https://github.com/imoonlab/dhgformer
DHGFormer: Dynamic Hierarchical Graph Transformer for Disorder Brain Disease Diagnosis
https://github.com/imoonlab/dhgformer
brain graph-transformer
Last synced: 10 days ago
JSON representation
DHGFormer: Dynamic Hierarchical Graph Transformer for Disorder Brain Disease Diagnosis
- Host: GitHub
- URL: https://github.com/imoonlab/dhgformer
- Owner: iMoonLab
- Created: 2025-06-20T10:02:07.000Z (9 months ago)
- Default Branch: main
- Last Pushed: 2025-06-24T03:56:19.000Z (9 months ago)
- Last Synced: 2025-06-24T04:38:07.389Z (9 months ago)
- Topics: brain, graph-transformer
- Language: Python
- Homepage:
- Size: 4.28 MB
- Stars: 2
- Watchers: 0
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
DHGFormer: Dynamic Hierarchical Graph Transformer for Disorder Brain Disease Diagnosis
Rundong Xue, Hao Hu, Zeyu Zhang, Xiangmin Han*, Juan Wang, Yue Gao, Shaoyi Du*
Accepted by _**MICCAI 2025**_
[[Paper]](https://papers.miccai.org/miccai-2025/paper/1576_paper.pdf)
## 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](https://drive.google.com/file/d/14UGsikYH_SQ-d_GvY2Um2oEHw3WNxDY3/view?usp=sharing).
### 2. Usage
Run the following command to train the model.
```bash
python main.py --config_filename setting/abide_DHGFormer.yaml
```
## Cite our work
```bibtex
@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](https://github.com/Wayfear/FBNETGEN). Thanks for their wonderful works.