ST shape classification in ECG by constructing reference ST set.

Abnormal changes in the ST segment of an electrocardiogram (ECG) are very important diagnostic parameters for detecting myocardial ischemia. ST segment analysis requires a long-term ECG recording because of the transient change of the ST segment. Deviations of the ST segment are generally related to myocardial abnormality. In this study, we classify the ST segments by their morphology. First, a set of reference ST shapes is given. The ECG analysis algorithm developed in this study consists of feature point detection and ST shape classification. S wave and J-point detection are performed during the process of feature point detection, and the proposed algorithm classifies the STs into reference ST shapes. To improve the performance of ST shape classification, the rules for the trend of previous beats and the shape type of previous beats are used. The results from the proposed algorithm can provide information on the change in the ST shape. In our evaluation for classification of STs by their morphology using the test ECG data, the global correct rate was 83.14%. The best accuracy of existing ST level detection algorithms are 90% and over. However, considering that ST level detection algorithms cannot show the change of ST morphology; and that there are no studies about the classification of STs by their morphology using a reference ST set, the proposed algorithm is worthy of note.