Daily Arxiv

This is a page that curates AI-related papers published worldwide.
All content here is summarized using Google Gemini and operated on a non-profit basis.
Copyright for each paper belongs to the authors and their institutions; please make sure to credit the source when sharing.

From Video to EEG: Adapting Joint Embedding Predictive Architecture to Uncover Visual Concepts in Brain Signal Analysis

Created by
  • Haebom

Author

Amirabbas Hojjati, Lu Li, Ibrahim Hameed, Anis Yazidi, Pedro G. Lind, Rabindra Khadka

Outline

In this paper, we propose a novel self-supervised learning (SSL) model, EEG-VJEPA, for effective analysis of electroencephalography (EEG) signals. EEG-VJEPA applies the V-JEPA model for video processing to EEG, using joint embedding and adaptive masking to capture spatiotemporal dependencies. Experimental results using the publicly available TUH Abnormal EEG dataset demonstrate that EEG-VJEPA outperforms existing state-of-the-art models in classification accuracy, and can support collaboration with clinicians by capturing physiologically relevant spatiotemporal patterns and providing interpretable embeddings. This demonstrates the potential of EEG-VJEPA as a scalable and reliable EEG analysis framework in real clinical settings.

Takeaways, Limitations

•
Takeaways:
◦
We first applied V-JEPA to EEG classification, achieving excellent performance.
◦
Achieves higher classification accuracy than existing methods.
◦
Capture physiologically meaningful spatiotemporal patterns to provide interpretable embeddings.
◦
Presenting the potential of a scalable and reliable EEG analysis framework in real clinical settings.
•
Limitations:
◦
In this paper, only experimental results for a specific dataset (TUH Abnormal EEG) are presented, so verification of generalization performance for other datasets is necessary.
◦
Further research is needed on the interpretability of the model (more detailed analysis is needed on the claim that it captures physiologically meaningful patterns).
◦
There may be a lack of detailed description of the specific parameter tuning and optimization process of V-JEPA.
👍
Made with Slashpage