A shape deformation-based framework for constructing individualized subcortical morphological networks.

Individualized morphological brain networks have emerged as an important framework for imaging-based connectomics. However, existing methods have focused predominantly on the cerebral cortex, leaving subcortical structures largely unexplored despite their central roles in cognition and neuropsychiatric disorders. Here, we developed a novel framework for constructing individualized subcortical morphological networks from structural magnetic resonance imaging. Specifically, subcortical surfaces were reconstructed using a Point2Mesh-optimized multilayer perceptron architecture and vertex-wise shape deformation was quantified as the signed Euclidean distance between each vertex on an individual mesh and the corresponding vertex on the group-average template mesh. Interregional morphological connectivity in local shape deformation distributions was subsequently estimated using Jensen-Shannon Divergence. Applying this approach to the Human Connectome Project dataset, we found that subcortical morphological networks exhibited complex topological organization, including small-world architecture, modular organization, and highly connected hubs, which were consistently preserved at finer subregional scales. Analyses using an independent test-retest dataset further demonstrated fair-to-good reliability of these network properties. Importantly, these shape deformation-based networks showed significant coupling with subcortical functional networks estimated from resting-state functional magnetic resonance imaging. Leveraging large-scale multisite clinical datasets, we further demonstrated alterations of subcortical morphological networks in schizophrenia and major depressive disorder. Schizophrenia was characterized predominantly by right caudate-centered disruptions in morphological connectivity and nodal centrality, whereas major depressive disorder exhibited alterations only in nodal centrality of the right amygdala. Together, these findings establish a novel and reliable framework for investigating subcortical morphological network organization in the human brain across both healthy and psychiatric populations.