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

TL;DR

This paper introduces CRCC, a novel contrastive learning framework that enhances the generalization of EEG-based models across different sites and subjects by addressing site-dependent biases.

Contribution

The paper reformulates cross-site EEG learning as a bias-factorized problem and proposes a two-stage training method with data standardization, contrastive learning, and adversarial optimization.

Findings

CRCC outperforms existing methods on a multi-site EEG benchmark.

Achieves 10.7% higher balanced accuracy in zero-shot site transfer.

Effectively mitigates site-dependent biases in EEG models.

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.