SegSTRONG-C: Segmenting Surgical Tools Robustly On Non-adversarial Generated Corruptions

This repo hosts the code for implementing the baseline algorithms for SegSTRONG-C.

This challenge is originated from:

CaRTS: Causality-driven Robot Tool Segmentation from Vision and Kinematics Data,
Hao Ding, Jintan Zhang, Peter Kazanzides, Jie Ying Wu, Mathias Unberath Proc. MICCAI, 2022
arXiv preprint (arXiv 2203.09475)

Rethinking Causality-driven Robot Tool Segmentation with Temporal Constraints, Hao Ding, Jie Ying Wu, Zhaoshuo Li, Mathias Unberath Int J CARS 18, 1009–1016 (2023) arXiv preprint (arXiv 2203.09475)

SegSTRONG-C: Segmenting Surgical Tools Robustly On Non-adversarial Generated Corruptions -- An EndoVis'24 Challenge, Hao Ding, Tuxun Lu, Yuqian Zhang, Ruixing Liang, Hongchao Shu, Lalithkumar Seenivasan, Yonghao Long, Qi Dou, Cong Gao, Mathias Unberath 2024 arXiv preprint (arXiv 2203.09475)

Installation

We provided docker for easy installation, the environment can be easily set up via:

cd docker
docker build ./ -t segstrongc:latest
docker run --rm -v "LOCAL_DATADIR":/workspace/data --gpus='"device={GPU_IDS}"' -it segstrongc:latest

Usage

We only used one GPU for training and inference so we haven't implement multi-gpu version.

To run training, find the right name for the config you want in the file:

python train.py --config CONFIG_FILENAME

for example:

python train.py --config UNet_SegSTRONGC

To run inference on validation set, give the name of the config and the path to the checkpoint file for networks to load:

python validate.py --config CONFIG_FILENAME --model_path CHECKPOINT_PATH --domain DOMAIN_NAME

for example:

python validate.py --config UNet_SegSTRONGC --model_path checkpoints/unet_segstrongc/model_39.pth --domain regular

The final test will be on test set(for example):

python validate.py --config UNet_SegSTRONGC --model_path checkpoints/unet_segstrongc/model_39.pth --test True --domain smoke --save_dir /workspace/data/SegSTRONG-C/results/smoke

Dataset preparation:

Please refer to our (website) for registration and data downloading

Citations

Please consider citing our papers in your publications if this repo helps you.

@inproceedings{ding2022carts,
  title={CaRTS: Causality-Driven Robot Tool Segmentation from Vision and Kinematics Data},
  author={Ding, Hao and Zhang, Jintan and Kazanzides, Peter and Wu, Jie Ying and Unberath, Mathias},
  booktitle={International Conference on Medical Image Computing and Computer-Assisted Intervention},
  pages={387--398},
  year={2022},
  organization={Springer}
}

@article{Ding2022RethinkingCR,
  title={Rethinking causality-driven robot tool segmentation with temporal constraints},
  author={Hao Ding and Jie Ying Wu and Zhaoshuo Li and M. Unberath},
  journal={International Journal of Computer Assisted Radiology and Surgery},
  year={2022},
  pages={1009 - 1016},
}

@misc{ding2024segstrongcsegmentingsurgicaltools,
      title={SegSTRONG-C: Segmenting Surgical Tools Robustly On Non-adversarial Generated Corruptions -- An EndoVis'24 Challenge}, 
      author={Hao Ding and Tuxun Lu and Yuqian Zhang and Ruixing Liang and Hongchao Shu and Lalithkumar Seenivasan and Yonghao Long and Qi Dou and Cong Gao and Mathias Unberath},
      year={2024},
      eprint={2407.11906},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2407.11906}, 
}

License

For academic use, this project is licensed under the 2-clause BSD License - see the LICENSE file for details. For commercial use, please contact Hao Ding (email) and Mathias Unberath(email）