Identifing two distinct cortical progression subtypes of Parkinson's disease through multimodal neuroimaging.

PURPOSE: To investigate whether distinct patterns of cortical involvement exist in Parkinson's disease (PD) and to characterize their potential progression trajectories using multimodal neuroimaging and data-driven disease progression modeling. METHODS: In this cross-sectional multimodal imaging study, we enrolled 317 patients with clinically diagnosed PD and 61 healthy controls. All participants underwent simultaneous FDG-PET and MRI scanning. We applied the Subtype and Stage Inference (SuStaIn) model to cortical glucose metabolism and thickness data to identify latent disease progression patterns. Network-level characteristics were further examined within a whole-brain gradient framework. Robustness was assessed through age- and sex-balanced sensitivity analyses in the local cohort and external validation using harmonized T1-weighted MRI data from the Parkinson's Progression Markers Initiative (PPMI) dataset. RESULTS: Two distinct cortical involvement subtypes were identified. One subtype showed predominant alterations in higher-order association networks, including the default mode and frontoparietal networks, whereas the other was characterized by greater involvement of lower-order sensorimotor and limbic systems. These subtype patterns remained stable across sensitivity analyses and external validation. Disease duration showed a significant correlation with the inferred disease stage (r = 0.15, p = 0.01). Imaging findings further revealed hypermetabolism in brainstem and trans-entorhinal regions accompanied by widespread cortical hypometabolism. CONCLUSIONS: Our findings reveal two robust cortical progression patterns in PD, highlighting substantial heterogeneity in network-level metabolic and structural involvement. This framework provides new insights into PD phenotypic variability and may support future efforts toward disease stratification and personalized research.