nnUNet_MSWAL

nnU-Net models for MSWAL lesion segmentation.

This repository contains nnU-Net models trained on the MSWAL dataset for 1000 and 4000 epochs.

Available model directories:

• nnUNetTrainer__nnUNetResEncUNetLPlans__3d_fullres
• nnUNetTrainer_4000epochs__nnUNetResEncUNetLPlans__3d_fullres

Inference

Place model in nnUNet_results directory; ensure nnU-Net environment variables are set; prediction can be run as follows:

nnUNetv2_predict \
  -i INPUT_FOLDER \
  -o OUTPUT_FOLDER \
  -d 201 \
  -c 3d_fullres \
  -f 0 1 2 3 4 \
  # use nnUNetTrainer_4000epochs for the 4000-epoch model
  -tr nnUNetTrainer \
  -p nnUNetResEncUNetLPlans

Reference

Please cite the original MSWAL work and refer to the official project resources.

@inproceedings{wu2025mswal,
  title={Mswal: 3d multi-class segmentation of whole abdominal lesions dataset},
  author={Wu, Zhaodong and Zhao, Qiaochu and Hu, Ming and Li, Yulong and Xue, Haochen and Jiang, Zhengyong and Stefanidis, Angelos and Wang, Qiufeng and Razzak, Imran and Ge, Zongyuan and others},
  booktitle={International Conference on Medical Image Computing and Computer-Assisted Intervention},
  pages={378--388},
  year={2025},
  organization={Springer}
}

Official MSWAL repository: https://github.com/haochen-MBZUAI/MSWAL-