Pretraining Deformable Image Registration Networks with Random Images

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Chen, Junyu, Wei, Shuwen, Liu, Yihao, Carass, Aaron, Du, Yong
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866913866798071808
author Chen, Junyu
Wei, Shuwen
Liu, Yihao
Carass, Aaron
Du, Yong
author_facet Chen, Junyu
Wei, Shuwen
Liu, Yihao
Carass, Aaron
Du, Yong
contents Recent advances in deep learning-based medical image registration have shown that training deep neural networks~(DNNs) does not necessarily require medical images. Previous work showed that DNNs trained on randomly generated images with carefully designed noise and contrast properties can still generalize well to unseen medical data. Building on this insight, we propose using registration between random images as a proxy task for pretraining a foundation model for image registration. Empirical results show that our pretraining strategy improves registration accuracy, reduces the amount of domain-specific data needed to achieve competitive performance, and accelerates convergence during downstream training, thereby enhancing computational efficiency.
format Preprint
id arxiv_https___arxiv_org_abs_2505_24167
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Pretraining Deformable Image Registration Networks with Random Images
Chen, Junyu
Wei, Shuwen
Liu, Yihao
Carass, Aaron
Du, Yong
Computer Vision and Pattern Recognition
Recent advances in deep learning-based medical image registration have shown that training deep neural networks~(DNNs) does not necessarily require medical images. Previous work showed that DNNs trained on randomly generated images with carefully designed noise and contrast properties can still generalize well to unseen medical data. Building on this insight, we propose using registration between random images as a proxy task for pretraining a foundation model for image registration. Empirical results show that our pretraining strategy improves registration accuracy, reduces the amount of domain-specific data needed to achieve competitive performance, and accelerates convergence during downstream training, thereby enhancing computational efficiency.
title Pretraining Deformable Image Registration Networks with Random Images
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2505.24167