Related Experiment Video
Updated: Apr 29, 2026
07:12
Individualized rTMS Treatment for Depression using an fMRI-Based Targeting Method
Published on: August 2, 2021
3.5K
GCTNet: a graph convolutional transformer network for major depressive disorder detection based on EEG signals
Yuwen Wang1, Yudan Peng1, Mingxiu Han1
1Beijing Advanced Innovation Center for Biomedical Engineering, School of Biological Science and Medical Engineering, Beihang University, Beijing, People's Republic of China.
Journal of Neural Engineering
|May 24, 2024
Summary
This study introduces a novel Graph Convolutional Transformer Network (GCTNet) for detecting major depressive disorder (MDD) using electroencephalogram (EEG) signals. The GCTNet shows high accuracy, offering a promising tool for objective MDD diagnosis.
Area of Science:
- Neuroscience
- Artificial Intelligence
- Biomedical Engineering
Background:
- Accurate identification of major depressive disorder (MDD) remains challenging, particularly using objective physiological measures.
- Electroencephalogram (EEG) signals offer a promising avenue for objective MDD detection due to their rich spatiotemporal information.
Purpose of the Study:
- To develop and validate a novel deep learning framework, the Graph Convolutional Transformer Network (GCTNet), for accurate and reliable MDD detection using EEG signals.
- To enhance feature extraction and classification performance through the integration of spatial and temporal information processing and a novel contrastive loss function.
Main Methods:
- The proposed GCTNet integrates residual graph convolutional network blocks for spatial feature extraction and Transformer blocks for temporal dynamics analysis.
- A contrastive cross-entropy (CCE) loss function was developed, combining contrastive learning with cross-entropy to improve feature stability and discriminability.
- The model was evaluated on a dataset of 41 MDD patients and 44 controls, alongside a public dataset, using subject-independent partitioning and 10-fold cross-validation.
Main Results:
- The GCTNet model achieved significant performance, with an average Area Under the Curve (AUC) of 0.7693 on the in-house dataset and 0.9755 on the public dataset.
- Comparative analysis confirmed the superiority of the GCTNet framework with CCE loss over existing state-of-the-art algorithms for MDD detection.
- The proposed method demonstrated robust performance in subject-independent classification.
Conclusions:
- The GCTNet framework provides an objective and effective method for detecting major depressive disorder (MDD) using EEG signals.
- The integration of graph convolutional and transformer networks, coupled with CCE loss, significantly enhances classification accuracy.
- This approach offers valuable support for clinical-assisted diagnosis of MDD.
