REVE: A Foundation Model for EEG -- Adapting to Any Setup with Large-Scale Pretraining on 25,000 Subjects

2025-10-24

AI Summary

Existing EEG foundation models exhibit poor cross-dataset generalization—particularly in linear probing—due to heterogeneity in acquisition devices, experimental protocols, and electrode configurations. To address this, we propose the first universal foundation model for EEG across all scenarios. Our method introduces a scalable 4D positional encoding scheme supporting arbitrary signal duration and electrode layouts; a largest-ever pretraining paradigm leveraging 92 datasets, 25,000 subjects, and over 60,000 hours of EEG data; and a unified architecture integrating masked self-supervised reconstruction, large-scale contrastive learning, spatiotemporal attention, and a flexible embedding structure adaptable to diverse electrode configurations. Evaluated on ten downstream tasks—including motor imagery, epilepsy detection, and sleep staging—our model achieves state-of-the-art performance. Notably, it enables high-accuracy spatiotemporal pattern modeling via linear probing without fine-tuning.

Technology Category

• Machine Learning: Deep Neural Architectures and Foundation Models
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding
• Natural Language Processing: (Large) Language Models

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Web Mining and Content Analysis: Large pretrained models with web data
• User Modeling, Personalization and Recommendation: Practical large-scale studies of user experience

Abstract

Foundation models have transformed AI by reducing reliance on task-specific data through large-scale pretraining. While successful in language and vision, their adoption in EEG has lagged due to the heterogeneity of public datasets, which are collected under varying protocols, devices, and electrode configurations. Existing EEG foundation models struggle to generalize across these variations, often restricting pretraining to a single setup, resulting in suboptimal performance, in particular under linear probing. We present REVE (Representation for EEG with Versatile Embeddings), a pretrained model explicitly designed to generalize across diverse EEG signals. REVE introduces a novel 4D positional encoding scheme that enables it to process signals of arbitrary length and electrode arrangement. Using a masked autoencoding objective, we pretrain REVE on over 60,000 hours of EEG data from 92 datasets spanning 25,000 subjects, representing the largest EEG pretraining effort to date. REVE achieves state-of-the-art results on 10 downstream EEG tasks, including motor imagery classification, seizure detection, sleep staging, cognitive load estimation, and emotion recognition. With little to no fine-tuning, it demonstrates strong generalization, and nuanced spatio-temporal modeling. We release code, pretrained weights, and tutorials to support standardized EEG research and accelerate progress in clinical neuroscience.

Problem

• Addressing EEG signal heterogeneity across varying protocols and devices
• Overcoming limited generalization of existing EEG foundation models
• Enabling flexible processing of arbitrary EEG signal configurations

Innovation

• 4D positional encoding for arbitrary EEG setups
• Pretrained on 60,000 hours using masked autoencoding
• Generalizes across 10 tasks with minimal fine-tuning