BrainHGT: A Hierarchical Graph Transformer for Interpretable Brain Network Analysis

2025-11-18

AI Summary

Existing brain network analyses typically model the brain as a flat graph, neglecting its intrinsic modular architecture and distance-dependent connectivity patterns—thus failing to capture hierarchical interactions among local and long-range regions, within and across functional modules. To address this, we propose a Hierarchical Graph Transformer framework: (1) a parallel short- and long-range attention mechanism explicitly models brain-region interactions across multiple spatial scales; (2) a neuroanatomically informed clustering module jointly leverages data-driven learning and biological priors for multi-scale functional module identification; and (3) cross-attention integrated with hierarchical graph structure enhances both interpretability and neurobiological plausibility. Evaluated on disease classification tasks, our method achieves significant performance gains and robustly identifies sub-functional modules, demonstrating dual advantages in predictive accuracy and mechanistic interpretability.

Technology Category

• Machine Learning: Graph-based Machine Learning
• Humans and AI: Brain-Sensing and Analysis
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding

Application Category

• Graph Algorithms and Modeling for the Web: Algorithms and analysis for heterogeneous, signed, attributed, multi-relational, temporal, higher-order, and annotated Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

Graph Transformer shows remarkable potential in brain network analysis due to its ability to model graph structures and complex node relationships. Most existing methods typically model the brain as a flat network, ignoring its modular structure, and their attention mechanisms treat all brain region connections equally, ignoring distance-related node connection patterns. However, brain information processing is a hierarchical process that involves local and long-range interactions between brain regions, interactions between regions and sub-functional modules, and interactions among functional modules themselves. This hierarchical interaction mechanism enables the brain to efficiently integrate local computations and global information flow, supporting the execution of complex cognitive functions. To address this issue, we propose BrainHGT, a hierarchical Graph Transformer that simulates the brain's natural information processing from local regions to global communities. Specifically, we design a novel long-short range attention encoder that utilizes parallel pathways to handle dense local interactions and sparse long-range connections, thereby effectively alleviating the over-globalizing issue. To further capture the brain's modular architecture, we designe a prior-guided clustering module that utilizes a cross-attention mechanism to group brain regions into functional communities and leverage neuroanatomical prior to guide the clustering process, thereby improving the biological plausibility and interpretability. Experimental results indicate that our proposed method significantly improves performance of disease identification, and can reliably capture the sub-functional modules of the brain, demonstrating its interpretability.

Problem

• Modeling hierarchical brain interactions from local to global levels
• Addressing over-globalization in graph attention mechanisms
• Capturing modular brain architecture with neuroanatomical prior guidance

Innovation

• Hierarchical Graph Transformer simulates brain information processing
• Long-short range attention encoder handles local and global connections
• Prior-guided clustering module groups regions into functional communities