Estimation of the ECG signal spectrum during ventricular fibrillation using the fast Fourier transform and maximum entropy methods

The aim of this study was to compare two approaches for estimating the spectrum of the surface ECG during ventricular fibrillation (VF): the fast Fourier transform (FFT) and maximum entropy spectral analysis (MESA). The first 10 s of 10 recordings of clinical VF sampled at 250 Hz were separated into 1 s epochs for analysis by the FFT, zero padded FFT and Burg algorithm with 5, 10, 20 and 50 coefficients. The mean difference in dominant frequency between the FFT and Burg algorithm with 50 coefficients was 0.04 Hz (SD 0.56). The mean frequency of the dominant peak in the spectrum of VF can be measured accurately using either the FFT, zero padded FFT or MESA with a model order of more than 10. The overall accuracy of the standard FFT when compared to MESA is not limited by its poor resolution as long as results are averaged over large numbers of recordings