CLEF: EEG Foundation Model for Learning Clinical Semantics

2026-05-11

AI Summary

This work addresses the limitations of existing EEG foundation models, which typically operate on short signal windows and fail to model entire recording sessions or incorporate clinical context. The authors propose CLEF, a long-context EEG foundation model that, for the first time, aligns full EEG sessions with clinical semantics by transforming raw signals into three-dimensional multi-taper spectrogram tokens. CLEF leverages a Transformer architecture and contrastive learning to jointly perform masked reconstruction pretraining and alignment with neurologist reports and structured electronic health records (EHR). Evaluated on a new benchmark encompassing 234 tasks across more than 260,000 sessions, CLEF outperforms prior methods on 229 tasks, raising the average AUROC from 0.65 to 0.74, and demonstrates strong generalization across diverse clinical concepts and external cohorts.

Abstract

Clinical EEG interpretation requires reasoning over full EEG sessions and integrating signal patterns with clinical context. Existing EEG foundation models are largely designed for short-window decoding and do not incorporate clinical context. We introduce CLEF, a clinically grounded long-context EEG foundation model. CLEF represents EEG sessions as 3D multitaper spectrogram tokens, enabling tractable Transformer modeling at session scale, and aligns embeddings with neurologist reports and structured EHR data through contrastive objectives. We evaluate CLEF on a new 234-task benchmark spanning disease phenotypes, medication exposures, and EEG findings, with more than 260k EEG sessions from over 108k patients. CLEF outperforms prior EEG foundation models on 229 of 234 tasks, improving mean AUROC from 0.65 to 0.74. Reconstruction-only pretraining surpasses prior EEG foundation models, while report and EHR alignment yields further gains. Held-out concept and external-cohort experiments suggest that these representations transfer beyond observed alignment targets. These results support session-scale, clinically grounded representation learning as a promising foundation-model paradigm for clinical EEG.

Problem

• EEG foundation model
• clinical semantics
• long-context modeling
• clinical context integration
• session-scale representation

Innovation

• EEG foundation model
• long-context modeling
• clinical semantics
• multitaper spectrogram
• contrastive alignment