Overview

Figure 1. The framework of the proposed DHGFormer.

Abstract - The functional brain network exhibits a hierarchical characterized organization, balancing localized specialization with global integration through multi-scale hierarchical connectivity. While graph-based methods have advanced brain network analysis, conventional graph neural networks (GNNs) face interpretational limitations when modeling functional connectivity (FC) that encodes excitatory/inhibitory distinctions, often resorting to oversimplified edge weight transformations. Existing methods usually inadequately represent the brain's hierarchical organization, potentially missing critical information about multi-scale feature interactions. To address these limitations, we propose a novel brain network generation and analysis approach--Dynamic Hierarchical Graph Transformer (DHGFormer). Specifically, our method introduces an FC-inspired dynamic attention mechanism that adaptively encodes brain excitatory/inhibitory connectivity patterns into transformer-based representations, enabling dynamic adjustment of the functional brain network. Furthermore, we design hierarchical GNNs that consider prior functional subnetwork knowledge to capture intra-subnetwork homogeneity and inter-subnetwork heterogeneity, thereby enhancing GNN performance in brain disease diagnosis tasks. Extensive experiments on the ABIDE and ADNI datasets demonstrate that DHGFormer consistently outperforms state-of-the-art methods in diagnosing neurological disorders.

Get Started

1. Data Preparation

Download the ABIDE dataset from here.

2. Usage

Run the following command to train the model.

python main.py --config_filename setting/abide_DHGFormer.yaml

Cite our work

@inproceedings{xue2025dhgformer,
  title={DHGFormer: Dynamic Hierarchical Graph Transformer for Disorder Brain Disease Diagnosis},
  author={Xue, Rundong and Hu, Hao and Zhang, Zeyu and Han, Xiangmin and Wang, Juan and Gao, Yue and Du, Shaoyi},
  booktitle={International Conference on Medical Image Computing and Computer-Assisted Intervention},
  pages={268--278},
  year={2025},
  organization={Springer}
}

License

The source code is free for research and education use only. Any comercial use should get formal permission first.

This repo benefits from FBNETGEN. Thanks for their wonderful works.