Parkinson’s disease (PD) is a progressive movement disorder and the second most common neurodegenerative disease worldwide. It is highly heterogeneous—patients vary in symptoms, symptom severity, and progression rate—which makes predicting individual outcomes and tailoring treatments challenging. This study tackles this issue by using brain imaging (MRI) data to identify distinct subtypes of PD based on changes in brain atrophy over time.

Using a data-driven approach applied to a large, longitudinal patient cohort, we identified PD subtypes characterized by distinct longitudinal trajectories of brain atrophy, associated with specific patterns of progression in clinical symptoms. Notably, structural MRI allowed us to identify patients at risk for faster disease progression.

Additionally, we investigated brain age—an estimate of the brain’s biological age derived from MRI scans using machine learning trained on MRI of healthy individuals. We examined brain age across PD subtypes and its relationship to clinical symptoms and found that higher brain age was associated with greater severity of various non-motor symptoms.

Our findings contribute to the emerging field of data-driven PD subtyping, which aims to improve prediction of disease course and support the development of personalized treatment strategies.