Non-stationary physiological noise in the cardiovascular system during sympatho-vagal changes.

This study introduces a novel estimation methodology for identifying non-stationary physiological noise, specifically applied to complex biomedical signals such as heart rate variability (HRV) series. By treating physiological noise as a dynamical recursive realization of independent and identically distributed (IID) Gaussian random variables, we employ an information-theoretic quantifier, the Approximate Entropy, to estimate noise power through a sliding window process. Our method effectively identifies noise levels in synthetic time series with varying dynamical noise powers, demonstrating accuracy even with relatively short window lengths. We further exploit this approach on real cardiovascular variability recordings during different postural changes, namely, stand-up, slow tilt, and fast tilt. The results reveal significant time-resolved variations in physiological noise, functionally linked with changes in autonomic regulation due to postural shifts. Specifically, in the absolute sense, physiological noise in the cardiovascular system tends to increase in the first 60 s of upright position with respect to a supine resting state, directly following sympathetic dynamics and inversely following vagal dynamics. Then, over 60 s physiological noise tends to decrease with respect to the resting state, almost monotonically. Moreover, results corroborate earlier findings where elevated stochasticity in HRV series biases complexity assessment through entropy analysis. Our work highlights the method's robustness and potential to improve the understanding of physiological noise dynamics, with implications for more accurate cardiovascular signal analysis and potential clinical applications.