ST Segment Change Classification Based on Multiple Feature Extraction Using ECG

ST deviation detection using electrocardiogram (ECG) is of great significance for ischemia heart disease diagnosis. In this paper, we proposed an algorithm based on multiple feature extraction to classify the ST deviation beat by beat. First, the ST segment was located. Then, morphological and Poincaré features of ST segment were extracted and combined with global feature. Finally, random forest was adopted to classify the ST segment change into normal, elevated or depressed. The algorithm was evaluated on the European ST-T Database and the average sensitivity of normal, depressed and elevated ST segment was 85.2%, 86.9% and 88.8% respectively. The result shows that the developed algorithm is helpful in automatically detecting the ST segment elevation and depression, showing more details of the ischemic syndrome.