAI > 1.03 rad, there was no TWA. Poincare maps for T-wave detection are shown in

Spectral method (SM) widely applied to the current detection of TWA is a frequency domain analysis method, and has high accuracy [_{twa}, and the correlation analysis proved the effectiveness of PM. Spectral method for T-wave detection are shown in

In the article, we adopted MIT/BIH Arrhythmia database and European ECG ST-T database. The sampling fre-

quency in former was 360 HZ while in the latter 250 HZ. For purposes of brevity, the signal from these databases was resampled with 200 HZ so that the analysis became simpler. We showed part of the simulation results in

From the following data in _{twa} and VAI is γ = 0.8601. In the Matlab7.0 environment, the discrete data V_{twa} and VAI in

The fitting curve of V_{twa} and VAI are shown in

According to a lot of the simulation data results, we accepted that the presence of TWA by way of PM is determined on the basis of 0.9 rad ≤ VAI ≤ 1.03 rad. Because of the cross-correlation coefficient between V_{twa} and VAI 0.8601, the strong correlation between VAI via PM and V_{twa} by means of the mature method of SM is proved. At the same time, it is also shown that Vector Angle can be applied to TWA detection. SP demands the complex of plenty of beat-to-beat circles to detect the presence or absence TWA accurately. By way of PM T-wave alternans voltage between any adjacent beats can be get, and Compared with SP, PM is more simpler [

This project was supported by Shandong Provincial Nature Science Foundation (ZR2010M020) and Shandong Provincial Key Scientific and Technological Project (2007GG10001018).