Classification of ST segment in ECG signals based on cross correlated supervised data

Abstract

This paper describes an automated selection of the ST segment in 12 leads electrocardiogram (ECG) as well as its classification based on cross correlation. Our proposed method classifies five categories of ST segment which are (a) Up slop (b) Down slop (c) Horizontal (Normal) (d) Concave (e) Convex using cross correlation process. We compare the main ECG (patient ECG) ST segment with the above-mentioned reference ST segments. In this work we have used MIT-BIH ST change database and European ST-T change database where every database contains minimum 30 min and maximum 1-h episode. Our method contains the following steps (1) Filtering ECG signal and Detrending it (2) R peak and S peak detection (3) Starting and ending point detection of ST segment (4) Comparing with ST segment supervised data (5) Classifying the ST segment. We have used total 1,34,879 beats where 58,331 beats from MIT-BIH ST change database and 74,609 beats from European ST-T change database. We have correctly selected total 126,608 ST segments. ST segment classification accuracy is 88.20% for MIT-BIH ST change database and 96.18% for European ST-T change database. The method confirms satisfactory performance with an overall accuracy of 92.1% which is helpful to the detection of major heart diseases like myocardial ischemia.