Research On Feature Extraction And ST Segment Shape Recognition Method For ECG Signal

Myocardial ischemia is a common disease in the elderly population. However, with the improvement of people’s living standards, myocardial ischemia is rising and gradually becoming younger, and serious damages the health of people in our country. ECG as recoded by the body of cardiac electrical activity, is the most common and effective way for clinical diagnosis of myocardial ischemia. The QT-interval and ST-T change are the most common characteristics and the main diagnostic indicators for the detection of myocardial ischemia diseases. In order to accurately detect the QT-interval and ST-T change, we made a deep research on ECG feature point detection and analyzed the change of ST segment in this paper. The main research works are as follows:(1) ECG signal preprocessing:We used the combined method of the band pass filter with 0.5-45Hz and the Gaussian low pass filter for ECG signal preprocessing. This method was used to filter the ECG signal which selected from MIT-BIH database. The experimental results show that this method not only effectively filtered out the noise and drift, but also retained the features and shape information of the ECG signal.(2) The research of ECG signal feature point extraction. Firstly, the dyadic wavelet transform was used for ECG signal decomposition and reconstruction. Secondly, the Mallat algorithm was used to locate the P wave, R wave and T wave, then the morphology of the R wave and T wave can be distinguished by using the modulus maxima. Finally, we proposed an adaptive selection T-wave end point detection algorithm based on T-wave morphology according to comparative analysis of three kinds classical T wave end detection algorithm. This method was tested on QT-database (20 records of 3,569 beats each). The experimental results show that the adaptive algorithm not only ensured the real-time, but also showed a better detection performance than a single T wave end point detection algorithm.(3) The research of the ST-T change. ST-T change includes ST deviation and ST shape change. In this paper, we proposed to use the curvature scale-space technique to locate the start and end point of ST segment, and detect the level of ST segment. Then the corner point of the ST segment would be located by using the multi-scale method based on the CSS technology. Finally, the ST shape can be distinguished according to the corner point. This method was tested on the QT database and ST-T database, and the detection accuracy reached 96.54% and 86.66%, respectively. This experiment results demonstrated the effectiveness of this proposed method.In conclusion, a series of automatic ECG signal analysis method were proposed in this paper. It not only could accurately located ECG key feature points, but also accurately extracted the QT interval and classified the change of the ST segment. This research can provide theoretical basis for ECG clinical diagnosis.