Ventricular Ectopic Beat Classification Using KNN Multi-Feature Classifier

Abstract

The detection of ventricular ectopic beats (VEB) in Electrocardiogram (ECG) plays a vital role in the diagnosis of cardiovascular diseases like ventricular tachycardia and ventricular fibrillation, which causes sudden death. This paper discusses the beat interval based features for the automated detection of ventricular ectopic beats. Five-time domain features, pre RR interval, post RR interval, QRS width, QR slope and RS slope are considered and a KNN classifier is employed to classify normal and ventricular ectopic beats. The proposed study make use of ECG signals derived from MIT-BIH arrhythmia database. The quantitative results are compared with qualitative assessment recorded already in the database. The overall classification accuracy achieved is 98.67% and the accuracy for normal & ventricular ectopic beats is 98.81% & 95.12% respectively with the Sensitivity of 95.12%, Specificity of 99.11% and Positive Predictivity of 92.85%.