A Novel Depression Recognition Network Based on Brain Functional Connectivity from Prefrontal Emotional State EEG
Yubing Sun1, Zijian Zhou1, Jiaqi Sun1
1Research Field of Medical Instruments and Bioinformation Processing, College of Instrumentation & Electrical Engineering, Jilin University, No. 938 West Democracy Street, Changchun 130061, China.
Brain Sciences
|July 28, 2026
Summary
This study introduces DR-Net, an advanced framework for recognizing depression using electroencephalogram (EEG) signals. DR-Net achieves high accuracy by modeling dynamic brain activity patterns, offering a reliable tool for objective depression diagnosis.
Area of Science:
- Neuroscience
- Computational Psychiatry
- Machine Learning
Background:
- Electroencephalogram (EEG) signals offer a noninvasive method for detecting neural dysfunctions in depression.
- Existing methods face limitations due to small sample sizes and inadequate spatiotemporal feature modeling.
Purpose of the Study:
- To develop an EEG-based depression recognition framework that captures dynamic neural activity patterns.
- To address limitations in current depression detection methods using EEG data.
Main Methods:
- Proposed DR-Net framework integrating generative adversarial network (GAN), transition propagation graph convolution network (TPGCN), and transformer architectures.
- Utilized GAN for data augmentation to enhance data diversity and model generalization.
- Employed TPGCN to model dynamic brain functional connectivity and transformer for long-range temporal dependencies, focusing on the frontal lobe.
Main Results:
- Achieved 98.7% accuracy in depression recognition, outperforming baseline and state-of-the-art methods.
- Demonstrated strong robustness and generalizability with 96.13% accuracy on clinical datasets.
- The method showed no significant class bias.
Conclusions:
- DR-Net effectively models EEG spatiotemporal dynamics for reliable and generalizable objective depression diagnosis.
- The framework advances computational neurodiagnostics by integrating dynamic graph propagation, self-attention, and data augmentation.
- Findings suggest DR-Net's potential for future EEG-based clinical depression assessment.
