CRCC: Contrast-Based Robust Cross-Subject and Cross-Site Representation Learning for EEG

2026-02-22

AI Summary

This study addresses the poor generalization of EEG neural decoding models across acquisition sites, revealing that such limitations stem from implicit reliance on site-specific structured biases during training. To mitigate this, the authors formulate cross-site clinical EEG learning as a bias factorization problem, identifying three fundamental bias factors and proposing a unified training framework that integrates data normalization with representation-level constraints. They introduce a novel two-stage training paradigm—the first to combine cross-subject/cross-site contrastive learning with site-adversarial optimization—and establish the first standardized multi-site EEG benchmark dataset for depression. Under a strict zero-shot site-transfer setting, their approach achieves a 10.7 percentage point improvement in balanced accuracy over the current state of the art, substantially enhancing robust generalization in unseen environments.

Technology Category

• Cognitive Modeling & Cognitive Systems: Neural Spike Coding
• Machine Learning: Multimodal Learning
• Natural Language Processing: Learning & Optimization for NLP

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

EEG-based neural decoding models often fail to generalize across acquisition sites due to structured, site-dependent biases implicitly exploited during training. We reformulate cross-site clinical EEG learning as a bias-factorized generalization problem, in which domain shifts arise from multiple interacting sources. We identify three fundamental bias factors and propose a general training framework that mitigates their influence through data standardization and representation-level constraints. We construct a standardized multi-site EEG benchmark for Major Depressive Disorder and introduce CRCC, a two-stage training paradigm combining encoder-decoder pretraining with joint fine-tuning via cross-subject/site contrastive learning and site-adversarial optimization. CRCC consistently outperforms state-of-the-art baselines and achieves a 10.7 percentage-point improvement in balanced accuracy under strict zero-shot site transfer, demonstrating robust generalization to unseen environments.

Problem

• cross-site generalization
• EEG decoding
• domain shift
• site-dependent bias
• cross-subject learning

Innovation

• contrastive learning
• adversarial optimization
• bias factorization
• cross-site generalization
• EEG representation learning