Binary Brains: How Excitable Dynamics Simplify Neural Connectomes

Abstract

Fiber networks connecting different brain regions are the structural foundation of brain dynamics and function. Recent studies have provided detailed characterizations of neural connectomes with weighted connections. However, the topological analysis of weighted networks still has conceptual and practical challenges. Consequently, many investigations of neural networks are performed on binarized networks, and the functional impact of unweighted versus weighted networks is unclear. Here we show, for the widespread case of excitable dynamics, that the excitation patterns observed in weighted and unweighted networks are nearly identical, if an appropriate network threshold is selected. We generalize this observation to different excitable models, and formally predict the network threshold from the intrinsic model features. The network-binarizing capacity of excitable dynamics suggests that neural activity patterns may primarily depend on the strongest structural connections. Our findings have practical advantages in terms of the computational cost of representing and analyzing complex networks. There are also fundamental implications for the computational simulation of connectivity-based brain dynamics and the computational function of diverse other systems governed by excitable dynamics such as artificial neural networks.