Support Vector Machine Classification of Obsessive-Compulsive Disorder Based on Whole-Brain Volumetry and Diffusion...

Area of Science:

• Neuroimaging
• Computational Psychiatry
• Biomarker Discovery

Background:

• Magnetic resonance imaging (MRI) has identified brain differences between individuals with obsessive-compulsive disorder (OCD) and healthy controls (HC).
• Machine learning (ML) offers potential for individual-level OCD discrimination, but multi-modal approaches for biomarker identification are underexplored.

Purpose of the Study:

• To investigate the efficacy of a machine learning approach using multi-modal MRI data for individual-level OCD detection.
• To identify potential neuroimaging biomarkers for obsessive-compulsive disorder.

Main Methods:

• Acquired high-resolution structural MRI and diffusion tensor imaging (DTI) data from 48 OCD patients and 45 HC.
• Extracted gray matter volume (GMV), white matter volume (WMV), fractional anisotropy (FA), and mean diffusivity (MD) as features.
• Utilized support vector machine (SVM) for classification and identified key brain regions contributing to discrimination.

Main Results:

• SVM achieved high classification accuracies: 72.08% for GMV, 61.29% for WMV, 80.65% for FA, and 77.42% for MD.
• Discriminative regions included orbitofronto-striatal and dorsolateral prefronto-striatal circuits, cerebellum, uncinate fasciculus, cingulum, corticospinal tract, and cerebellar peduncle.
• Multi-modal analysis demonstrated significant accuracy in distinguishing OCD patients from healthy subjects at the individual level.

Conclusions:

• Support vector machine analysis of structural MRI and DTI data can accurately differentiate individuals with OCD from healthy controls.
• This computer-aided diagnostic method provides accurate information and offers a novel perspective for clinical OCD assessment.
• Identified brain regions and diffusion metrics serve as potential neuroimaging biomarkers for obsessive-compulsive disorder.