Arrhythmia classification using the RR-interval duration signal

Abstract

In this paper we explore the RR-interval duration signal in classifying arrhythmias. Initially the RR-interval duration signal is extracted from the ECGs. Each RR interval is characterized using the MIT-BIH arrhythmia database annotation. We use a 3 RR interval sliding window, centered in the middle RR interval, to classify each beat using a rule-based approach. The rules are obtained from cardiologists and are related to the tachogram morphology in arrhythmic events. Classification is performed for five categories of arrhythmic beats. The classification system is tested using selected beats from the MIT-BIH database (30,000 annotated beats were used). We present results in terms of sensitivity, positive prediction and system's performance. The scores achieved indicate high performance (95.85%) for the classification of the different categories of arrhythmic beats.