Spatial entropy of brain network landscapes: a novel method to assess spatial disorder in brain networks

Abstract

In this work, we introduce a method for mapping the spatial entropy of functional brain network community structure images in brain space. Entropy maps indicate the extent to which the network communities present in a local area are ordered or disordered. We demonstrate how spatial entropy can be quantified for each voxel in the brain according to the network community affiliations of surrounding voxels. This process results in interpretable maps of brain network entropy. We show that local entropy decreases in predictable brain regions during working memory and music-listening tasks. We suggest that these regional entropy reductions reflect self-organization of neural processes in support of functionally localized cognitive tasks. Analyses in this work provide a framework for future analyses of spatial entropy in complex networks that can be mapped to Euclidean space - both within the brain and in other contexts. Significance StatementWe introduce an approach for quantifying the spatial entropy of functional brain network community structure. We demonstrate the biological relevance of the measure in three independent datasets. This approach for analyzing brain network data is data-driven, easy to implement, and highly interpretable. It also allows investigators to visualize complex data by mapping values into the brain rather than storing values in extremely high-dimensional and abstract data structures. We believe this will make the method highly accessible even to investigators with minimal experience analyzing human neuroimaging data.