Spatiotemporal and progression heterogeneity of Parkinson’s disease leveraging multimodal data

Parkinson’s disease (PD) is a clinically and biologically heterogeneous neurodegenerative disorder with divergent progression trajectories. Conventional longitudinal cohorts often lack the temporal resolution required for comprehensive disease progression modeling. This study leverages multimodal biomarkers to systematically identify and characterize interpretable PD subtypes with distinct temporal progression patterns. A cohort of 147 participants (107 PD, 40 healthy controls) underwent comprehensive multimodal assessments, including clinical evaluations, quantitative gait analysis, and neuroimaging. Biomarkers demonstrating monotonic associations with disease progression and significant group differences were selected, adjusted for covariates, and normalized. The subtype and stage inference (SuStaIn) model was applied to delineate progression subtypes, which were externally validated. Analysis revealed three reproducible PD subtypes: gait-predominant, intermediate motor, and cognitive-predominant, each defined by unique progression trajectories. Subtype classification exhibited high confidence and generalizability. The gait-predominant subtype showed early impairments in step length, gait speed, and turning, with late cognitive decline. The intermediate motor subtype featured early deterioration in axial and appendicular metrics alongside intermediate structural changes. The cognitive-predominant subtype displayed early cerebellar/brainstem involvement, rapid cognitive decline, and severe multimodal deficits, reflecting a malignant phenotype with the highest disease burden. These findings demonstrate biologically grounded divergence in PD progression, offering a scalable framework for individualized staging, mechanistic stratification, and precision-medicine trial design.