Divergent Multimodal Age-Association Profiles Across the Human Brain.

Chronological age accounts for much of the variance in structural brain measures, yet aging is spatially distributed, modality-dependent and often nonlinear, features that conventional T1-only brain-age models overlook. We introduce a population-calibrated, multimodal framework in 2,796 adults that integrates gray-matter volume with diffusion-derived metrics to fit linear and quadratic age-association profiles across cortical and subcortical regions. We define the Age of Structural Transition (AST) as the vertex of the preferred trajectory, marking maximal macrostructural or microstructural refinement before age-related reduction. AST maps show early peaks in sensory-motor cortex and later peaks in association networks. Trajectory-based clustering uncovers distributed co-aging networks that group distant regions by shared multimodal trends rather than anatomy. A regionally stacked predictor improves age estimation (mean absolute error 3.45 years) and yields interpretable spatial importance maps. Sliding-window analyses show that brain-age-gap associations with phenotypes are age-dependent, weak in younger adults and amplified later in life.