Machine Learning Prediction of Rapid Parkinson Disease Progression Using Combined Imaging and Clinical Biomarkers

Area of Science:

• Neuroscience
• Medical Imaging
• Machine Learning

Background:

• Parkinson disease (PD) exhibits variable progression rates, necessitating accurate prediction for patient management.
• Identifying patients with rapid PD progression is crucial for effective clinical trial design and personalized therapeutic strategies.

Purpose of the Study:

• To evaluate machine learning (ML) models for predicting rapid PD progression.
• To integrate clinical and imaging biomarkers for enhanced predictive accuracy in Parkinson disease.

Main Methods:

• Retrospective analysis of 683 Parkinson's Progression Markers Initiative (PPMI) patients.
• Defined rapid PD progression (RPPD) using MDS-UPDRS scores and SPECT Putamen-Specific Binding ratio (SBR).
• Employed MRMR feature selection, SVM, and decision tree models with 5-fold cross-validation and ROC AUC analysis.

Main Results:

• The lowest putamen SBR was the most predictive feature for RPPD.
• SVM model achieved an ROC AUC of 0.86 using 6 features.
• Decision tree model achieved an ROC AUC of 0.81 using 2 clinical features (UPDRS score, SBR).

Conclusions:

• ML models, especially SVM, effectively predict rapid PD progression using multimodal data.
• Integrating SPECT imaging and clinical data improves disease trajectory characterization.
• Findings support personalized treatment strategies and optimized clinical trial design for Parkinson disease.