Fig. 1.

Modeling of adolescent changes in morphometric similarity. We estimated morphometric similarity networks (MSNs) from multimodal MRI brain scanning data. (A) These images were parcellated into 358 predefined cortical regions. (B) Macrostructural MRI phenotypes, like CT, GM, and SA, were estimated for each cortical area overall. (C) Additionally, depth-dependent profiling was used to construct multiple cortical surfaces between the white matter surface and the pial surface for estimation of MT, a microstructural MRI phenotype, with fine-grained laminar resolution at 70% of cortical depth. (D) We constructed a feature matrix of multiple features estimated at each region for each individual subject, resulting in a subject-specific $\{Regions\times Features\}$ multimodal MRI data matrix. (E) We estimated the similarity between each pair of cortical areas in terms of the pairwise correlations between regional feature vectors comprising multiple normalized macro- and microstructural MRI features estimated at each region. (F) We compiled all possible interareal similarity measures in a subject-specific $\{Regions\times Regions\}$ association matrix or morphometric similarity network. (G) We estimated adolescence-related changes in weighted degree, i.e., the mean weight over all of each node’s edges: the baseline functional connectivity, as the predicted nodal degree at age 14, and the rate of change of hubness, as the slope of a linear regression of age on weighted degree.