Classifying Obsessive-Compulsive Disorder from Resting-State EEG Using Convolutional Neural Networks: A Pilot Study.

Area of Science:

• Neuroscience
• Computational Psychiatry
• Machine Learning in Medicine

Background:

• Identifying obsessive-compulsive disorder (OCD) using brain data is challenging.
• Resting-state electroencephalography (EEG) is a noninvasive, affordable neuroimaging technique.
• Traditional machine learning methods have shown limited success in detecting predictive EEG signals for OCD.

Purpose of the Study:

• To explore the effectiveness of convolutional neural networks (CNNs) for classifying individuals with OCD.
• To compare CNN performance against traditional support vector machine (SVM) methods.
• To investigate if multimodal fusion of clinical and demographic data enhances classification accuracy.

Main Methods:

• Collected resting-state EEG data from 20 participants (10 with OCD, 10 healthy controls).
• Transformed 4-second EEG segments into time-frequency representations.
• Trained a 2D CNN and an SVM using a leave-one-subject-out cross-validation framework.

Main Results:

• The CNN achieved 85.0% accuracy and an AUC of 0.88 in subject-level classification.
• The SVM baseline performed at chance levels (45.0% accuracy, AUC: 0.47).
• Clinical and demographic data did not improve classification accuracy when added via multimodal fusion.

Conclusions:

• CNNs applied to resting-state EEG show significant potential for identifying OCD.
• Deep learning can uncover complex diagnostic patterns in neural data, outperforming traditional methods.
• Further research with larger, diverse samples is warranted to explore multimodal models for psychiatric classification.