arXiv:2510.21585v1 Announce Type: new 
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.

تقديم REVE، نموذج أساسي لتخطيط الدماغ الكهربائي (EEG)، يمثل تقدمًا كبيرًا في مجال الذكاء الاصطناعي. من خلال الاستفادة من التدريب المسبق على نطاق واسع باستخدام بيانات من 25,000 موضوع، يهدف REVE إلى التغلب على التحديات التي تطرحها البروتوكولات والأجهزة المتنوعة المستخدمة في جمع بيانات EEG. هذه الابتكار مهم لأنه يعد بتحسين تعميم نماذج EEG، مما يجعلها أكثر قابلية للتكيف وفعالية عبر إعدادات مختلفة. مع تزايد الطلب على تحليل EEG موثوق، يمكن أن يمهد REVE الطريق لتطبيقات أكثر قوة في علم الأعصاب والبيئات السريرية.

La introducción de REVE, un modelo fundamental para EEG, marca un avance significativo en el campo de la inteligencia artificial. Al aprovechar el preentrenamiento a gran escala con datos de 25,000 sujetos, REVE busca superar los desafíos que presentan los diversos protocolos y dispositivos utilizados en la recolección de datos de EEG. Esta innovación es crucial, ya que promete mejorar la generalización de los modelos de EEG, haciéndolos más adaptables y efectivos en diferentes configuraciones. A medida que crece la demanda de un análisis de EEG confiable, REVE podría allanar el camino para aplicaciones más sólidas en neurociencia y entornos clínicos.

L'introduction de REVE, un modèle fondamental pour l'EEG, marque une avancée significative dans le domaine de l'intelligence artificielle. En s'appuyant sur un préentraînement à grande échelle sur des données provenant de 25 000 sujets, REVE vise à surmonter les défis posés par les protocoles et dispositifs divers utilisés dans la collecte de données EEG. Cette innovation est cruciale car elle promet d'améliorer la généralisation des modèles EEG, les rendant plus adaptables et efficaces dans différents contextes. Alors que la demande pour une analyse EEG fiable augmente, REVE pourrait ouvrir la voie à des applications plus robustes en neurosciences et en milieu clinique.

The introduction of REVE, a foundation model for EEG, marks a significant advancement in the field of artificial intelligence. By leveraging large-scale pretraining on data from 25,000 subjects, REVE aims to overcome the challenges posed by the diverse protocols and devices used in EEG data collection. This innovation is crucial as it promises to enhance the generalization of EEG models, making them more adaptable and effective across different setups. As the demand for reliable EEG analysis grows, REVE could pave the way for more robust applications in neuroscience and clinical settings.

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

One More Thing in AI – Your Shortcut to AI Mastery

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

Was this article worth reading? Share it

One More Thing in AI

LucidQuery AI

Research AI

Brainactive

Zemith-3bda3b

Octofy

Ready to build your own newsroom?