Detection of QRS complexes in electrocardiogram using support vector machine

Abstract

This paper presents the application of a support vector machine (SVM) for the detection of QRS complexes in the electrocardiogram (ECG). The ECG signal is filtered using digital filtering techniques to remove noise and baseline wander. The support vector machine is used as a classifier to delineate QRS and non-QRS regions. Two different algorithms are presented for the detection of QRS complexes. The first uses a single-lead ECG at a time for the detection of QRS complexes, while the second uses 12-lead simultaneously recorded ECG. Both algorithms have been tested on the standard CSE ECG database. A detection rate of 99.3% is achieved when tested using a single-lead ECG. This improves to 99.75% for the simultaneously recorded 12-lead ECG signal. The percentage of false negative detection is 0.7% and the percentage of false positive detection is 12.4% in the single-lead QRS detection and it reduces to 0.26% and 1.61% respectively for QRS detection in simultaneously recorded 12-lead ECG signals. The performance of the algorithms depends strongly on the selection and the variety of the ECGs included in the training set, data representation and the mathematical basis of the classifier.