Please use this identifier to cite or link to this item: http://bura.brunel.ac.uk/handle/2438/31901
Title: Multi-scale Lipschitz Neural Fields Incorporating Frequency Decoupling for medical image registration
Authors: Huang, J
Jiang, X
Li, X
Wu, J
Li, G
Meng, H
Li, Z
Keywords: deformable image registration;neural fields;Lipschitz continuity constraints;deep similarity prior;implicit dual-domain feature representation;frequency decoupling
Issue Date: 16-Aug-2025
Publisher: Elsevier
Citation: Huang, J. et al. (2026) 'Multi-scale Lipschitz Neural Fields Incorporating Frequency Decoupling for medical image registration', Biomedical Signal Processing and Control, 112, Multi-scale Lipschitz Neural Fields Incorporating Frequency Decoupling for medical image registration, pp. 1 - 14. doi: 10.1016/j.bspc.2025.108493.
Abstract: Deformable image registration is a pivotal task in medical image analysis, essential for applications such as surgical navigation, lesion localization, and treatment planning. However, the accuracy of registration is frequently constrained by the ability to analyze complex textural features, which impedes the capture of fine-grained deformations and elevates surgical risks. To overcome these limitations, we propose the Multi-Scale Lipschitz Neural Fields Incorporating Frequency Decoupling (MLNFFD) framework, which reformulates the registration task as the correction of high-frequency residuals based on a low-frequency coarse registration foundation. Specifically, we employ a dual-branch architecture, comprising a band-limited U-Net low-frequency branch for coarse deformation estimation, and a Multi-Scale Lipschitz Neural Fields (MLNF) high-frequency branch dedicated to refining fine-grained deformation residuals. Through a complementary joint optimization strategy, the low-frequency branch provides an initial estimation that accelerates the optimization of the high-frequency branch, while the high-frequency branch refines the deformation field and enhances local details, fostering mutual reinforcement and improving registration accuracy. Additionally, the MLNF branch incorporates a Multi-Scale Implicit Dual-Domain Feature Representation (MIDFR), a deep similarity prior, and Lipschitz continuity constraints, enhancing the representation of complex textural features while ensuring the smoothness of the deformation field. Extensive experimental results on brain MRIs and cine cardiac MRIs datasets show that our proposed approach outperforms existing methods in terms of both registration accuracy and deformation field smoothness.
Description: Data availability: Data will be made available on request.
URI: https://bura.brunel.ac.uk/handle/2438/31901
DOI: https://doi.org/10.1016/j.bspc.2025.108493
ISSN: 1746-8094
Other Identifiers: ORCiD: Jing Huang https://orcid.org/0009-0001-6644-3571
ORCiD: Xinwei Li https://orcid.org/0000-0003-0713-9366
ORCiD: Jia Wu https://orcid.org/0009-0007-6710-4137
ORCiD: Guoquan Li https://orcid.org/0000-0001-8022-743X
ORCiD: Hongying Meng https://orcid.org/0000-0002-8836-1382
ORCiD: Zhangyong Li https://orcid.org/0000-0002-3918-069X
Article number: 108493
Appears in Collections:Dept of Electronic and Electrical Engineering Research Papers

Files in This Item:
File Description SizeFormat 
FullText.pdfEmbargoed until 16 August 2026. Copyright © 2025 Elsevier Ltd. All rights reserved. This manuscript version is made available under the CC-BY-NC-ND 4.0 license https://creativecommons.org/licenses/by-nc-nd/4.0/ (see: https://www.elsevier.com/about/policies/sharing).2.49 MBAdobe PDFView/Open


This item is licensed under a Creative Commons License Creative Commons