A pairwise maximum entropy model uncovers the white matter scaffold underlying emergent dynamics in intracranial EEG

Abstract

A major challenge in systems neuroscience is to understand how the brain's structural architecture gives rise to its complex functional dynamics. Here, we address this challenge by examining the inter-ictal activity of five patients with medically refractory epilepsy during ∼15 hours of multi-channel intracranial recording. By constructing a pairwise maximum entropy model (MEM) of the observed neural dynamics, we seek to uncover the fundamental relationship between functional activity and its underlying structural substrate. Despite only incorporating the pairwise correlations in the observed neural activity, we find that the pairwise MEM robustly fits large-scale patterns of inter-ictal power dynamics across a wide range of frequency bands, notably displaying time-invariance and cross-frequency similarity. Furthermore, across all frequency bands, we demonstrate that the pairwise MEM accurately identifies the structural white matter connections between brain regions, outperforming other common model-free measures of functional connectivity. Together, our findings show that a simple pairwise MEM, which is explicitly ignorant of higher-order correlations between three or more brain regions, not only captures complex spatiotemporal patterns of neural activity across the frequency spectrum, but also suggests that the network of structural connections in the human brain is a plausible scaffold capable of supporting observed wide-band neural dynamics.

Inferring excitation-inhibition dynamics using a maximum entropy model unifying brain structure and function

“…In this work, we first convert our BOLD time series to z-scores, ensuring that our BOLD date is represented as zero-mean with unitary variance, without altering the correlations between brain regions. As maximum entropy models of neural activity are developed based on Ising dynamics, studies investigating pairwise interactions using BOLD time course data are binarized to define activation states (either +1 for active, or -1 for inactive) in both simulated and empirical fMRI-based studies (Ashourvan et al, 2021;Cofré et al, 2019;Ezaki et al, 2017Ezaki et al, , 2020Gu et al, 2018;Nghiem et al, 2018;Niu et al, 2019;Watanabe et al, 2013). We will show how the binarization strategy may be validated using monte carlo simulations, whereby using the inferred interaction networks to reconstruct functional correlations.…”

“…Further, at the macro-scale, Ashourvan et al recently developed a maximum entropy-based framework that derives functional connectivity measures from intracranial EEG recordings; their findings suggest that structural connections in the brain give rise to large-scale patterns of functional connectivity by promoting co-activation between connected structures (Ashourvan et al, 2021). Thus, MEM may be an ideal tool to model functional connectivity and ultimately link micro-scale interactions (such as excitation and inhibition in neuronal circuits) to the functional connectome (FC) captured through fMRI BOLD activity.…”

Detection of functional connectivity in the brain during visuo‐guided grip force tracking tasks: A functional near‐infrared spectroscopy study

“…Pairwise MEM, which was originally used to measure the complexity of neural activity patterns, has been demonstrated to provide an estimation of FC more similar to structural connections and thus could more accurately estimate FC than Pearson's correlation method (Watanabe et al., 2013). It has been proven that a pairwise MEM can be well fitted to fMRI and EEG signals and provide accurate FC estimations during resting states (Ashourvan et al., 2018; Watanabe et al., 2013). In our study, during grip tracking tasks, the pairwise MEM provided richer information for task‐state FC estimation and showed an FC estimation more consistent with physiological evidence.…”