This repository is the official implementation of NOVEL OCT MOSAICKING PIPELINE WITH FEATURE- AND PIXEL-BASED REGISTRATION. A novel OCT image mosaicking pipeline that bridges the gap between feature-based and pixel-based registration. We also investiage how to apply SAM into the mosaicking evaluation pipeline.
Please check the original paper and follow the steps specified in the IEEE website.
In our experiment, we only utilized the OCTA_6M data. We utilized monai to generate the known homography/affine transform and deform displacement. By this way, we can generate a synthetic dataset with known transform to evaluate the performance of the registration result. You can also implement it by cv.Warp; we also have the openCV implementation.
Please check the full implementation synthetic_octa-500.ipynb
To install ANTs on your desktop, I highly recommend you follow the great tutorial. This is designed for MAC, but it follows a similar installation logic for other UNIX systems. Please refer to the ANTs official Github to submit any issues regarding that.
If you found this repo interesting, please check our paper
@article{wang2023novel,
title={Novel OCT mosaicking pipeline with Feature-and Pixel-based registration},
author={Wang, Jiacheng and Li, Hao and Hu, Dewei and Tao, Yuankai K and Oguz, Ipek},
journal={arXiv preprint arXiv:2311.13052},
year={2023}
}Please contact Jiacheng Wang(jiacheng.wang.1@vanderbilt.edu) if you have any questions on the implementation or leave a issue.