NestedFormer (MICCAI2022) is a multimodal segmentation model on 3D medical images. The features of different modalities are fused through tri-oriented self-attention and cross-attention. We also improve the Poolfromer structure (CVPR2022) as an efficient encoder.

Read our paper https://arxiv.org/abs/2208.14876 on ArXiv for a formal introduction.

pip install monai
pip install tqdm
pip install tensorboardX

Please download the brats2020 datasets. Of course, switching to other datasets is ok.

python main.py --distributed  --logdir=log_train_nestedformer --fold=0 --json_list=./brats2020_datajson.json --max_epochs=1000 --lrschedule=warmup_cosine --val_every=10 --data_dir=/data/MICCAI_BraTS2020_TrainingData/  --out_channels=3 --batch_size=1 --infer_overlap=0.5

--data_dir is the location of the data.

The data processing code is in utils/data_utils.py. You can modify this code for your own dataset.

Our implementation is mainly based on the following codebases. We gratefully thank the authors for their wonderful works.

pytorch, monai, monai-research-contributions