Neuro-GPT: Towards A Foundation Model for EEG

TL;DR

Neuro-GPT introduces a foundation model for EEG data that leverages large-scale pre-training and fine-tuning, significantly improving motor imagery classification performance in low-data scenarios.

Contribution

The paper presents Neuro-GPT, a novel EEG foundation model combining an encoder and GPT, pre-trained on large datasets to enhance BCI task performance with limited data.

Findings

Pre-trained Neuro-GPT outperforms models trained from scratch.

Foundation model effectively addresses EEG data scarcity.

Demonstrates generalizability across subjects.

Abstract

To handle the scarcity and heterogeneity of electroencephalography (EEG) data for Brain-Computer Interface (BCI) tasks, and to harness the power of large publicly available data sets, we propose Neuro-GPT, a foundation model consisting of an EEG encoder and a GPT model. The foundation model is pre-trained on a large-scale data set using a self-supervised task that learns how to reconstruct masked EEG segments. We then fine-tune the model on a Motor Imagery Classification task to validate its performance in a low-data regime (9 subjects). Our experiments demonstrate that applying a foundation model can significantly improve classification performance compared to a model trained from scratch, which provides evidence for the generalizability of the foundation model and its ability to address challenges of data scarcity and heterogeneity in EEG. The code is publicly available at github.com/wenhui0206/NeuroGPT.