基于时序-拓扑一致性图对比模型的EEG抑郁识别

Temporal-Topological Consistency Graph Contrastive Model for EEG Depression Recognition

摘要

重度抑郁症 (Major Depressive Disorder，MDD) 是严重影响人类健康的精神疾病，现阶段基于对比学习 (Contrastive Learning，CL) 的脑电信号 (ElectroEncephaloGrams，EEG) 抑郁识别取得良好效果，但仍存在如下局限性：目前的 CL 方法主要聚焦于 EEG 的时序一致性，忽略了 EEG 的拓扑一致性，即各通道的抗干扰稳定性与通道间的固有神经生理学关联性。为解决上述问题，本文提出基于时序-拓扑一致性图对比学习 (Temporal-Topological Consistency Graph Contrastive Learning，TTGCL) 的抑郁识别模型。首先，设计自适应图增强策略，包括基于信噪比与频域能量分布的动态节点增强和基于动态正则约束的边增强，以生成高质量的对比视图。其次，提出跨窗口时序一致性对比 (Cross-Window Temporal Consistency Contrast，CWTCC)，通过双视图交叉预测学习 EEG 的时序一致性。最后，构建区域 -全局图对比 (Regional-Global Graph Contrast，RGGC) 框架，从局部脑区到全局样本层面学习拓扑一致性。本文在两个公开 MDD 数据集上分别取得了 96.53% 和 96.33% 的识别准确率，优于现有基线模型。

Abstract

Major Depressive Disorder (MDD) is a mental illness that severely affects human health. Currently, depression recognition based on Contrastive Learning (CL) for Electroencephalogram (EEG) signals has achieved promising results, yet it still has certain limitations: Mainstream CL methods primarily focus on the temporal consistency of EEG signals while neglecting the topological consistency of EEG data, namely the anti-interference stability of individual channels and the inherent neurophysiological correlations between channels. To address the above issues, this paper proposes a novel depression recognition model based on Temporal-Topological Consistency Graph Contrastive Learning (TTGCL). First, an adaptive graph augmentation strategy is designed, which includes dynamic node augmentation based on signal-to-noise ratio and frequency-domain energy distribution, as well as edge augmentation based on dynamic regular constraints, to generate high-quality contrastive views. Second, Cross-Window Temporal Consistency Contrast (CWTCC) is proposed to learn the temporal consistency of EEG signals through cross-prediction between dual views. Finally, a Regional-Global Graph Contrast (RGGC) framework is constructed to learn topological consistency from the level of local brain regions to global samples. This study achieves recognition accuracies of 96.53% and 96.33% on two public MDD datasets, respectively, outperforming existing baselines.

关键词

脑电图 / 抑郁识别 / 对比学习 / 图神经网络 / 拓扑一致性

Key words

EEG / depression recognition / contrastive learning / graph neural network / topological consistency

参考文献

[1] Institute of Health Metrics and Evaluation. Global health data exchange[EB/OL]. (2023-03-04). https://vizhub.healthdata.org/gbd-results/.

[2] KIM S A, BAE S M. Impacts of perceived stress, neglect, victim and respect for human rights on depression of adolescents[J].Child Psychiatry & Human Development, 2024, 55(5):1325-1332.

[3] ZHONG P, WANG D, MIAO C. EEG-based emotion recognition using regularized graph neural networks[J].IEEE Transactions on Affective Computing, 2020, 13(3):1290-1301.

[4] ACHARYA U R, SUDARSHAN V K, ADELI H, et al. A novel depression diagnosis index using nonlinear features in EEG signals[J].European Neurology, 2015, 74(1/2):79-83.

[5] BAI R, GUO Y, TAN X, et al. An EEG-based depression detection method using machine learning model[J].International Journal of Pharma Medicine and Biological Sciences, 2021, 10(1):17-22.

[6] WANG D, LEI C, ZHANG X, et al. Identification of depression with a semi-supervised GCN based on EEG data[C]//2021 IEEE International Conference on Bioinformatics and Biomedicine (BIBM). Piscataway, NJ, USA: IEEE, 2021:2338-2345.

[7] WANG Y, ZHENG W, LI Y, et al. A hybrid graph neural network for enhanced EEG-Based depression detection[C]//2025 International Joint Conference on Neural Networks (IJCNN). Piscataway, NJ, USA: IEEE, 2025:1-8.

[8] LU H, YOU Z, GUO Y, et al. Mast-gcn: Multi-scale adaptive spatial-temporal graph convolutional network for EEG-based depression recognition[J].IEEE Transactions on Affective Computing, 2024, 15(4):1985-1996.

[9] WANG Y, XU Y, YANG J, et al. Fully-connected spatial-temporal graph for multivariate time-series data[C]//Proceedings of the AAAI Conference on Artificial Intelligence. Washington, DC, USA: AAAI, 2024:15715-15724.

[10] CUI W, SUN M, DONG Q, et al. A multiview sparse dynamic graph convolution-based region-attention feature fusion network for major depressive disorder detection[J].IEEE Transactions on Computational Social Systems, 2023, 11(2):2691-2702.

[11] JIA Z, LIN Y, WANG J, et al. Multi-view spatial-temporal graph convolutional networks with domain generalization for sleep stage classification[J].IEEE Transactions on Neural Systems and Rehabilitation Engineering, 2021, 29:1977-1986.

[12] TIAN Y, SUN C, POOLE B, et al. What makes for good views for contrastive learning?[J].Advances in Neural Information Processing Systems, 2020, 33:6827-6839.

[13] ELDELE E, RAGAB M, CHEN Z, et al. Time-series representation learning via temporal and contextual contrasting[EB/OL]. (2021-06-26). https://arxiv.org/abs/2106.14112.

[14] TONEKABONI S, EYTAN D, GOLDENBERG A. Unsupervised representation learning for time series with temporal neighborhood coding[EB/OL]. (2021-06-01). https://arxiv.org/abs/2106.00750.

[15] PöPPELBAUM J, CHADHA G S, SCHWUNG A. Contrastive learning based self-supervised time-series analysis[J].Applied Soft Computing, 2022, 117:108397.

[16] KHAERTDINOV B, GHALEB E, ASTERIADIS S. Contrastive self-supervised learning for sensor-based human activity recognition[C]//2021 IEEE International Joint Conference on Biometrics (IJCB). Piscataway, USA: IEEE, 2021:1-8.

[17] CHOI H, KANG P. Multi-task self-supervised time-series representation learning[J].Information Sciences, 2024, 671:120654.

[18] FRIEDMAN N P, ROBBINS T W. The role of prefrontal cortex in cognitive control and executive function[J].Neuropsychopharmacology, 2022, 47(1):72-89.

[19] HUANG M H, FAN S Y, LIN I M. EEG coherences of the fronto-limbic circuit between patients with major depressive disorder and healthy controls[J].Journal of Affective Disorders, 2023, 331:112-120.

[20] HELM K, VIOL K, WEIGER T M, et al. Neuronal connectivity in major depressive disorder: A systematic review[J].Neuropsychiatric Disease and Treatment, 2018:2715-2737.

[21] NYSTROM C, MATOUSEK M, HALLSTROM T. Relationships between EEG and clinical characteristics in major depressive disorder[J].Acta Psychiatrica Scandinavica, 1986, 73(4):390-394.

[22] SAWAYA H, JOHNSON K, SCHMIDT M, et al. Resting-state functional connectivity of antero-medial prefrontal cortex sub-regions in major depression and relationship to emotional intelligence[J].International Journal of Neuropsychopharmacology, 2015, 18(6):pyu112.

[23] WANG Y, XU Y, YANG J, et al. Graph-aware contrasting for multivariate time-series classification[C]//Proceedings of the AAAI Conference on Artificial Intelligence. Washington, DC, USA: AAAI, 2024:15725-15734.

[24] ROMMEL C, PAILLARD J, MOREAU T, et al. Data augmentation for learning predictive models on EEG: A systematic comparison[J].Journal of Neural Engineering, 2022, 19(6):066020.

[25] ARI B, SIDDIQUE K, ALçIN ö F, et al. Wavelet ELM-AE based data augmentation and deep learning for efficient emotion recognition using EEG recordings[J].IEEE Access, 2022, 10:72171-72181.

[26] BOGGESS A, NARCOWICH F J. A First Course in Wavelets with Fourier Analysis[M]. Hoboken, New Jersey, USA: John Wiley & Sons,2009.

[27] Wang Y, Wu M, Li X, et al. Multivariate time-series representation learning via hierarchical correlation pooling boosted graph neural network[J].IEEE Transactions on Artificial Intelligence, 2023, 5(1):321-333.

[28] YOU Y, CHEN T, SUI Y, et al. Graph contrastive learning with augmentations[J].Advances in Neural Information Processing Systems, 2020, 33:5812-5823.

[29] OORD A, LI Y, VINYALS O. Representation learning with contrastive predictive coding[EB/OL]. (2018-07-10). https://arxiv.org/abs/1807.03748.

[30] VASWANI A, SHAZEER N, PARMAR N, et al. Attention is all you need[C]//Advances in Neural Information Processing Systems. [s.l.]: Curran Associated, Inc,2017:5998-6008.

[31] CAI H, GAO Y, SUN S, et al. Modma dataset: A multimodal open dataset for mental-disorder analysis[EB/OL]. (2020-02-20). https://arxiv.org/abs/2002.09283.

[32] MUMTAZ W, XIA L, Mohd Yasin M A, et al. A wavelet-based technique to predict treatment outcome for major depressive disorder[J].PloS One, 2017, 12(2):e0171409.

[33] Comon P. Independent component analysis, a new concept?[J].Signal Processing, 1994, 36(3):287-314.

[34] FU Z, ZHU H, ZHAO Y, et al. GMAEEG: A self-supervised graph masked autoencoder for eeg representation learning[J].IEEE Journal of Biomedical and Health Informatics, 2024, 28(11):5048-5057.

[35] LEE C H, KIM H, HAN H, et al. Neuronet: A novel hybrid self-supervised learning framework for sleep stage classification using single-channel eeg[EB/OL]. (2024-04-10). https://arxiv.org/abs/2404.17585.

[36] ZHENG K, YU S, CHEN B. CI-GNN: A Granger causality-inspired graph neural network for interpretable brain network-based psychiatric diagnosis[J].Neural Networks, 2024, 172:106147.

[37] VAN DER VINNE N, VOLLEBREGT M A, VAN PUTTEN M J A M, et al. Frontal alpha asymmetry as a diagnostic marker in depression: Fact or fiction? A meta-analysis[J].Neuroimage: Clinical, 2017, 16:79-87.

[38] NEWSON J J, THIAGARAJAN T C. EEG frequency bands in psychiatric disorders: A review of resting state studies[J].Frontiers in Human Neuroscience, 2019, 12:521.