A Novel Depression Recognition Network Based on Brain Functional Connectivity from Prefrontal Emotional State EEG

Objective: Electroencephalogram (EEG) signals provide a noninvasive and objective method for detecting neural dysfunctions associated with depression. To address the limitations of limited sample sizes and inadequate modeling of spatiotemporal features, the study aims to develop an EEG-based depression recognition framework to capture dynamic patterns of neural activity related to depression. Methods: We propose DR-Net, a novel depression recognition framework integrating generative adversarial network (GAN), transition propagation graph convolution network (TPGCN), and transformer architectures. First, GAN-based data augmentation generates synthetic EEG samples that preserve the statistical properties of real data, improving data diversity and model generalization. Second, brain functional connectivity networks are constructed using the phase lag index, and the TPGCN models the dynamic propagation of information across channels, capturing the spatiotemporal evolution of brain network topologies beyond conventional static graph models. Concurrently, the transformer module employs self-attention to strengthen the modeling of long-range temporal dependencies. The framework also emphasizes the frontal lobe brain region, reducing input dimensionality while improving feature discriminability. Results: The proposed method achieved an accuracy of 98.7%, surpassing baseline and state-of-the-art methods, with no significant class bias. Validation on clinical datasets yielded an accuracy of 96.13%, demonstrating strong robustness and generalizability. Conclusions: By effectively modeling the spatiotemporal dynamics of EEG signals, DR-Net provides a reliable and generalizable approach to the objective diagnosis of depression and advances computational neurodiagnostic by integrating dynamic graph propagation, self-attention mechanisms, and data-efficient augmentation. These findings suggest that DR-Net may support future EEG-based clinical assessment of depression.