EUVIMEDEUVIMED entdecken
Crossref

SemSTNet: Medical EEG Semantic Metric Learning With Class Prototypes Generated by Pretrained Language Model

Quanlin Chen, Chunjin Ye, Rui Xiao, Jiahui Pan, Jingcong Li

IEEE Transactions on Biomedical Engineering · 2026-05 · 73 (5) · 2010-2020

DOI: 10.1109/tbme.2025.3620754 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Transactions on Biomedical Engineering, 2026-05. Verfasst von Quanlin Chen und 4 weiteren Autor:innen. Hauptthemen laut MEDLINE: Electroencephalography, Deep Learning, Signal Processing, Computer-Assisted. Die Arbeit wurde bisher 2-mal zitiert (Crossref) und nennt 53 Literaturangaben. Gefördert von: STI 2030-Major Projects.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Electroencephalography (EEG) feature learning is crucial for brain-machine interfaces and medical diagnostics. Existing deep learning models for classification often overlook the intrinsic semantic relationships between different EEG classes and rely on overly complex models with a large number of parameters. To address these challenges, we propose SemSTNet, a novel and lightweight framework for EEG analysis. Firstly, we designed an efficient, lightweight convolutional architecture that decouples spatial and temporal feature extraction. Then we propose a framework which introduces a novel semantic metric learning paradigm that uses class prototypes generated by a pretrained language model to better capture inter-class relationships and enhance intra-class compactness. These prototypes are extracted and stored offline, requiring no additional inference from the language model during training or deployment. This design significantly reduces model complexity, resulting in a model with only 23 K parameters-over 100 times fewer than common Transformer-based models. Extensive experiments demonstrate that SemSTNet outperforms state-of-the-art approaches on tasks such as epilepsy classification and sleep staging, highlighting its effectiveness and efficiency. Our work demonstrates that integrating semantic knowledge with a purpose-built lightweight architecture provides a highly effective and efficient solution.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

ElectroencephalographyDeep LearningSignal Processing, Computer-AssistedHumansSemanticsEpilepsyAlgorithmsBrain-Computer Interfaces

Förderung

  • STI 2030-Major Projects

Zitieren

APA

Chen, Q., Ye, C., Xiao, R., Pan, J., & Li, J. (2026). SemSTNet: Medical EEG Semantic Metric Learning With Class Prototypes Generated by Pretrained Language Model. IEEE Transactions on Biomedical Engineering, 73(5), 2010-2020. https://doi.org/10.1109/tbme.2025.3620754

Vancouver

Chen Q, Ye C, Xiao R, Pan J, Li J. SemSTNet: Medical EEG Semantic Metric Learning With Class Prototypes Generated by Pretrained Language Model. IEEE Transactions on Biomedical Engineering. 2026;73(5):2010-2020. doi:10.1109/tbme.2025.3620754

BibTeX
@article{Chen2026,
  author = {Chen, Quanlin and Ye, Chunjin and Xiao, Rui and Pan, Jiahui and Li, Jingcong},
  title = {SemSTNet: Medical EEG Semantic Metric Learning With Class Prototypes Generated by Pretrained Language Model},
  journal = {IEEE Transactions on Biomedical Engineering},
  year = {2026},
  volume = {73},
  number = {5},
  pages = {2010--2020},
  doi = {10.1109/tbme.2025.3620754}
}
Kostenlos starten →