Real-time Heartbeat Classification Based on Parallel QRS Detections

The main challenge for automatic classification of cardiac arrhythmias with battery-operated devices is to achieve good classification results without high computational load and related high energy consumption.In this paper, we present a heartbeat classification method that uses the differences in R-peak time locations detected by multiple low power QRS detection algorithms operating in parallel.The outputs of the detectors are the inputs to the decision tree classifier.The classification is divided into three QRS morphology types: ( 1) Normal and Atrial Premature, (2) Ventricular, and (3) Other.The overall accuracy (ACC) of the proposed classification method for test dataset is 90.52%.The detailed results are as follows: Positive Predictivity (PPV) 93.82%, Sensitivity (Se) 95.51% for Normal and Atrial Premature class; PPV=84.66%,Se=76.33% for Ventricular class; PPV=84.33%,Se=79.29% for Other class.The proposed arrhythmia classification method is applicable for realtime mobile ischemia detection and HRV analysis.