Long-range Brain Graph Transformer

TL;DR

ALTER introduces a brain graph transformer that explicitly captures long-range dependencies among ROIs by leveraging an Adaptive Long-range Aware (ALGA) strategy based on biased random walks guided by ROI correlations. Long-range embeddings are injected into a transformer to jointly model short- and long-range connectivity, achieving superior neurological disease diagnosis on ABIDE and ADNI compared with both generalized graph learners and brain-specific baselines. The approach is validated through extensive experiments, ablations, and interpretability analyses, with ablation showing the critical role of adaptive factors and hop-based long-range sampling. The work provides a principled framework for incorporating brain-wide communication patterns into graph representations and opens avenues for multimodal extensions and improved brain-network analysis.

Abstract

Understanding communication and information processing among brain regions of interest (ROIs) is highly dependent on long-range connectivity, which plays a crucial role in facilitating diverse functional neural integration across the entire brain. However, previous studies generally focused on the short-range dependencies within brain networks while neglecting the long-range dependencies, limiting an integrated understanding of brain-wide communication. To address this limitation, we propose Adaptive Long-range aware TransformER (ALTER), a brain graph transformer to capture long-range dependencies between brain ROIs utilizing biased random walk. Specifically, we present a novel long-range aware strategy to explicitly capture long-range dependencies between brain ROIs. By guiding the walker towards the next hop with higher correlation value, our strategy simulates the real-world brain-wide communication. Furthermore, by employing the transformer framework, ALERT adaptively integrates both short- and long-range dependencies between brain ROIs, enabling an integrated understanding of multi-level communication across the entire brain. Extensive experiments on ABIDE and ADNI datasets demonstrate that ALTER consistently outperforms generalized state-of-the-art graph learning methods (including SAN, Graphormer, GraphTrans, and LRGNN) and other graph learning based brain network analysis methods (including FBNETGEN, BrainNetGNN, BrainGNN, and BrainNETTF) in neurological disease diagnosis. Cases of long-range dependencies are also presented to further illustrate the effectiveness of ALTER. The implementation is available at https://github.com/yushuowiki/ALTER.

Figures (6)

• Figure 1: An illustration of long-range dependencies and short-range dependencies within human brain.
• Figure 2: The overall framework of the proposed ALTER.
• Figure 3: Performance comparison with varying readout functions (%).
• Figure 4: In-depth analysis of ALTER and adaptive long-range aware strategy.
• Figure 5: Example brain graphs from the ABIDE dataset and the corresponding attention heatmap.