Discrimination of fibrillatory rhythms from intracardiac electrograms using real-time magnitude-squared coherence

We investigate the ability of moving discrete Fourier transform (MDFT) MSC to discriminate fibrillatory rhythms (Fib) from non-fibrillatory rhythms (Non-Fib). Four paroxysmal supraventricular tachycardia (PSVT), 6 atrial flutter (AFL), 9 atrial fibrillation (AF), 6 sinus rhythm (SR), 24 monomorphic ventricular tachycardia (MVT), and 25 ventricular fibrillation (VF) obtained from 42 patients were studied. An average mean MSC was computed over several time intervals for 3 frequency bands. Each rhythm was also analyzed by a moving rate algorithm with different combinations of threshold and blanking period, and a moving PDF algorithm. Mean MSC completely separates AF from non-AF in both 0-60 Hz and 20-50 Hz bands for 32, 64, and 128 point segment lengths. Mean MSC separates VF from non-VF more completely than rate and PDF.