Extended detrended cross-correlation analysis of physiological data

Abstract

Nonstationary dynamics of complex systems affects their reliable characterization from experimentally recorded datasets. In this regard, extensions of signal processing tools are often a required task to improve their diagnostic capabilities. Here, we address the problem of modifying cross-correlation analysis based on the DCCA-method. We suggest to estimate two scaling exponents, one of which is associated with DCCA and quantifies cross-correlations in interrelated time series, and the other exponent takes into account the effects of nonstationarity. Analysis of the dependence of local rms deviations of signal profiles from trend gives an opportunity to select a suitable range of scales where power-law statistics are observed, and to avoid combining ranges with distinct scaling properties. Such approach was used to characterize the entrainment phenomena in the dynamics of coupled Lorenz models that produce chaotic oscillations. In addition to quantifying the transition from asynchronous to synchronous chaos with increasing coupling strength, we analyzed the effects of nonstationarity on this characterization. Further, we applied this approach to study cross-correlations in the dynamics of adjacent cerebral vessels and their changes caused by abrupt growth in peripheral blood pressure. The results obtained confirm the conclusions for the simulated datasets and illustrate the advantages of the proposed DCCA extension in the numerical description of pharmacologically induced changes in CBF dynamics. We hope that this tool can be useful in various studies of interactions of physiological systems, where combined effects of dynamics and nonstationary behavior are observed, e.g., in network physiology.

Speaker

Alexander A. Koronovskii Jr.
Saratov State University
Russia

Discussion

Ask question