Session S71.3

Computer Algorithm for Tracking ECG Spectral Dynamics in Ventricular Tachyarrhythmias

P Langley*, A Murray

Newcastle University
Newcastle upon Tyne, UK

Spectral components of the ECG might provide information about the mechanisms underlying cardiac arrhythmias. We present a computer algorithm for tracking the temporal dynamics of the spectral components of the ECG of ventricular tachyarrhythmias (VT). The ECGs of 10 patients, 5 classified as monomorphic ventricular tachycardia (MVT) and 5 classified as ventricular fibrillation (VF), of durations ranging from 26 s to 110 s were analysed. The ECGs were recorded at a sample rate of 250 Hz and the power spectral density estimated by the periodogram using a 2048 point FFT (corresponding to an ECG time interval of approximately 8 s) and Bartlett-Hann window providing a frequency resolution of 0.12 Hz. Spectra were calculated across the recordings in steps of 10 sample points (corresponding to an ECG time step of 0.04 s). Spectral peaks were defined as discrete peaks above the noise threshold. An estimate of the spectral noise was obtained from the maximum power in the range 1 to 3 Hz which was outside the range of frequencies of VT. Dynamics of the spectra were measured by tracking changes in power and frequency of spectral peaks at each step. Peaks were considered to evolve continuously if the change in frequency of the peaks from interval to interval was less than or equal to 0.12 Hz (the spectral resolution) and the power was greater than the noise threshold. Using the algorithm we tracked the dynamics of spectral peaks across each recording and counted the number and duration of evolution of the peaks. MVT was characterised by a single peak lasting for the full duration of the recording in all cases. VF was characterised by spectra with many more peaks than MVT with on average 4.6 (across patients ranging from 3.8 to 5.7) peaks during any interval and with continuous evolution of the longest duration peaks of on average 21.9 (range 19.0 to 28.0) s. Using the algorithm we were able to quantify the number and duration of spectral peaks and compare these values for VF and MVT. The algorithm provides a simple non-invasive measurement and facilitates further research into whether the number of spectral peaks in the ECG correlates with the number of active sources underlying these arrhythmias and the mechanisms of human VF.

(Abstract Control Number: 40)