Spectral analysis of atrial fibrillation cycle lengths: comparison between fast Fourier transform analysis and autocorrelation function analysis using multipurpose physio-informatic analysis software.

Abstract:  BACKGROUND: Fast Fourier transform (FFT) analysis is a popular method of spectral analysis of atrial fibrillation cycle lengths (AFCL). Autocorrelation function (ACF) analysis is also available, so the aim of this study was to elucidate the relationship between FFT and ACF analyses in the spectral analysis of AFCLs. METHODS AND RESULTS: A total of 75 atrial fibrillation (AF) data from 39 patients were subjected to analysis. The dominant frequencies (DFs) from 4 different spectral resolutions of the FFT and peak AFCL from the ACF analysis were compared. In the FFT analysis using rectified signals, the DF was influenced by spectral resolution, no matter how the signals were tapered by the Hanning or Hamming window or filtered with the low-pass filter. There was a significant relationship between the DF from each spectral resolution and the peak AFCL. The DF from the 4,096-point FFT analysis had the strongest relationship to the peak AFCL with the smallest difference, when using 30-s AF data. In a study of the different lengths of the atrial fibrillation data, the DF also had a strong correlation to the peak AFCL with a small difference. CONCLUSIONS: The peak AFCL obtained from ACF analysis was not of the same quality as that from FFT analysis, but had the same value as the DF from FFT analysis.