Lesion-Level Subtypes of White Matter Hyperintensity Evolution Beyond Spatial Location

White matter hyperintensities (WMHs) are common neuroimaging markers of cerebrovascular pathology in aging and neurodegeneration. Despite their clinical relevance, WMH are typically quantified using global burden measures that assume a relatively homogeneous pathologic process. However, growing evidence suggests substantial biological heterogeneity across lesions. We aimed to identify lesion-level WMH subtypes beyond anatomic location and evaluate their associations with neurodegeneration and vascular risk.

We conducted a longitudinal observational study analyzing 3224 MRI scans from 403 participants spanning cognitively normal aging, mild cognitive impairment, Alzheimer, and Parkinson disease. Imaging at baseline and 2-year follow-up included structural, diffusion, and resting-state MRI. A total of 2107 WMH lesions were identified, and lesion-wise longitudinal changes were used to derive subtypes using unsupervised clustering. Associations with neurodegeneration and vascular risk factors were assessed using multivariable models with false discovery rate correction.

Three lesion subtypes (L₁-L₃) were identified, frequently coexisting within the same individual. L₁ lesions were the most prevalent (48.1%), predominated in cognitively normal individuals, and exhibited relatively stable trajectories without association with brain atrophy. L₂ lesions represented a less frequent (11.3%) unstable subtype associated with weight gain (odds ratio [OR] 1.33, 95% CI 1.22-1.45; p_FDR ≤ 0.001), suggesting metabolic vulnerability. L₃ lesions (40.6%) represented an unstable subtype associated with brain atrophy (β = -0.11, 95% CI -0.16 to -0.05; p_FDR < 0.001), older age (OR 1.15, 95% CI 1.07-1.23; p_FDR < 0.001), and vascular risk reflected by pulse pressure changes (OR 1.09, 95% CI 1.03-1.15; p_FDR = 0.006). Global WMH burden was no longer associated with brain atrophy after accounting for L₃ lesion burden. Clustering robustness was supported by sensitivity analyses excluding anatomical location and by external validation in an independent cohort reproducing the main atrophy-related findings.

WMH are not a homogeneous entity but comprise biologically distinct lesion subtypes with differential neurobiological and clinical significance. Lesion composition may therefore offer a more informative framework than global WMH burden for understanding cerebrovascular contributions to aging and neurodegeneration, with potential implications for risk stratification, clinical interpretation, and targeted interventions.