MM-HiFuse

This repo. is the official implementation of MM-HiFuse: Multi-modal Multi-task Hierarchical Feature Fusion for Esophageal Cancer Staging and Differentiation Classification
Authors: Xiangzuo Huo, Shengwei Tian, Long Yu, Wendong Zhang, Aolun Li, Qimeng Yang & Jinmiao Song.
Enjoy the code and find its convenience to produce more awesome works!

Overview

Multi-modal Stem

MHF Block

Task of MM-HiFuse

Run

0. Requirements:

• python3
• pytorch 1.10
• torchvision 0.11.1

1. Training:

• Prepare the required images and store them in categories, set up training image folders and validation image folders respectively
• Run python train.py

2. Resume training:

• Modify parser.add_argument('--RESUME', type=bool, default=True) in python train.py
• Run python train.py

TensorBoard

Run tensorboard --logdir runs --port 6006 to view training progress

Reference

Some of the codes in this repo are borrowed from:

• HiFuse
• Swin Transformer
• ConvNeXt

Citation

If you find our paper/code is helpful, please consider citing:

@article{huo2025mm,
  title={MM-HiFuse: multi-modal multi-task hierarchical feature fusion for esophagus cancer staging and differentiation classification},
  author={Huo, Xiangzuo and Tian, Shengwei and Yu, Long and Zhang, Wendong and Li, Aolun and Yang, Qimeng and Song, Jinmiao},
  journal={Complex \& Intelligent Systems},
  volume={11},
  number={1},
  pages={1--12},
  year={2025},
  publisher={Springer}
}
@article{huo2024hifuse,
  title={HiFuse: Hierarchical multi-scale feature fusion network for medical image classification},
  author={Huo, Xiangzuo and Sun, Gang and Tian, Shengwei and Wang, Yan and Yu, Long and Long, Jun and Zhang, Wendong and Li, Aolun},
  journal={Biomedical Signal Processing and Control},
  volume={87},
  pages={105534},
  year={2024},
  publisher={Elsevier}
}