MBURegNet: an unsupervised model-based deep unrolling registration network for deformable image registration in dual-energy CT

Rui Liao; Peng Wang; Wenjing Cao; Yuan Bao; Jian Xu; Guotao Quan; Wenying Wang

doi:10.1117/12.3006887

1 April 2024 MBURegNet: an unsupervised model-based deep unrolling registration network for deformable image registration in dual-energy CT

Rui Liao, Peng Wang, Wenjing Cao, Yuan Bao, Jian Xu, Guotao Quan, Wenying Wang

Author Affiliations +

Proceedings Volume 12925, Medical Imaging 2024: Physics of Medical Imaging; 129254O (2024) https://doi.org/10.1117/12.3006887
Event: SPIE Medical Imaging, 2024, San Diego, California, United States

Conference Poster

Abstract

Dual-energy CT (DECT) provides additional material-based contrast using spectral information. The realization of DECT using a rotation-to-rotation kVp switching may suffer from structure misalignment due to patient’s motion and requires deformable image registration (DIR) between the two kVp images. Recent studies in DIR has highlighted deep-learning-based methods which can achieve superior registration accuracy with reasonable computational time. However, current deep-learning-based DIR methods may eliminate important anatomical features or hallucinate faked structures. The lack of interpretability complicates the robustness verification. Alternatively, recent studies have introduced the algorithm unrolling method that provides a concrete and systematic connection between model-based iterative methods and data-driven methods. In this work, we present an unsupervised Model-Based deep Unrolling Registration Network (MBURegNet) for DIR in DECT. MBURegNet comprises a sequence of stacked update blocks to unroll the Large Deformation Diffeomorphic Metric Mapping method, where each block samples the velocity field that follows the diffeomorphism physics. Preliminary studies using clinical data has shown that the proposed network can achieve superior performance improvement compared to the baseline deep-learning-based method, as evidenced by both qualitative and quantitative analyses. Additionally, the network can generate a sequence of intermediate images connecting the initial and final motion states, effectively illustrating the continuous flow of diffeomorphisms.

(2024) Published by SPIE. Downloading of the abstract is permitted for personal use only.

Citation Download Citation

Rui Liao, Peng Wang, Wenjing Cao, Yuan Bao, Jian Xu, Guotao Quan, and Wenying Wang "MBURegNet: an unsupervised model-based deep unrolling registration network for deformable image registration in dual-energy CT", Proc. SPIE 12925, Medical Imaging 2024: Physics of Medical Imaging, 129254O (1 April 2024); https://doi.org/10.1117/12.3006887

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available