Nonlinear Neural Patterns Are Revealed In High Frequency fNIRS Analysis

Abstract

AbstractVasomotor tone has a direct implication in oxygen transport to neural tissue, and its dynamics are known to be under constant control from feedback loops with visceral signals, such as sympathovagal interactions. Functional Near Infrared Spectroscopy (fNIRS) offers a nuanced measure of hemoglobin concentration that also comprises high frequencies, though most fNIRS literature studies traditional frequency ranges of hemodynamics (< 0.2 Hz). Linear theory of the hemodynamic response function supports this low frequency band, but we hypothesize that nonlinear effects elicited from the complex system sustaining vasomotor tone presents itself in higher frequencies. To characterize these effects, we investigate how plausible modulation of autoregulatory effects impact aforementioned high frequency components of fNIRS through simulations of mechanistic hemodynamic models. Then, we compare representational similarities between fast (0.2 Hz to 0.6 Hz) and slow (< 0.2 Hz) wave fNIRS to demonstrate that representations acquired through nonlinear analysis are distinct between the frequency bands, whereas when using linear time-domain analysis they are not. Furthermore, by comparing topoplots of significant detectors using nonlinear random vector correlation methods (distance correlation), we demonstrate through a 2nd level group analysis that the median concentrations acquired by fNIRS are independent when analyzing the nonlinearity of their dynamics in their fast and slow component, while they are dependent when utilizing linear time-domain analysis. This study not only provides motivation for researchers to also include higher frequency components in their analysis, but also provides motivation to explore nonlinear effects, e.g. topological entropy. The results of this study motivate future research to explore the nonlinear autoregulatory impacts of regional blood flow and hemoglobin concentrations.Author summaryConventionally, hemodynamic response from induced neural metabolic demand is studied as a slow signal, i.e < 0.2 Hz. Though this may be justified in linear analysis of hemodynamics, vascular mechanics nonlinearly transform the neural metabolic demand to hemodynamic response, where a nonlinear spectral profile may show higher frequency responses. Higher frequency ranges may give insight into local vascular dynamics, particularly their reflection of autoregulatory phenomena, hypothesized to be controlled by sympathovagal feedback loops, thus opening a new avenue for studying brain-body interactions. Functional near infrared spectroscopy (fNIRS) offers a method with high temporal resolution (10 Hz) for observing these effects in hemoglobin concentrations. In this study, we utilize stochastic dynamical simulations of plausible autoregulatory phenomena and an open fNIRS dataset to study differences of fast and slow wave neurovascular representations. We demonstrate that, while linear time-domain analysis provides similar representations of fast and slow wave activity, representations derived from nonlinear methods are not. Furthermore, we show how stress tasks, which may elicit autonomic activity, further desynchronizes nonlinear activity between fast and slow wave signals compared to a non-stress inducing task, demonstrating unique high frequency neurovascular phenomena that is mediated by stress processing.