Contrastive learning based multi-scale spatio-temporal-spectral network for patient-specific EEG seizure prediction

Accurate seizure prediction using electroencephalogram (EEG) signals is crucial for improving patients’ quality of life. However, the latent representations of preictal and interictal samples are difficult to effectively distinguish within the feature space. In this study, we propose a patient-specific seizure prediction framework that combines contrastive pre-training with multi-scale spatio-temporal-spectral EEG modeling.

Instead of relying on cross-patient transfer, the proposed contrastive strategy regularizes the intra-patient latent space by enhancing the consistency of seizure-related representations before supervised fine-tuning. A multi-scale encoder is further designed to capture temporal dynamics, frequency-specific neural rhythms, and graph attention-guided spatial dependencies among EEG channels. Window-level predictions are finally converted into event-level alarms through a post-processing strategy.

Experiments on the CHB-MIT and Siena datasets demonstrate that the proposed method achieves superior performance compared with representative baselines, with an average AUC of 0.942, sensitivity of 94.61%, and FPR/h of 0.034 on CHB-MIT, and an AUC of 0.935, sensitivity of 92.34%, and FPR/h of 0.090 on Siena. These results suggest that the proposed method can improve both discriminability and alarm reliability in seizure prediction. This work was supported by the JSPS KAKENHI (Nos. 24K21599 and 26K00914), the Shenzhen Basic Research Program (No.

JCYJ20241202124804007), and the Key-Area Research and Development Program of Guangdong Province under Grant (No. 2023B0303030002). Yue Du and Yifei Han have authors contributed equally to this work. Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, Shenzhen, 518055, China Centre for Artificial Intelligence and Robotics, Hong Kong Institute of Science & Innovation, Chinese Academy of Sciences, Hong Kong SAR, 999077, China Department of Biomedical Engineering, City University of Hong Kong, Hong Kong SAR, 999077, China Shenzhen University of Advanced Technology, Shenzhen, 518107, Guangdong, China Southeast University School of Medicine, Nanjing, 210009, China Okayama University, Okayama, 700-8530, Japan Correspondence to Zhiqing Zhang or Jinglong Wu.

The authors declare no competing interests. Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material.

You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.