The Kalman filter based research of T-wave alternans detection algorithm

Abstract

As the second largest cause of death in human beings only behind cancer, sudden cardiac death (SCD - Sudden Cardiac Death) is featured as sudden onset and difficult to rescue. Therefore, for these patients, early diagnosis, early intervention is the most effective treatment. TWA (T wave alternans) as the detection of SCD-like illness is an important indicator. How to obtain accurate data in a TWA research is focused on in recent years. This paper presents a Kalman filter-based (non-steady state) TWA detection algorithm. Firstly, we preprocess, denose and baseline drift of ECG. Secondly, using wavelet modulus extremum to detect the positions of feature point, which belong to the QRS and T waves. Further to align T wave and extracts the T wave matrix. And re-group the T wave matrix according to the strategy (in accordance with the odd and even). Kalman filter is used on two groups of T wave matrix. After filtering, we calculate the difference matrix between the two matrix above, and get the absolute value of difference matrix. Finally, we use a serious of tactics, such as sorting average moving window to get the TWA value. According to simulation result, the correlation coefficient between the TWA detection values and real values reaches 0.97, and not only can it test the value of T wave alternans, but also can identify short-term T wave alternans.