🤖 AI Summary
This study investigates whether general-purpose time-series foundation models (TSMs) can serve as robust, interpretable feature extractors in low-resource clinical EEG settings—replacing task-specific models. We propose a domain-agnostic transfer learning framework and systematically evaluate diagnostic performance on three key tasks: age prediction, seizure detection, and clinical event classification. Feature attribution methods are integrated to identify neurophysiologically meaningful biomarkers. Our results demonstrate, for the first time: (1) general-purpose TSMs significantly outperform state-of-the-art specialized EEG models; (2) performance is highly sensitive to architectural choices—particularly context length; and (3) these models exhibit strong cross-task generalization and intrinsic interpretability, effectively mitigating data scarcity in clinical EEG analysis. These findings validate the practical utility and translational potential of foundation models for real-world EEG-based diagnostics.
📝 Abstract
The success of foundation models in natural language processing and computer vision has motivated similar approaches for general time series analysis. While these models are effective for a variety of tasks, their applicability in medical domains with limited data remains largely unexplored. To address this, we investigate the effectiveness of foundation models in medical time series analysis involving electroencephalography (EEG). Through extensive experiments on tasks such as age prediction, seizure detection, and the classification of clinically relevant EEG events, we compare their diagnostic accuracy with that of specialised EEG models. Our analysis shows that foundation models extract meaningful EEG features, outperform specialised models even without domain adaptation, and localise task-specific biomarkers. Moreover, we demonstrate that diagnostic accuracy is substantially influenced by architectural choices such as context length. Overall, our study reveals that foundation models with general time series understanding eliminate the dependency on large domain-specific datasets, making them valuable tools for clinical practice.
