Fig. 2: Region-specific models accurately estimate cortical thinning in PD.

a The infographic shows the HC and sPD subgroups used to derive the reference model of PD-related brain alterations b Selection of carefully matched groups of participants. Only 104 out of 139 HC and 104 out of 293 sPD participants are carried forward to derive estimates of brain alterations associated with PD. c Propensity scores enable careful participant matching. Participants from both subgroups are matched based on age, sex, and site using a flexible propensity score matching strategy. The histograms display frequencies of propensity scores separately for sPD and HC subgroups. d Linear models incorporating diagnosis, age, sex, and site were used to predict cortical thickness and subcortical volume, with colored brain maps indicating regions where PD was associated with reduced (blue) or increased (red) structural measures. e Effect size of diagnosis across all brain regions. The heatmap shows diagnosis beta coefficients among 68 regions based on the Desikan-Killiany cortical parcellation and 14 regions defined by Harvard-Oxford subcortical atlas. Detected cortical thinning and volume loss aligns with established PD changes, underscoring our analytical protocol.