A scale-invariant perturbative approach to study information communication in dynamic brain networks

How communication among neuronal ensembles shapes functional brain dynamics at the large scale is a question of fundamental importance to Neuroscience. To date, researchers have primarily relied on two alternative ways to address this issue 1) in-silico neurodynamical modelling of functional brain dynamics by choosing biophysically inspired non-linear systems, interacting via a connection topology driven by empirical data; and 2) identifying topological measures to quantify network structure and studying them in tandem with functional metrics of interest, e.g. co-variation of time series in brain regions from fast (EEG/ MEG) and slow (fMRI) timescales. While the modelling approaches are limited in scope to only scales of the nervous system for which dynamical models are well defined, the latter approach does not take into account how the network architecture and intrinsic regional node dynamics contribute together to inter-regional communication in the brain. Thus, developing a generalized scale-invariant measure of interaction between network topology and constituent regional dynamics can potentially resolve how transmission of perturbations in brain networks alter function e.g. by neuropathologies, or the intervention strategies designed to mitigate them. In this work, we introduce a recently developed theoretical perturbative framework in network science into a neuroscientific framework, to conceptualize the interaction of regional dynamics and network architecture in a quantifiable manner. This framework further provides insights into the information communication contributions of putative regions and sub-networks in the brain, irrespective of the observational scale of the phenomenon (firing rates to BOLD fMRI time series). The proposed approach can directly quantify network-dynamical interactions without reliance on a specific class of models or response characteristics: linear/nonlinear. By simply gauging the asymmetries in responses to perturbations, we obtain insights into the significance of regions in communication and their influence over the rest of the network. Moreover, coupling perturbations with functional lesions can also answer which regions contribute the most to information spread: a quantity termed Flow. The simplicity of the proposed technique allows translation to an experimental setting where the response asymmetries and flow can inversely act as a window into the dynamics of regions. For proof-of-concept, we apply the perturbative approach on in-silico data generated for human resting state network dynamics, using different established dynamical models that mimic empirical observations. We also apply the perturbation approach at the level of large scale Resting State Networks (RSNs) to gauge the range of network-dynamical interactions in mediating information flow across brain regions.