ELPG-DTFS: Prior-Guided Adaptive Time-Frequency Graph Neural Network for EEG Depression Diagnosis

2025-09-29

AI Summary

Existing EEG-based depression diagnosis models suffer from limited interpretability and accuracy due to static spectral representations, fixed graph topologies, and neglect of neuroscientific priors. To address these limitations, this paper proposes the Adaptive Time-Frequency Graph Neural Network (ATF-GNN). Our method introduces three key innovations: (1) a channel–frequency-band attention mechanism grounded in cross-band mutual information, enabling dynamic selection of discriminative time-frequency features; (2) a learnable, subject-specific adjacency matrix to model individualized functional brain connectivity; and (3) a prior-knowledge-informed residual graph pathway that enhances biological plausibility. Evaluated on the 128-channel MODMA dataset, ATF-GNN achieves 97.63% accuracy and 97.33% F1-score, significantly outperforming state-of-the-art methods. Ablation studies confirm the substantial contribution of each component. The model thus delivers both superior diagnostic performance and clinically interpretable neurophysiological insights.

Technology Category

• Machine Learning: Graph-based Machine Learning
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding
• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Timely and objective screening of major depressive disorder (MDD) is vital, yet diagnosis still relies on subjective scales. Electroencephalography (EEG) provides a low-cost biomarker, but existing deep models treat spectra as static images, fix inter-channel graphs, and ignore prior knowledge, limiting accuracy and interpretability. We propose ELPG-DTFS, a prior-guided adaptive time-frequency graph neural network that introduces: (1) channel-band attention with cross-band mutual information, (2) a learnable adjacency matrix for dynamic functional links, and (3) a residual knowledge-graph pathway injecting neuroscience priors. On the 128-channel MODMA dataset (53 subjects), ELPG-DTFS achieves 97.63% accuracy and 97.33% F1, surpassing the 2025 state-of-the-art ACM-GNN. Ablation shows that removing any module lowers F1 by up to 4.35, confirming their complementary value. ELPG-DTFS thus offers a robust and interpretable framework for next-generation EEG-based MDD diagnostics.

Problem

• Diagnosing major depressive disorder using EEG data objectively
• Overcoming limitations of static spectral and fixed graph models
• Integrating neuroscience priors for improved accuracy and interpretability

Innovation

• Channel-band attention with cross-band mutual information
• Learnable adjacency matrix for dynamic functional links
• Residual knowledge-graph pathway injecting neuroscience priors