Overview: This study, by Hao et al., evaluated the efficiency of a diagnostic model for schizophrenia (SCZ) by assessing electroencephalogram (EEG) signals utilizing deep learning (DL) techniques such as EEGNet, a Convolutional Neural Network (CNN). γ waves were selected as clinical studies indicated that disruptions in γ activity serve as a neurophysiological marker in SCZ, and are also considered informative in SCZ research. A 64-channel NeuroScan EEG cap was used at a sampling rate of 1000 Hz, and electrodes were placed using the 10-20 system on 14 SCZ patients and 14 healthy controls (HCs) before preprocessing was done to ensure consistency across participants. Preprocessed EEG signals were divided into 2s epochs, and a 1024 Fast Fourier Transform (FFT) was applied to each epoch using an EEGNet-based model to classify individuals with SCZ. The network comprised three convolution layers and classification, spanning from learning spatial patterns across channels, to differentiate between healthy and SCZ patterns. The dataset implemented the Leave-One-Subject-Out Cross-Validation (LOSOCV) method to divide the training and test set, where 30% of the training set was separated as a validation set. The average recognition accuracy for the SCZ group was 98.19% and 97.24% for the HC group when using γ frequency band features, showing that the model has good generalization and high recognition accuracy. When comparing this model against other, more traditional models, it was found that this model outperformed the others. This study puts forth a method for SCZ diagnosis using EEG γ bands.