Uni-NTFM: A Unified Foundation Model for EEG Signal Representation Learning

2025-09-28

AI Summary

Existing EEG foundation models suffer from three key limitations: (1) entanglement of time-frequency features, (2) neglect of electrode spatial topology, and (3) lack of task-adaptive architecture. To address these, we propose Uni-NTFM—the first decoupled universal EEG foundation model—featuring a parallel encoder architecture for raw signal, time-domain, and frequency-domain representations; a neuroscience-informed unified electrode topology embedding; and a Mixture-of-Experts Transformer backbone trained via dual-domain masked reconstruction. Trained on 28,000 hours of multi-source EEG data, Uni-NTFM achieves state-of-the-art performance across nine diverse downstream tasks. Notably, Uni-NTFM_large (1.9B parameters) demonstrates exceptional cross-task and cross-device generalization under both linear probing and fine-tuning paradigms. This work establishes a new paradigm for interpretable and scalable EEG representation learning.

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
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

Foundation models pretrained on various and unlabeled data have demonstrated significant success in natural language and vision, but their application to electroencephalography (EEG) remains challenged due to the signal's unique properties. Existing brain foundation models that inherit architectures designed for text or images lead to three limitations in pre-training: 1) conflating time-domain waveform patterns with frequency-domain rhythmic features in a single processing stream, 2) ignoring the critical spatial topology of electrodes with different standards, and 3) reliance on the inflexible, dense network to process functionally distinct EEG patterns. To address these challenges, we introduce the Unified Neural Topological Foundation Model (Uni-NTFM), which is designed based on neuroscience principles to produce universal and interpretable representations. Uni-NTFM integrates three core innovations: 1) a decoupled architecture parallelly encodes time, frequency, and raw signal representations before performing cross-domain feature integration; 2) a topological embedding mechanism to unify electrodes from different international standards and generate structured input sequences for brain regions; and 3) a Mixture-of-Experts neural Transformer that efficiently scales model capacity by routing signal patterns to specialized subnetworks. The largest model, Uni-NTFM$_{large}$, has a record-breaking 1.9B parameters and was pretrained on over 28,000 hours of diverse EEG data via a dual-domain masked reconstruction objective. Uni-NTFM significantly outperforms existing task-specific methods and foundation models across nine distinct downstream tasks under both linear probing and fine-tuning settings, demonstrating a superior ability to learn universal representations of brain activity.

Problem

• Addressing EEG signal's unique properties in foundation models
• Overcoming limitations in time-frequency and spatial processing
• Creating universal interpretable brain activity representations

Innovation

• Decoupled architecture for time-frequency cross-domain integration
• Topological embedding unifies electrodes across international standards
• Mixture-of-Experts Transformer routes patterns to specialized subnetworks