Answers are grounded in this paper and cite their sources.
Existing EEG functional network methods have provided important insights into affective disorders, but the direction, timing, and modulation of inter-regional influences are not fully characterized within a unified representation. To address this limitation, we propose an attention-based framework for interpretable modeling of asynchronous functional brain networks (aFBNs) from a causal-discovery perspective. Its Cross-Attention Causal Discovery (CACD) module jointly estimates directed influence strength, transmission delay, and modulation sign from multivariate EEG. CACD integrates a dual-branch decoder for separating intra- and inter-channel dynamics, a learnable cross-attention mask for connection selection, and an Attention-Gradient Consistency (AGC) mechanism for improving attribution consistency and delay localization. Experiments on nine causal discovery benchmark datasets demonstrate the effectiveness of CACD in recovering causal structures and estimating transmission delays. On three clinical EEG datasets, the framework achieves competitive subject-independent classification performance and reveals neurophysiologically plausible asynchronous connectivity patterns. Delay and modulation information provide additional discriminative value beyond directed influence strength. Together, these findings support aFBNs as an interpretable representation for characterizing affective-disorder-related neural dynamics.
No takes yet. Share an insight, caveat, or question.
Wang et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: