Adaptive classification of myocardial electrogram waveforms.

Abstract

The shape of myocardial electrogram complexes can change gradually in response to electrical and physiological transients. These changes could affect the reliability of morphologic-based electrogram classifiers proposed for use in implantable cardioverters. In this report, we present a method of detecting gradual changes in the shape of electrogram complexes and evaluate the method by incorporating it into a simple adaptive classification scheme. Of the six subjects recruited to take part in a previous comparative study of myocardial electrogram features, we observed extensive morphologic drift of normal sinus beats in two subjects. Our results indicate that the adaptive classification scheme proposed here can reduce observed classification error rates compared to rates obtained without adaptation.