MUIS

Keywords: Unsupervised Segmentation, OCT, Scale-invariant.

Overview

Source code of "Multiscale Unsupervised Retinal Edema Area Segmentation in OCT Images (MICCAI2022)".

Make Dataset

1. Download ai_challenger_fl2018 dataset from ai challenge.
2. Run below

cd pprocess
# change the varibale `adjust_to` and `new_size` to 96
# and get the dataset with image size 96x96
python make_edema.py  
# split the train-test
python split_edema.py

# change the varibale `adjust_to` and `new_size` to 256
# and get the dataset with image size 256x256
python make_edema.py  
# reuse the split file in Edema96 dataset

Train

./train.sh

Thanks

• DCCS

Citation

@inproceedings{yuan2022multiscale,
  title={Multiscale Unsupervised Retinal Edema Area Segmentation in OCT Images},
  author={Yuan, Wenguang and Lu, Donghuan and Wei, Dong and Ning, Munan and Zheng, Yefeng},
  booktitle={International Conference on Medical Image Computing and Computer-Assisted Intervention},
  pages={667--676},
  year={2022},
  organization={Springer}
}