Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2018 Sep:11070:739-746.
doi: 10.1007/978-3-030-00928-1_83. Epub 2018 Sep 26.

Adversarial Similarity Network for Evaluating Image Alignment in Deep Learning based Registration

Jingfan Fan  1 , Xiaohuan Cao  1   2 , Zhong Xue  3 , Pew-Thian Yap  1 , Dinggang Shen  1   2   3
Affiliations

Adversarial Similarity Network for Evaluating Image Alignment in Deep Learning based Registration

Jingfan Fan et al. Med Image Comput Comput Assist Interv. 2018 Sep.

Abstract

This paper introduces an unsupervised adversarial similarity network for image registration. Unlike existing deep learning registration frameworks, our approach does not require ground-truth deformations and specific similarity metrics. We connect a registration network and a discrimination network with a deformable transformation layer. The registration network is trained with feedback from the discrimination network, which is designed to judge whether a pair of registered images are sufficiently similar. Using adversarial training, the registration network is trained to predict deformations that are accurate enough to fool the discrimination network. Experiments on four brain MRI datasets indicate that our method yields registration performance that is promising in both accuracy and efficiency compared with state-of-the-art registration methods, including those based on deep learning.

PubMed Disclaimer

Figures

Fig. 1.
Fig. 1.
The proposed adversarial similarity network for deformable image registration. The input image pair is already linearly aligned.
Fig. 2.
Fig. 2.
The discrimination network.
Fig. 3.
Fig. 3.
Boxplot of DSC (%) in 54 ROIs for the 10 testing subjects from LPBA40 dataset, after performing registration under different training strategies: 1) supervised learning, 2) similarity metrics SSD and CC, and 3) the proposed adversarial similarity network. “+” marks improvements given by the proposed method over the three other methods.
Fig. 4.
Fig. 4.
Typical registration results from MGH10. The boxes mark significant improvements.
See this image and copyright information in PMC

References

    1. Sokooti H, et al. , Nonrigid Image Registration Using Multi-scale 3D Convolutional Neural Networks. in MICCAI, Springer; (2017)
    1. Yang X, et al. , Quicksilver: Fast predictive image registration-A deep learning approach. NeuroImage 158: p. 378–396 (2017) - PMC - PubMed
    1. Rohe M-M, et al. , SVF-Net: Learning Deformable Image Registration Using Shape Matching. in MICCAI, Springer; (2017)
    1. Li H and Fan Y, Non-Rigid Image Registration Using Self-Supervised Fully Convolutional Networks without Training Data. arXiv preprint arXiv:1801.04012 (2018) - PMC - PubMed
    1. Balakrishnan G, et al. , An Unsupervised Learning Model for Deformable Medical Image Registration. arXiv preprint arXiv:1802.02604 (2018)

LinkOut - more resources