Multichannel Complexity Index (MCI) for a multi-organ physiological complexity assessment

Quantitative measurements of multi-organ interplay are crucial for the assessment of multivariate physiological dynamics in health and disease. Nevertheless, current quantification of multivariate complexity for nonlinear physiological processes is limited by reliability issues on short-time series, and parameters sensitivity especially in case of a multiscale analysis. To overcome these limitations, we propose a new tool to characterize the complexity of interacting physiological processes that may have different temporal dynamics: the Multichannel Complexity Index (MCI). This metrics relies on a novel method for the reconstruction of the multivariate phase space, where each series is embedded using its proper time delay. MCI accounts for the estimation of phase space distances using fuzzy rules, and may be computed at two different ranges of time-scale values to investigate short- and long-term dynamics. We validated our algorithm using three-channel white gaussian noise and 1/f noise systems, with different levels of coupling. By applying our approach to these data, we demonstrate that the MCI method allows to discern not only the degree of complexity in the system dynamics, but also the across-channel coupling level. Results on synthetic series from the Henón map and Rössler attractor demonstrate that MCI effectively discerns between different dynamical behaviours, outperforming state of the art metrics such as the Refined Composite Multivariate Multiscale Fuzzy Entropy. On publicly-available physiological series, considering heartbeat dynamics and blood pressure variability, results demonstrate a MCI sensitivity to postural changes(p<10−2 for rest vs. slow-tilt, and p<0.05 for rest vs. rapid-tilt/stand-up conditions), as well as a MCI sensitivity to subjects’ age-range (data gathered while watching Fantasia Disney movie, 1940) with p<10−2 for short scales and p=0.03 for long scales. In conclusion, MCI is a viable tool for an effective multivariate physiological complexity assessment. The Matlab code implementing the proposed MCI algorithm is available online.