Usformer: A Small Network for Left Atrium Segmentation of 3D LGE MRI

More details are presented in the following papers, Video, and Slides:

If you find our work is useful in your research, please consider citing:

(1) Usformer: A Light Neural Network for Left Atrium Segmentation of 3D LGE MRI

@inproceedings{lin2023usformer,
  title={Usformer: A Light Neural Network for Left Atrium Segmentation of 3D LGE MRI},
  author={Lin, Hui and Tapia, Santiago Lopez and Schiffers, Florian and Wu, Yunan and Yang, Huili and Iakovlev, Nikolay and Allen, Bradley D and Avery, Ryan and Lee, Daniel C and Kim, Daniel and others},
  booktitle={2023 31st European Signal Processing Conference (EUSIPCO)},
  pages={995--999},
  year={2023},
  organization={IEEE}
}

(2) Usformer: A Small Network for Left Atrium Segmentation of 3D LGE MRI

@article{lin2024usformer,
  title={Usformer: A Small Network for Left Atrium Segmentation of 3D LGE MRI},
  author={Lin, Hui and L{\'o}pez-Tapia, Santiago and Schiffers, Florian and Wu, Yunan and Gunasekaran, Suvai and Hwang, Julia and Bishara, Dima and Kholmovski, Eugene and Elbaz, Mohammed and Passman, Rod S and others},
  journal={Heliyon},
  publisher={Elsevier}
}

Dataset

To obtain the 2018 Atria Segmentation challenge dataset

Instructions

• [Installation]

git clone https://github.com/HuiLin0220/Usformer.git
cd Usformer
pip install -e.

• Usformer's architecture, plan, and weight (Google drive).

• [Test]

  Challenge dataset's configuration plan

  python ./test.py

Median Performance

A case with median performance in terms of Dice scores in the challenge dataset.

Green lines: prediction

Red lines: groundtruth

Yellow: slice index

More results in two datasets: challenge dataset NU dataset

References and Acknowledgements:

Usformer is developed on the nnU-Net framework. The Left Atrium Segmentation project is funded by the American Heart Association and the National Institutes of Health. Any questions, please email [email protected]