A detection algorithm for multiform premature ventricular contractions

This paper reports an algorithm developed to identify and quantify multiform PVCs. The algorithm clusters PVCs of similar morphology using a combination of time-domain and frequency-domain analysis. Initially, PVCs are grouped together on the basis of four time-domain-based morphological feature measurements. However, these time-domain-based clusters many times are nonunique because commonly encountered signal changes can cause substantial variations in the feature measurements of clinically similar beats. These redundant clusters are consolidates using two frequency-domain parameters: The First Spectral Moment (FSM) (center of gravity) of the amplitude spectrum, and the 5-Hz phase angle.