Rule-Based Method for Morphological Classification of ST Segment in ECG Signals

Abstract

Abnormal change in ST morphology is an important indicator for heart disease, especially for myocardial ischemia. The automatic classification of ST morphology can provide valuable information for helping physicians in the diagnosis of myocardial ischemia, especially in long-term and remote electrocardiogram (ECG) monitoring environments. In order to provide more accurate and efficient ST morphology classification for long-term and remote ECG monitoring applications, a simple rule-based ST morphology classification method is proposed that identifies ST segments with the normal morphology type and five abnormal morphology sub-types: concave or convex elevation, up-sloping, down-sloping, or horizontal depression. The proposed method consists of the following steps: (1) 0.05–45 Hz band-pass filtering; (2) signal quality assessment; (3) R peak detection; (4) removal of ECG beats with anomalous RR interval; (5) identification of start and end points of ST segment; (6) determination of sliding baseline; and (7) rule-based morphological classification of each ST segment. A total of 17,314 ECG beats were selected from the European ST–T ECG database, with 11,461 beats used for training and 5853 beats used for testing. Four quantitative parameters, namely sensitivity, positive predictive rate, F-measure, and G-mean, were used as classification indices. The final classification accuracy for the training and testing data are 91.8 and 90.1 % respectively, indicating the clinical significance of the proposed method. This method is clinically and pathophysiologically important, and is an important contribution to the early detection of the potential risk of myocardial ischemia in heart disease patients in long-term and remote ECG monitoring.