The Research On T Wave Alternans Automatic Detection Technology

A lot of research has shown that T wave alternans (TWA) was closely related to sudden cardiac death and ventricular tachycardia. TWA was also gradually accepted as an important reference for heart disease diagnosis and evaluation. The early TWA detection could only be done with manual check which costs much time. Recently the TWA auto detection method has become a hotspot of research, but the TWA's quantitative measurement and time location were still unsolved.Preprocessing was done to the ECG signals for further analysis. After the comparison of several preprocessing methods, FIR low-pass filter and wavelet transform method were chosen to eliminate the high-frequency noise and baseline wander which may be applied in real-time ECG terminal.Derivative method was used to detect positive R-wave peak. Adaptive threshold, adaptive RR interval and R wave search-back method were also implied to improve the detection rate of R wave. The detection rate of this algorithm has reached 97.7% for most data in TWA database and 98% for data in MIT-BIH database with a time location precision below 2.8ms.Derivative method was also used to detect positive T wave for higher computing efficiency. Extremums were marked on derivative ECG signal in order to locate the subinterval. The maximum in this subinterval was picked as the T-peak. Then an additional correction was designed to get a higher accuracy of T-peak determination. The statistic results of MIT-QT Database demonstrated that this algorithm had good performance for the T-peak detection while the standard deviation of time location error would be controlled below 10ms.Traditional power spectral density method was revised for TWA amplitude quantitative measurement. Fixed length window integration was used instead of single reference point mode to avoid time domain interference. Synthetic ECG was made from the real ECG signal and the simulated result showed that the TWA amplitude can be approximately calculated out with an absolute error between±5μV when the noise is low enough.STFT method was used to resolve the position of TWA. The results showed that this method could recognize TWA segments as short as 20 beats, and can also linearly give out the amplitude of TWA signal. According to the results of STFT, a new algorithm was developed to automatically recognize the TWA segments which can reduce the manual computational complexity.