Coarse-graining spectral analysis: new method for studying heart rate variability

Abstract

Heart rate variability (HRV) spectra are typically analyzed for the components related to low- (less than 0.15 Hz) and high- (greater than 0.15 Hz) frequency variations. However, there are very-low-frequency components with periods up to hours in HRV signals, which might smear short-term spectra. We developed a method of spectral analysis suitable for selectively extracting very-low-frequency components, leaving intact the low- and high-frequency components of interest in HRV spectral analysis. Computer simulations showed that those low-frequency components were well characterized by fractional Brownian motions (FBMs). If the scale invariant, or self-similar, property of FBMs is considered a new time series (x′) was constructed by sampling only every other point (course graining) of the original time series (x). Evaluation of the cross-power spectra between these two (Sxx′) showed that the power of the FBM components was preserved, whereas that of the harmonic components vanished. Subtraction of magnitude of Sxx from the autopower spectra of the original sequence emphasized only the harmonic components. Application of this method to HRV spectral analyses indicated that it might enable one to observe more clearly the low- and high-frequency components characteristic of autonomic control of heart rate.