Support vector machine-driven Parkinson's disease identification: a 7-Tesla multidimensional structural MRI approach

Parkinson's disease (PD) is a progressive neurodegenerative disorder diagnosed clinically by cardinal motor symptoms, with structural brain changes associated with its diverse motor and non-motor manifestations. This study integrated multidimensional 7-Tesla structural Magnetic Resonance Imaging (MRI) features (gray matter volume, cortical thickness, etc.) using Support Vector Machine (SVM) to distinguish 98 PD patients from 74 healthy controls. The SVM model achieved 0.80 accuracy (sensitivity: 100%, F1-score: 0.85) and identified key biomarkers. Partial Least Squares Regression (PLSR) revealed these features correlated significantly with motor symptoms (Movement Disorder Society-Unified Parkinson's Disease Rating Scale [MDS-UPDRS]-III, tremor, rigidity, bradykinesia, postural instability; P < 0.05) and non-motor symptoms (cognition, anxiety, depression, MDS-UPDRS-I; P < 0.05). The findings highlight the potential of 7-Tesla MRI and machine learning as diagnostic tools for PD, while also providing insights into its pathophysiology. This approach may aid in detection and understanding of PD's motor and non-motor manifestations.