Multi-view ensemble learning for EEG-based detection of Obsessive-Compulsive Disorder.

Area of Science:

• Neuroscience
• Computational Psychiatry
• Machine Learning

Background:

• Psychiatric disorders, including Obsessive-Compulsive Disorder (OCD), present diagnostic challenges due to subjective methods and complex symptoms.
• Accurate and objective diagnosis of OCD is crucial for effective treatment but remains difficult.

Purpose of the Study:

• To develop a novel frequency-aware multi-view ensemble learning framework for electroencephalography (EEG)-based OCD classification.
• To enhance the diagnostic precision and objectivity of OCD recognition.

Main Methods:

• EEG features were partitioned into frequency-specific views (delta, theta, alpha, beta, gamma).
• Multiple base classifiers were trained for each frequency view.
• Particle Swarm Optimization (PSO) was used to optimize classifier weights for integrating predictions.

Main Results:

• The proposed framework achieved high performance: 90.1% accuracy, 90.6% precision, 90.1% recall, and 89.7% F1-score.
• Significantly outperformed traditional methods like Support Vector Machines (SVM) and Random Forest.
• Demonstrated superior diagnostic capabilities compared to state-of-the-art baselines.

Conclusions:

• The frequency-aware multi-view ensemble learning framework shows significant clinical potential for OCD diagnosis.
• This non-invasive, scalable computational tool can complement existing psychiatric assessments.
• The study highlights the utility of advanced machine learning techniques in improving psychiatric diagnostic accuracy.