This is the implementation for the paper:

Modality-aware Mutual Learning for Multi-modal Medical Image Segmentation

Early Accepted to MICCAI 2021

• Data Preparation

  • Download the data from MICCAI 2018 BraTS Challenge.

  • Convert the files' name by

  python dataset_conversion/Task032_BraTS_2018.py

  • Preprocess the data by

  python experiment_planning/nnUNet_plan_and_preprocess.py -t 32 --verify_dataset_integrity

• Train

  • Train the model by

  python run/run_training.py 3d_fullres MAMLTrainerV2 32 0

• Test

  • inference on the test data by

  python inference/predict_simple.py -i INPUT_PATH -o OUTPUT_PATH -t 32 -f 0 -tr MAMLTrainerV2

MAML is integrated with the out-of-box nnUNet. Please refer to it for more usage.

If you find this code and paper useful for your research, please kindly cite our paper.

@inproceedings{zhang2021modality,
  title={Modality-Aware Mutual Learning for Multi-modal Medical Image Segmentation},
  author={Zhang, Yao and Yang, Jiawei and Tian, Jiang and Shi, Zhongchao and Zhong, Cheng and Zhang, Yang and He, Zhiqiang},
  booktitle={International Conference on Medical Image Computing and Computer-Assisted Intervention},
  pages={589--599},
  year={2021},
  organization={Springer}
}

MAML is integrated with the out-of-box nnUNet.