Self-Supervised Bulk Motion Artifact Removal in Optical Coherence Tomography Angiography

Jiaxiang Ren¹, Kicheon Park², Yingtian Pan², Haibin Ling¹

¹Department of Computer Science, ²Department of Biomedical Engineering

Stony Brook University

This repository is the official Keras implementation of Content-Aware BMA Removal model (CABR). Paper

Environment:

tensorflow == 1.14.0
keras == 2.3.1

Inference

Ensure the trained model weights is in model_weights/bestmodel.hdf5. Then run

python -u inference.py

DICE score will be printed after inference.

Find the predicted mask and enhanced image in

.
└── dataset
    ├── ...
    ├── AwakeOCA_mask_pred_DiceXXXX.tif # Predicted mask
    └── AwakeOCA_enhanced.tif # Enhanced image

The rest of dataset used in this work is not available for now due to data policy.

Citation

@inproceedings{ren2022self,
  title={Self-Supervised Bulk Motion Artifact Removal in Optical Coherence Tomography Angiography},
  author={Ren, Jiaxiang and Park, Kicheon and Pan, Yingtian and Ling, Haibin},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  pages={20617--20625},
  year={2022}
}

Contact

jiaxren@cs.stonybrook.edu

Acknowledgment