Correlation Ratio for Multi- and Mono-modal Image Registration

keywords: correlation ratio, image registration, multi-modal image registration

This repository hosts PyTorch implementation of Correlation Ratio for medical image registration, originally proposed in this paper. We have evaluated CR has a loss function for the application of both affine and deformable registration.

1. Affine Registration: Chen, Junyu, et al. "Unsupervised Learning of Multi-modal Affine Registration for PET/CT,” 2024 IEEE NSS/MIC
2. Deformable Registration: X. Chen, et al., "Correlation ratio for unsupervised learning of multi-modal deformable registration", Proceedings of SPIE Medical Imaging (SPIE-MI 2025), San Diego, CA, February 16–20, 2025

You can find the PyTorch implementation of the correlation ratio and local-patch-based correlation ratio here:

• correlation ratio
• local correlation ratio

PET/CT Multi-Modal Affine Registration

The source code for PET/CT affine registration can be found here, and you will need to install the required packages listed in the requirements.txt file.

Multi-scale Instance-specific Optimization

Qualitative Results

T1/T2 Brain MRI Multi-Modal Deformable Registration

CR and diffusion regularization as the loss function for unsupervised learning of multimodal deformable image registration.

Qualitative Results

Citation

If you find this code is useful in your research, please consider to cite:

Affine Registration

@misc{chen2024unsupervised,
  title={Unsupervised Learning of Multi-modal Affine Registration for PET/CT}, 
  author={Junyu Chen and Yihao Liu and Shuwen Wei and Aaron Carass and Yong Du},
  year={2024},
  eprint={2409.13863},
  archivePrefix={arXiv},
  primaryClass={eess.IV},
  url={https://arxiv.org/abs/2409.13863}, 
}

Deformable Registration

@inproceedings{chen2024correlation,
  title={Correlation ratio for unsupervised learning of multi-modal deformable registration}, 
  author={Xiaojian Chen, Yihao Liu, Shuwen Wei, Aaron Carass, Yong Du, and Junyu Chen},
  booktitle={Medical Imaging 2025: Image Processing}
  year={2025},
  organization={SPIE}
}