On the physiological validity and the effects of detrending in the multiscale entropy analysis of heart rate variability

Quantifying the complexity of a physiological time series has been of considerable interest. The recently developed multiscale entropy (MSE) analysis is aimed at evaluating the complexity of a physiological time series. MSE robustly separates healthy and pathologic groups and consistently yields higher values for simulated long-range correlated noise compared to uncorrelated noise. In this study, we found that strongly correlated noise, such as fractional Brownian motion (fBm), has a higher complexity than uncorrelated noise, such as Gaussian noise, for sufficiently large time scales, and that pathologic dynamics like congestive heart failure is associated with a decreased fluctuation due to the strong correlation property of fBm. The multiscale entropy assigned to a coarse-grained time series takes into account both its fluctuation and its correlation properties. Short- and long-term fluctuations and correlations in time series could represent an artifact of non-stationarity related to linear or higher-order trends. We also examined how removal of local or global trends from a signal of interbeat intervals can have some effects on the changes in MSE curves at different time scales. These findings raise the possibility that understanding the origin of such complexity and its alterations with disease may have potential for clinical monitoring.