This paper, presented by Ghasemi et al., examines the use of a scalp EEG-based ensemble of convolutional neural networks (CNN) to identify obsessive-compulsive disorder (OCD). The three advanced CNN models of interest were EEGNet, Shallow ConvNet, and Deep ConvNet as they have yet to be used in diagnosing OCD although they are trained on EEG datasets. The researchers used these models designed for scalp-EEG data analysis and optimized them using the Differential Evolution algorithm to improve the accuracy of diagnosis. To produce input images for the CNN models, an electroencephalogram (EEG) segmentation approach was used. 2880 images from 6 participants were used as the final test data. After the models were trained and optimized, the test data was used to evaluate their performance in metrics such as accuracy, sensitivity, specificity, F1-score, AUC, and confusion matrix. EEG patterns were used to distinguish between individuals with OCD and healthy individuals. Among the individual models, Shallow ConvNet outperformed the rest with an accuracy of 85.91 ± 0.72, sensitivity of 82.19 ± 0.72, specificity of 93.34 ± 2.91, F1-score of 88.61 ± 0.49, AUC of 87.77 ± 0.12, and PAM of 74.89 ± 1.36. Furthermore, the ensemble of the three scalped EEG-based CNNs performed extremely well with an accuracy of 87.03 ± 0.46, sensitivity of 82.21 ± 0.56, specificity of 96.69 ± 1.28, F1-score of 89.42 ± 0.37, and AUC of 89.45 ± 0.06. These findings show that EEG-based CNN models, particularly ensembles, may be able to improve the detection of OCD with high accuracy and specificity.