Leveraging Generic Time Series Foundation Models for EEG Classification

TL;DR

The paper addresses EEG classification under data scarcity and inter-subject variability by evaluating time-series foundation models (TSFMs) as cross-domain backbones. It leverages a generalist TSFM, Mantis, along with an EEG-specific baseline, CBraMod, and a traditional EEGNet baseline to test two pretraining regimes: real heterogeneous data and synthetic CauKer data, with end-to-end fine-tuning. The results show that Mantis achieves strong transfer to EEG tasks, often surpassing EEGNet and CBraMod, and that synthetic pretraining can be highly effective, highlighting the potential of cross-domain pretraining for brain signals. This work suggests a paradigm shift toward generalist TSFMs for EEG, enabling robust performance with limited task-specific data and guiding future cross-domain development for neural data analysis.

Abstract

Foundation models for time series are emerging as powerful general-purpose backbones, yet their potential for domain-specific biomedical signals such as electroencephalography (EEG) remains rather unexplored. In this work, we investigate the applicability a recently proposed time series classification foundation model, to a different EEG tasks such as motor imagery classification and sleep stage prediction. We test two pretraining regimes: (a) pretraining on heterogeneous real-world time series from multiple domains, and (b) pretraining on purely synthetic data. We find that both variants yield strong performance, consistently outperforming EEGNet, a widely used convolutional baseline, and CBraMod, the most recent EEG-specific foundation model. These results suggest that generalist time series foundation models, even when pretrained on data of non-neural origin or on synthetic signals, can transfer effectively to EEG. Our findings highlight the promise of leveraging cross-domain pretrained models for brain signal analysis, suggesting that EEG may benefit from advances in the broader time series literature.