Power spectral, cepstral, and Hartley analyses of intracardiac electrograms for detection of tachyarrhythmias

The authors have designed and evaluated three signal processing algorithms for use in automatic tachyarrhythmia detection that do not require detection of individual events in the intracardiac electrogram. The algorithms use Fourier, cepstral, and Hartley analyses of arbitrary segments of intracardiac signals. These three integral transforms were applied to 5.12-s passages of bipolar ventricular electrograms recorded during supraventricular tachycardia, ventricular tachycardia, and ventricular fibrillation. Each tachyarrhythmic rhythm was individually paired with an episode of sinus rhythm from the same patient. The running integral function of the normalized Fourier power spectrum separated sinus rhythm from supraventricular tachycardia in three out of five paired episodes and separated sinus rhythm from ventricular tachycardia and ventricular fibrillation in both individual patients and the group. However, it did not distinguish among the three tachyarrhythmias. Cepstral analysis revealed distinct peaks related to the cycle lengths of all tachyarrhythmic rhythms independent of waveform morphology and gain setting. Cepstral analysis was found to be superior in presenting the rate information. The Hartley analysis did not produce suitable discriminatory features for pattern recognition