A Competition-Cooperation Graph Adversarial Augmentation Learning with Application to Brain Disease Detection

作者：Manman Yuan, Yi-Xiang Wang, Wenjun Xiong, Jiapei Li, Ting Xu, Mengyi Shao · 年份：2026 · DOI：10.1109/icassp55912.2026.11461227 · 研究领域：Advanced Graph Neural Networks、Machine Learning in Healthcare、EEG and Brain-Computer Interfaces

Graph learning exhibits varying capabilities in modeling brain networks. However, existing methods often overlook critical factors: negative interactions between ROIs and distribution shifts caused by inter-subject physiological variations. This hampers functional connectivity modeling, reducing interpretability and the ability to identify out-of-distribution samples. We propose the Signed Graph Adversarial Augmentation Contrastive Learning Network (SGA-CLNet), a competition–cooperation framework for brain disease detection. SGACLNet builds a signed functional network (SFN) based on balance theory to model cooperative–competitive interactions. To counteract distribution shifts, we introduce Adversarial Stability Augmentation (ASA) to mitigate variations caused by physiological differences. Furthermore, a signed graph contrastive learning strategy preserves disease-relevant connectivity patterns. On three datasets, SGACLNet consistently surpasses state-of-the-art methods in accuracy and interpretability, providing a principled approach for brain disease detection.