For training medical image registration models
OFG is a general training framework that provides an alternative to weakly-supervised and unsupervised training for image registration models. By iteratively optimizing the prediction result of the trained registration model on-the-fly, OFG introduces pseudo ground truth to an unsupervised training process. This supervision provides more direct guidance towards model training compared with unsupervised methods.
OFG is a two stage training method, integrating optimization-based methods with registration models. It optimize the model's output in training time, this process generates a pseudo label on-the-fly, which will provide supervision for the model, yielding a model with better registration performance.
OFG consistently improves the registration methods it is used on, and achieves state-of-the-art performance. It has better trainability than unsupervised methods while not using any manually added labels.
Cite our work when comparing results:
@article{ofg2023,
title={On-the-Fly Guidance Training for Medical Image Registration},
author={Yicheng Chen and Shengxiang Ji and Yuelin Xin and Kun Han and Xiaohui Xie},
year={2023},
eprint={2308.15216},
archivePrefix={arXiv},
primaryClass={cs.CV}
}