Unsupervised subtyping of motor dysfunction of Parkinson's disease and its structural brain imaging correlates

Parkinson's disease (PD) is a clinical neurodegenerative disorder. The Unified Parkinson's Disease Rating Scale (UPDRS) has been used as a standard measure of the PD symptom profile, and magnetic resonance (MR) imaging is widely used for identifying the critical brain regions involved in PD progression.

The present study aimed to (1) identify PD subtypes based on the motor dysfunction profile in the MDS-UPDRS and (2) find the differences in gray matter volumes of brain regions, and (3) compare non-motor features between the subtypes to explore their distinct clinical profiles.

In total, 299 patients with PD and 173 healthy participants from the Parkinson's Progression Markers Initiative were included. A software package, Generalized Association Plots, was used to cluster the motor dysfunction profile in the MDS-UPDRS. Regression models and the Artificial Intelligence Platform as a Service were used to quantify the differences in gray matter volume of brain regions between subtypes.

We identified three PD subtypes-resting tremor, intermediate, and akinetic-rigid-using motor symptom clustering. MRI analysis revealed significant differences in brain regions, including the posterior cingulate gyrus, lenticular nucleus, olfactory cortex, and cerebellum. Non-motor features, such as cognitive decline and autonomic dysfunctions, varied across subtypes, highlighting distinct systemic profiles. Akinetic-rigid patients exhibited the most severe impairments, while tremor-dominant patients showed milder non-motor symptoms.

Three PD subtypes of motor dysfunction were identified. Structural brain imaging revealed subtype-specific differences not only in cingulum and putamen regions, but also in the olfactory cortex, parahippocampal gyrus, and cerebellum, correlating with motor symptoms. Non-motor features varied by subtype, with increasing severity from tremor-dominant to akinetic-rigid.