Analysis of cardiovascular time series using multivariate sample entropy: A comparison between normal and congestive heart failure subjects

The cardiovascular (CV) system typically exhibits complex dynamical behavior, which is reflected not only within a single data channel, but more importantly across data channels. Multivariate sample entropy (MSE) has been proven as a useful tool to analyze both the within-and cross-channel coupled dynamics, providing an insight into the underlying system complexity and coupling relationship. In this study, the MSE method was used to monitor both the univariate and multivariate CV time series variability, focusing on identifying the differences between normal and congestive heart failure (CHF) subjects. Electrocardiogram, phonocardiogram and radial artery pressure waveforms were simultaneously recorded from 30 normal and 30 CHF subjects to determine three CV time series: RR interval, cardiac systolic time interval (STI) and pulse transit time (PTT). The MSE method was applied to univariate (RR, STI, PTT), bivariate (RR & STI, RR & PTT, STI & PTT) and trivariate (RR & STI & PTT) time series. The results showed that all MSE values in the CHF group were significantly lower than for the normal group (all P<;0.05, except for the univariate PTT series), which indicates that the complexity of univariate series decreased and the synchronization of multivariate series increased for CHF subjects. Moreover, the statistical significance between the two subject groups increased from using univariate to multivariate time series (with P<;0.05 to P<;0.001), confirming the advantage of multivariate analysis.