Double-Mix Pseudo-Label Framework

Here is the code for our proposed Double-Mix Pseudo-Label Framework.

Data Preprocess

To prepare the dataset, you can follow the work of DHC.

You can also use the dmp/code/data/preprocess_amos.py to prepare the dataset.

The splits are available at {dataset}_data/split.

Model Training

Run

cd dmp
bash train3times_seeds_20p.sh -c 0 -t synapse -m cdifw_dmp_ours -e 'test' -l 3e-2 -w 0.1

Training Data Percentage:

The notation 20p represents training with 20% labeled data. You can modify this value to train3times_seeds_40p, train3times_seeds_5p, etc., to indicate training with 40%, 5%, and so on.

Command-line Parameters:

-c: Specifies which GPU to use for training.

-t: Defines the task, which can be either synapse or amos.

-m: Specifies the training method. The available methods include:

(i) cdifw_dmp_ours (our proposed method)

(ii)cdifw (ablation studies)

-e: Defines the name of the current experiment. default: 'test'

-l: Sets the learning rate. In this experiment, it was set to 0.1

-w: Specifies the weight of the unsupervised loss.

Have fun.

Cite

If this code is helpful for your study, welcome to cite our paper

@article{zhang2025double,
  title={Double-mix pseudo-label framework: enhancing semi-supervised segmentation on category-imbalanced CT volumes},
  author={Zhang, Luyang and Hayashi, Yuichiro and Oda, Masahiro and Mori, Kensaku},
  journal={International Journal of Computer Assisted Radiology and Surgery},
  pages={1--12},
  year={2025},
  publisher={Springer}
}

License

This repository is released under MIT License.