Multimodal brain age prediction reveals dissociable signatures of health, cognition and disease risk in 24,648 UK Biobank participants

Brain aging proceeds through multiple tissue compartments - grey matter, white matter, iron-rich subcortical nuclei and the cerebrovascular tree - yet most deep-learning brain-age models rely on a single MRI contrast, leaving compartment-specific contributions unresolved. Here we train 3D DenseNet121 models to predict chronological age from five MRI modalities - T1, T2 FLAIR, T1+T2 early fusion, diffusion MRI (dMRI) and susceptibility-weighted imaging (SWI) - in up to 24,648 UK Biobank participants, with external validation in the Parkinson's Progression Markers Initiative (PPMI). T1+T2 fusion achieved the lowest within-cohort error (mean absolute error 2.19 yr; Pearson $r=0.934$), yet downstream associations revealed striking modality specificity that accuracy alone did not predict. Diffusion MRI brain age gap (BAG) was uniquely associated with arterial stiffness and most strongly predicted incident type 2 diabetes (hazard ratio per s.d. 1.12, $P=2.4\times10^{-11}$); SWI BAG showed the largest cognitive effect sizes, particularly for reaction time and processing speed; T2 FLAIR BAG was the strongest predictor of incident all-cause dementia (HR 1.26) and cerebrovascular disease (HR 1.11); and dMRI BAG carried the highest hazard for Alzheimer's disease (HR 1.34). Grad-CAM attribution showed that T1+T2 fusion redistributes 7 percentage points from grey to white matter relative to T1 alone - a qualitative reshaping of the neuroanatomical basis rather than a linear combination of single-modality signals. Two-sample Mendelian randomization established causal protective effects on BAG for physical activity, oily fish, fresh fruit and coffee intake, and causal acceleration by smoking initiation and alcohol intake frequency. Together, these results establish brain age as a family of modality-specific biomarkers with dissociable phenotypic, prognostic and neuroanatomical profiles, and identify modifiable lifestyle targets for slowing structural brain aging.