Coarse-graining and criticality in the human connectome

TL;DR

This paper introduces a modular coarse-graining method for the human connectome, enabling the study of critical phenomena with reduced computational complexity, and demonstrates that qualitative behaviors are preserved in smaller networks.

Contribution

The authors develop a modularity-based coarse-graining approach for the human connectome, facilitating analysis of criticality in reduced networks while maintaining key qualitative behaviors.

Findings

Coarse-grained networks reflect original network behavior, enabling analysis of critical phenomena.

The transition point in Wilson-Cowan models correlates with the critical temperature in Ising models.

Qualitative insights are preserved in smaller, computationally manageable networks.

Abstract

In the face of the stupefying complexity of the human brain, network analysis is a most useful tool that allows one to greatly simplify the problem, typically by approximating the billions of neurons comprising the brain by means of a coarse-grained picture with a practicable number of nodes. But even such relatively small and coarse networks, such as the human connectome with its 100-1000 nodes, may present challenges for some computationally demanding analyses that are incapable of handling networks with more than a handful of nodes. With such applications in mind, we set out to further coarse-grain the human connectome by taking a modularity-based approach, the goal being to produce a network of a relatively small number of modules. We applied this approach to study critical phenomena in the brain; we formulated a hypothesis based on the coarse-grained networks in the context of criticality in the Wilson-Cowan and Ising models, and we verified the hypothesis, which connected a transition value of the former with the critical temperature of the latter, using the original networks. We found that the qualitative behavior of the coarse-grained networks reflected that of the original networks, albeit to a less pronounced extent. This, in principle, allows for the drawing of similar qualitative conclusions by analysing the smaller networks, which opens the door for studying the human connectome in contexts typically regarded as computationally intractable, such Integrated Information Theory and quantum models of the human brain.