MRI subtypes in Parkinson’s disease: Unraveling differences across diverse populations and clustering approaches

Abstract Parkinson’s disease (PD) is clinically heterogeneous, which suggests the existence of subtypes; however, there has been no consensus regarding their characteristics. This study included 633 PD individuals across distinct cohorts: unmedicated de novo PD, medicated PD, mild-moderate PD and a cohort based on diagnostic work-up in clinical practice. Additionally, 233 controls were included. Clustering based on cortical and subcortical grey matter measures was conducted with and without adjusting for global atrophy in the entire sample and validated within each cohort. Subtypes were characterized using baseline and longitudinal demographic and clinical data. Unadjusted results identified three clusters showing a gradient of neurodegeneration and symptom severity across both the entire sample and individual cohorts. When adjusting for global atrophy eight clusters were identified, lacking consistency across cohorts in validation. This study identified atrophy-based subtypes in PD, emphasizing the significant impact of global atrophy on subtype number, patterns, and interpretation in cross-sectional analyses.