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

Area of Science:

• Neuroimaging
• Machine Learning
• Neurology

Background:

• Parkinson's disease (PD) is a progressive neurodegenerative disorder characterized by motor and non-motor symptoms.
• Structural brain changes are associated with PD manifestations.
• Accurate diagnosis and understanding of PD pathophysiology remain critical.

Purpose of the Study:

• To differentiate Parkinson's disease patients from healthy controls using advanced neuroimaging and machine learning.
• To identify key brain imaging biomarkers associated with PD.
• To explore the correlation between identified biomarkers and motor/non-motor symptoms of PD.

Main Methods:

• Utilized multidimensional 7-Tesla structural Magnetic Resonance Imaging (MRI) features, including gray matter volume and cortical thickness.
• Applied Support Vector Machine (SVM) algorithm for classification between 98 PD patients and 74 healthy controls.
• Employed Partial Least Squares Regression (PLSR) to correlate MRI features with clinical symptom scales (MDS-UPDRS).

Main Results:

• The SVM model achieved 0.80 accuracy, 100% sensitivity, and an F1-score of 0.85 in distinguishing PD patients.
• Identified significant correlations between key MRI-derived biomarkers and motor symptoms (MDS-UPDRS-III, tremor, rigidity, bradykinesia, postural instability).
• Found significant associations between these biomarkers and non-motor symptoms (cognition, anxiety, depression, MDS-UPDRS-I).

Conclusions:

• 7-Tesla MRI integrated with machine learning shows high potential as a diagnostic tool for Parkinson's disease.
• The study provides insights into the pathophysiology of PD by linking structural brain changes to clinical symptoms.
• This approach may enhance the early detection and comprehensive understanding of PD's diverse manifestations.