Code for OCT2Confocal: 3D CycleGAN based Translation of Retinal OCT Images to Confocal Microscopy (ISBI 2024)
Xin Tian, Nantheera Anantrasirichai, Lindsay Nicholson, and Alin Achim
The dataset is organized within the dataset
directory, containing two subfolders:
trainA
: OCT imagestrainB
: Confocal imagesEnsure the images are pre-processed as per the specifications outlined in the paper for optimal results.
Install the necessary dependencies with the following command:
conda create --name 3dcyclegan python=3.10.4
pip install torch==2.0.1+cu117 torchvision==0.15.2+cu117 torchaudio==2.0.2+cu117 -f https://download.pytorch.org/whl/torch_stable.html
To train the 3DCycleGAN model, execute:
python train.py --dataroot './dataset/Depth11' --name train11 --model cycle_gan --n_epochs 200 --n_epochs_decay 200 --save_epoch_freq 20 --load_size 232 --crop_size 212 --lr 0.00002
To test the 3DCycleGAN model, execute:
python test.py --dataroot './dataset/Depth11' --name train11 --model cycle_gan --load_size 212 --epoch latest
For customized training settings or modifications to the network architecture, refer to the detailed documentation.
There are 22 OCT images acquired in the same manner as the primary dataset, but without corresponding confocal matches. These images are for evaluating model performance and can be further used to advance multimodal image analysis.
The OCT2Confocal dataset can be applied to:
The full dataset will be released upon the publication of our paper. This release will allow researchers and practitioners full access to the dataset for their studies and applications.
All mice experiments were approved by the local Animal Welfare and Ethical Review Board (Bristol AWERB), and were conducted under a Home Office Project Licence.
To request early access, please email xin.tian@bristol.ac.uk. The download link will be shared post-submission.
If you use OCT2Confocal in your research, please cite:
@inproceedings{tian2024oct2confocal,
title={OCT2Confocal: 3D CycleGAN based Translation of Retinal OCT Images to Confocal Microscopy},
author={Tian, Xin and Anantrasirichai, Nantheera and Nicholson, Lindsay and Achim, Alin},
booktitle={IEEE International Symposium on Biomedical Imaging (ISBI)},
year={2024},
url={https://arxiv.org/abs/2311.10902}
}