Investigating the Specific Spectral Features of Obsessive-Compulsive Disorder in Quantitative Electroencephalography

Obsessive-compulsive disorder (OCD) is a debilitating mental illness with symptoms typically manifesting during childhood and adolescence. Quantitative electroencephalography (QEEG) is a favored brain mapping technique for investigating psychiatric disorders due to its accessibility and ease of use. Spectral features of QEEG, particularly differences in neural activity across frequency bands and brain regions, may underlie certain diseases. This study investigates the performance of a machine learning algorithm using spectral features as input to provide insight into the neural activities associated with OCD.

This analytical cross-sectional study obtained EEG signals from healthy controls and OCD patients in a closed-eyes condition using a 21-channel cap. After noise removal via independent component analysis (ICA), features were calculated, augmented by 1-minute EEG segmentation, and used as input for the machine learning algorithm. Statistical tests were used to compare features between groups.

The study included 42 participants, 27 OCD patients, and 14 healthy controls. All features followed a non-normal distribution. The Mann-Whitney test revealed no significant differences between groups for all features (p>0.05) except for frontal beta (p<0.002). Using 10-fold cross-validation, the machine learning algorithm achieved an accuracy of 82.1%, with a sensitivity of 83.3% and specificity of 80.0%. The false discovery rate (FDR) was 11.8%.

In the studied population, there were no significant differences between OCD patients and healthy controls for any of the investigated features except for the frontal beta feature. However, the machine learning algorithm accurately detected 82.1% of patients, which was comparable to the performance of other features and imaging methods reported in previous studies.