https://doi.org/10.1140/epjs/s11734-025-02053-5
Regular Article
A pre-modeling interpretable epilepsy detection model with multi-feature fusion and graph attention network
1
School of Aerospace Engineering, Xi’an Jiaotong University, 710049, Xi’an, Shaanxi, People’s Republic of China
2
State Key Laboratory for Strength and Vibration of Mechanical Structures, Xi’an Jiaotong University, 710049, Xi’an, Shaanxi, People’s Republic of China
a
This email address is being protected from spambots. You need JavaScript enabled to view it.
Received:
30
July
2025
Accepted:
28
October
2025
Published online:
17
November
2025
Combining electroencephalogram (EEG) signals and deep learning algorithms to automatically detect epileptic seizure has achieved prospective effectiveness. But classical deep learning methods struggle to utilize the spatial information of non-Euclidean EEG signals, leading to accuracy and sensitivity losses, and their interpretability is also a great challenge. Here, we proposed a pre-modeling interpretable epilepsy detection model which adopts the firing information and interaction between EEG channels. The raw EEG was compressed into graph signals. The edges were computed by the phase lag index to represent the spatial information of EEG signals, and the nodes were fusion features that contain temporal, dynamical, and cross-frequency coupling information of the clinical EEG recordings. A graph attention-based detection model was built to contain graph attention network layers and a node-wise feed-forward layer. Experiments on the CHB-MIT dataset showed that the accuracy, precision, and sensitivity of the model can reach 98.77%, 99.22%, and 98.54%, respectively, which is better or comparable to the state-of-the-art models. Furthermore, we conducted ablation experiments to assess the impact of different features on the model’s performance. The results were compared with the statistical analysis of these features during different seizure periods, validating the pre-modeling interpretability of the model. This research provides a new perspective in neuro disease automatic detection.
Copyright comment Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
© The Author(s), under exclusive licence to EDP Sciences, Springer-Verlag GmbH Germany, part of Springer Nature 2025
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
