Tracking ventricular arrhythmias with an artificial neural network

Presented are results of experiments using an artificial neural network (ANN) for tracking ventricular arrhythmias. Given two parameters (the current R-R interval and the probability of the current QRS being a ventricular premature beat (VPB)), the ANN determines the prevailing heart rate and the instantaneous rhythm (sinus, isolated VPB, couplet, ventricular tachycardia, bigeminy or trigeminy). The current R-R interval, together with the six previous R-R intervals and the prevailing heart rate, are used by the sub-network which tracks the prevailing heart rate. The current R-R interval, prevailing heart rate and current beat classification, together with the previous rhythm classification, are used by the sub-network which determines the rhythm classification. Using 236 beats from 19 independent subjects, the ANN classified the rhythm correctly 100% of the time, given the probability of correct classification of the QRS waveform in the range of 60% to 100%