Abstract
White matter hyperintensity, a key imaging biomarker for brain health, has prognostic implications for stroke. Using a multicenter MRI dataset of 9179 stroke patients plus the UK Biobank (n = 36,210 low/high risk controls), we employ Subtype and Stage Inference modeling and identify three distinct white matter hyperintensity progression subtypes: fronto-parietal, radial, and temporo-occipital. Longitudinal validation confirms classification stability. The fronto-parietal subtype shows delayed onset and more hypertension, while the temporo-occipital subtype has more atrial fibrillation and coronary heart disease. The fronto-parietal and radial subtypes are linked to small vessel stroke, while the temporo-occipital subtype is linked to cardioembolism. The fronto-parietal subtype has higher 1-year ischemic stroke recurrence, while the temporo-occipital subtype shows a higher incidence of early neurological deterioration by symptomatic hemorrhagic transformation and worse 3-month outcomes. Beyond capturing progression, demographics, and vascular risks, and improving post-stroke outcome prediction, this subtyping-staging model also holds potential for stroke prediction.