NeuroDyn-EEG: An Interpretable Pre-trained Model for EEG Based on Neural Dynamics

概要

Clinical scalp electroencephalography (EEG) offers a noninvasive window into neural dynamics of neuropsychiatric disorders. However, discriminative deep models often lack anatomically indexed physiological interpretability. We propose NeuroDyn-EEG, a pretraining framework integrating generative priors from neural dynamics. It couples an extended Jansen-Rit neural mass model, leadfield-based source projection, and simulation-based parameter inversion. Trained on synthetic parameter-EEG pairs within physiological ranges, NeuroDyn-EEG estimates 11 regional parameter families across 90 AAL regions plus one global parameter from standard 19-channel EEG, using only ~2.43M trainable parameters. We evaluate the framework across three levels. First, controlled simulations demonstrate robust parameter recovery under diverse noise conditions, while real resting-state EEG evaluations confirm spectral and phase consistency in an inverse-forward closed loop. Second, on four clinical benchmarks (AD65, PD31, Figshare MDD, and TUAB), NeuroDyn-EEG achieves competitive classification performance, securing the highest BACC, AUROC, and AUCPR on PD31 and MDD, and highest BACC on AD65. Third, post hoc regional analyses reveal disease-specific alterations: local synaptic connectivity C_1 involves the most altered regions in AD65, whereas the firing threshold theta ranks first in MDD, offering testable mechanistic hypotheses. Overall, NeuroDyn-EEG maps scalp EEG to anatomically indexed dynamical parameters, bridging representation learning and mechanistic neurophysiology. Code: https://github.com/Gnosis-Neurodynamics/NeuroDyn-EEG.

参考リンク

• Paper ↗
• GitHub ↗

リンク先のライセンス・再現性・データ条件をご確認ください。自動分類や実装の目安は保証ではありません。

実装を始めるプロンプト

参考リンクを確認したうえで、次の内容をコピーして利用できます。

# 論文・記事
NeuroDyn-EEG: An Interpretable Pre-trained Model for EEG Based on Neural Dynamics

# 概要
Clinical scalp electroencephalography (EEG) offers a noninvasive window into neural dynamics of neuropsychiatric disorders. However, discriminative deep models often lack anatomically indexed…

# 原典
https://arxiv.org/abs/2609.36773

上記の手法を実装してください。実行前に依存関係とライセンスを確認してください。