Cardiac rhythm detection and classification by WOLA filterbank analysis of EGM signals

A method of cardiac rhythm detection based on time-frequency analysis provided by a weighted overlap-add (WOLA) oversampled filterbank is proposed. Cardiac rhythms obtained from intracardiac electrogram signals are decomposed into the time-frequency domain and analyzed by parallel peak detectors in selected frequency subbands. Based on the coherence (synchrony) of the subband peaks, an optimal peak sequence representing the beat locations is obtained. By analyzing the synchrony of the subband beats and the periodicity and regularity of the optimal beat, the electrogram segment is classified into various possible classes including fibrillation, flutter, and tachycardia. This multi-tier method is evaluated using Ann Arbor Electrogram library in clean and in additive noise. The presented results show that in clean and in 15 dB SNR noise, the method never misses fibrillation or flutter. Other misclassification errors were within the acceptable limits.