Comparison of autoregressive and Fourier transform based techniques for estimating RR interval spectra

Two fundamentally different approaches have been used to estimate spectra of RR interval series. The classical approach uses a Fourier transform (FT), and the model based approach uses an autoregressive (AR) model. The aim of this study was to compare estimates of both low frequency (LF) and high frequency (HF) heart rate variability (HRV) obtained using both FT and AR techniques. Five minute RR interval series were obtained from a group of normals (N=9) and a group of heart transplant patients with decreased HRV (N=9). For the normal group, mean LF power was 1164 ms² (SD 817 ms²) with the FT and 1358 ms² (SD 906 m²s) with the AR method. Mean HF power was 1010 ms² (SD 1321 ms²) with the FT and 442 ms² (SD 718 ms²) with the AR method. In the transplant group LF HRV was unmeasureable, and mean HF power was 2.39 ms ² with the FT and 5.31 m²s with the AR method. Overall the two measures were well correlated (r=0.74). Both FT and AR spectra provide a comparable measure of LF and HF HRV