Multi-policy models of interregional communication in the human connectome

Abstract

Network models of communication, e.g. shortest paths, diffusion, navigation, have become useful tools for studying structure-function relationships in the brain. These models generate estimates of communication efficiency between all pairs of brain regions, which can then be linked to the correlation structure of recorded activity, i.e. functional connectivity (FC). At present, however, communication models have a number of limitations, including difficulty adjudicating between models and the absence of a generic framework for modeling multiple interacting communication policies at the regional level. Here, we present a framework that allows us to incorporate multiple region-specific policies and fit them to empirical estimates of FC. Briefly, we show that many communication policies, including shortest paths and greedy navigation, can be modeled as biased random walks, enabling these policies to be incorporated into the same multi-policy communication model alongside unbiased processes, e.g. diffusion. We show that these multi-policy models outperform existing communication measures while yielding neurobiologically interpretable regional preferences. Further, we show that these models explain the majority of variance in time-varying patterns of FC. Collectively, our framework represents an advance in network-based communication models and establishes a strong link between these patterns and FC. Our findings open up many new avenues for future inquiries and present a flexible framework for modeling anatomically-constrained communication.

Comparing models of information transfer in the structural brain network and their relationship to functional connectivity: diffusion versus shortest path routing

“…Considering past success with the hybrid measures combining the diffusion and shortest path routing models of information transfer (Goñi et al 2014 ), this research will apply the exclusively diffusion-based measures of mean first passage time and communicability as well as the shortest path routing measure of shortest path length to structural connectivity, and the results will be used to predict functional connectivity to determine to what extent these measures are able to account for variance in functional connectivity. Crucially, this research will extend past research that has examined the ability of multiple graph theory communication measures to predict functional connectivity from structural connectivity (Betzel et al 2022 ; Vázquez-Rodríguez et al 2019 ; Zamani Esfahlani et al 2022 ) and benchmarking the ability for different communication measures to predict functional connectivity (Seguin et al 2018 , 2020 , 2022 ), by directly comparing two commonly used models (diffusion and shortest path routing) using multiple linear regression analyses, partial least squares regression, and principal components analysis to determine which graph theory model is most important in this relationship. Research suggests that brain networks (at both the macroscale and microscale) typically demonstrate a balance of diffusion efficiency and global efficiency (Goñi et al 2013 ), while also suggesting that this balance may lean more towards dominance of diffusion efficiency in human brains, in which case we expect that the diffusion measures examined here will be more relevant than shortest path length to the structure–function relationship in the brain.…”

Comparing Models of Information Transfer in the Structural Brain Network and Their Relationship to Functional Connectivity: Diffusion Versus Shortest Path Routing

“…Considering past success with the hybrid measures combining the diffusion and shortest path routing models of information transfer (Goñi et al, 2014), this research will apply the exclusively diffusion-based measures of mean first passage time and communicability as well as the shortest path routing measure of shortest path length to structural connectivity, and the results will be used to predict functional connectivity to determine to what extent these measures are able to account for variance in functional connectivity. Crucially, this research will extend past research that has examined the ability of multiple graph theory communication measures to predict functional connectivity from structural connectivity (Betzel et al, 2022; Vázquez-Rodríguez et al, 2019; Zamani Esfahlani et al, 2022) and benchmarking the ability for different communication measures to predict functional connectivity (Seguin et al, 2018, 2020, 2022), by directly comparing two commonly used models (diffusion and shortest path routing) using multiple linear regression analyses, partial least squares regression, and principal components analysis to determine which graph theory model is most important in this relationship. Research suggests that brain networks (at both the macroscale and microscale) typically demonstrate a balance of diffusion efficiency and global efficiency (Goñi et al, 2013), while also suggesting that this balance may lean more towards dominance of diffusion efficiency in human brains, in which case we expect that the diffusion measures examined here will be more relevant than shortest path length to the structure-function relationship in the brain.…”

Spatially-embedded recurrent neural networks reveal widespread links between structural and functional neuroscience findings

“…As such, we have seen considerations of space implemented in functional feedforward neural network models (Gozel & Doiron, 2022; Huang et al, 2019; Lee et al, 2020). Here we instantiated our core hypothesis mathematically within the seRNN model by providing two challenges to RNNs during supervised learning: (1) long connections should be minimized where possible – reflective of their metabolic cost (Kaiser & Hilgetag, 2006; Sporns, 2011), and (2) connections can only change their weights as a function of their underlying communication – reflective of signal propagation between neuronal units (Betzel et al, 2022; Seguin, Jedynak, et al, 2022; Seguin, Mansour L, et al, 2022; Shimono & Hatano, 2018). Both challenges are addressed at the local neuronal-level over the course of training which has the effect of continually shaping the networks global structural and functional properties over time.…”