Self-supervised 3D anatomy segmentation using self-distilled masked image transformer (SMIT). SMIT used both global, patch and pixel level information for self-supervised learning in self-distillization manner.
This is the official source code for the MICCAI 2022 paper SMIT
pip install requirements.txt
python train_self_supervised.py
python fine_tuning_swin_3D.py --resume_ckpt
We offered the pre-trained weight with imagee patch size of 96x96x96, depth= (2, 2, 4, 2), head= (4, 4, 4, 4), window size= (4,4,4).
If you find this repository useful, please consider giving a star ⭐ and citation:
@InProceedings{juejsmit,
title={Self-supervised 3D Anatomy Segmentation Using Self-distilled Masked Image Transformer (SMIT)},
author={Jiang, Jue and Tyagi, Neelam and Tringale, Kathryn and Crane, Christopher and Veeraraghavan, Harini},
journal={International Conference Medical Image Computing and Computer Assisted Intervention, 2022},
pages={556--566},
DOI={DOI: 10.1007/978-3-031-16440-8_53},
year={2022}
}