Modularity and community detection in human brain morphology.

Abstract

Humans possess morphologically complex brains, which are spatially constrained by their many intrinsic and extrinsic physical interactions. Anatomical network analysis can be used to study these constraints and their implications. Modularity is a key issue in this framework, namely, the presence of groups of elements that undergo morphological evolution in a concerted way. An array of community detection algorithms was tested on a previously designed anatomical network model of the human brain in order to provide a detailed assessment of modularity in this context. The algorithms that provide the highest quality partitions also reveal general phenotypic patterns underlying the topology of human brain morphology. Taken together, the community detection algorithms highlight the simultaneous presence of a longitudinal and a vertical modular partition of the brain's topology, the combination of which matches the organization of the enveloping braincase. Specifically, the longitudinal organization is in line with the different morphogenetic environments of the three endocranial fossae, while the vertical arrangement corresponds to the distinct developmental processes associated with the cranial base and vault, respectively. The results are robust and have the potential to be compared with equivalent network models of other species. Furthermore, they suggest a degree of concerted topological reciprocity in the spatial organization of brain and skull elements, and posit questions about the extent to which geometrical constraints of the cranial base and the modular partition of the corresponding brain regions may channel both evolutionary and developmental trajectories.