Classification of obsessive-compulsive disorder from resting-state fMRI

Area of Science:

• Neuroscience
• Computational Psychiatry
• Medical Imaging

Background:

• Obsessive-compulsive disorder (OCD) significantly impacts patient quality of life.
• Accurate OCD diagnosis is crucial for effective treatment but is often lengthy and complex.
• Functional magnetic resonance imaging (fMRI) offers potential for identifying objective biomarkers.

Purpose of the Study:

• To investigate the utility of Pearson's correlation and network-based features derived from fMRI data for OCD prediction.
• To develop a diagnostic tool that assists clinicians in identifying individuals with OCD.

Main Methods:

• Extracted mean time series from 112 brain regions and decomposed them into 5 frequency bands.
• Calculated Pearson's correlation matrices and network-based features for each frequency band.
• Employed Minimum Redundancy Maximum Relevance (mRMR) for feature selection and leave-one-out cross-validation for performance evaluation.

Main Results:

• Achieved 80% accuracy, 81% sensitivity, and 77% specificity in classifying subjects as having OCD or being controls.
• Identified specific frequency bands (0.06-0.11 Hz for correlation, 0.03-0.06 Hz for network features) as most discriminative.
• Demonstrated the predictive power of network features and selected Pearson's correlation values.

Conclusions:

• fMRI-derived network features and correlation values show significant potential as biomarkers for OCD diagnosis.
• The proposed methodology offers a promising, data-driven approach to aid in the clinical assessment of OCD.
• Further research can refine these biomarkers for improved diagnostic accuracy and clinical utility.