A comprehensive anatomical network analysis of human brain topology

Abstract

A network approach to the macroscopic anatomy of the human brain can be used to model physical interactions among regions in order to study their topological properties, as well as the topological properties of the overall system. Here, a comprehensive model of human brain topology is presented, based on traditional macroanatomical divisions of the whole brain, which includes its subcortical regions. The aim was to localise anatomical elements that are essential for the geometric balance of the brain, as to identify underlying phenotypic patterns of spatial arrangement and understand how these patterns may influence brain morphology in ontogeny and phylogeny. The model revealed that the parahippocampal gyrus, the anterior lobe of the cerebellum and the ventral portion of the midbrain are subjected to major topological constraints that are likely to limit or channel their morphological evolution. The present model suggests that the brain can be divided into a superior and an inferior morphological block, linked with extrinsic topological constraints imposed by the surrounding braincase. This information should be considered duly both in ontogenetic and phylogenetic studies of primate neuroanatomy.

Keywords: brain morphology; morphological integration; network theory; topological constraints.

FIGURE 1

Anatomical layout of the network simulating a dorsal view of the human brain (frontal lobe: red; central lobe: dark pink; parietal lobe: dark blue; temporal lobe: green; occipital lobe: light blue; insular lobe: brown; limbic lobe: orange; basal nuclei: light pink; diencephalon: yellow; cerebellum: light grey; cerebral trunk: dark grey). See Table 1 for labels.

FIGURE 2

Centrality metrics represented on an anatomical illustration of a lateral view of the human brain (a) and a medial view of the human brain (b) using a chromatic scale.

FIGURE 3

Regression models of the relations between degree and betweenness, and closeness and betweenness.

FIGURE 4

Jitter plot of Principal Component 1 based on the five centrality metrics (a). Scree plot (b). Loadings of the different centrality metrics in Principal Component 1 based on the correlation matrix (c).

FIGURE 5

Dendrogram of the cluster analysis based on degree, closeness, betweenness and eigenvector. Cophenetic correlation coefficient c = 0.82.

FIGURE 6

Distributions of degree and clustering coefficient (a) and corresponding regression models (b).

FIGURE 7

Topological modules represented on an anatomical illustration of a lateral view of the human brain (a) and a medial view of the human brain (b). Striped patterns indicate side discrepancies and hence overlap between distinct modules.