Wavelet-Guided Multi-Scale ConvNeXt for Unsupervised Medical Image Registration

Medical image registration is essential in clinical practices such as surgical navigation and image-guided diagnosis. The Transformer architecture of TransMorph demonstrates better accuracy in non-rigid registration tasks. However, its weaker spatial locality priors necessitate large-scale training...

Full description

Saved in:
Bibliographic Details
Main Authors: Xuejun Zhang, Aobo Xu, Ganxin Ouyang, Zhengrong Xu, Shaofei Shen, Wenkang Chen, Mingxian Liang, Guiqi Zhang, Jiashun Wei, Xiangrong Zhou, Dongbo Wu
Format: Article
Language:English
Published: MDPI AG 2025-04-01
Series:Bioengineering
Subjects:
Online Access:https://www.mdpi.com/2306-5354/12/4/406
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850183764574994432
author Xuejun Zhang
Aobo Xu
Ganxin Ouyang
Zhengrong Xu
Shaofei Shen
Wenkang Chen
Mingxian Liang
Guiqi Zhang
Jiashun Wei
Xiangrong Zhou
Dongbo Wu
author_facet Xuejun Zhang
Aobo Xu
Ganxin Ouyang
Zhengrong Xu
Shaofei Shen
Wenkang Chen
Mingxian Liang
Guiqi Zhang
Jiashun Wei
Xiangrong Zhou
Dongbo Wu
author_sort Xuejun Zhang
collection DOAJ
description Medical image registration is essential in clinical practices such as surgical navigation and image-guided diagnosis. The Transformer architecture of TransMorph demonstrates better accuracy in non-rigid registration tasks. However, its weaker spatial locality priors necessitate large-scale training datasets and a heavy number of parameters, which conflict with the limited annotated data and real-time demands of clinical workflows. Moreover, traditional downsampling and upsampling always degrade high-frequency anatomical features such as tissue boundaries or small lesions. We proposed WaveMorph, a wavelet-guided multi-scale ConvNeXt method for unsupervised medical image registration. A novel multi-scale wavelet feature fusion downsampling module is proposed by integrating the ConvNeXt architecture with Haar wavelet lossless decomposition to extract and fuse features from eight frequency sub-images using multi-scale convolution kernels. Additionally, a lightweight dynamic upsampling module is introduced in the decoder to reconstruct fine-grained anatomical structures. WaveMorph integrates the inductive bias of CNNs with the advantages of Transformers, effectively mitigating topological distortions caused by spatial information loss while supporting real-time inference. In both atlas-to-patient (IXI) and inter-patient (OASIS) registration tasks, WaveMorph demonstrates state-of-the-art performance, achieving Dice scores of 0.779 ± 0.015 and 0.824 ± 0.021, respectively, and real-time inference (0.072 s/image), validating the effectiveness of our model in medical image registration.
format Article
id doaj-art-09dca57a363d43d7afd6f4c6782c4f24
institution OA Journals
issn 2306-5354
language English
publishDate 2025-04-01
publisher MDPI AG
record_format Article
series Bioengineering
spelling doaj-art-09dca57a363d43d7afd6f4c6782c4f242025-08-20T02:17:14ZengMDPI AGBioengineering2306-53542025-04-0112440610.3390/bioengineering12040406Wavelet-Guided Multi-Scale ConvNeXt for Unsupervised Medical Image RegistrationXuejun Zhang0Aobo Xu1Ganxin Ouyang2Zhengrong Xu3Shaofei Shen4Wenkang Chen5Mingxian Liang6Guiqi Zhang7Jiashun Wei8Xiangrong Zhou9Dongbo Wu10School of Computer, Electronics and Information, Guangxi University, Nanning 530004, ChinaSchool of Computer, Electronics and Information, Guangxi University, Nanning 530004, ChinaDepartment of Electrical, Electronic and Computer Engineering, Gifu University, Gifu 501-1193, JapanSchool of Computer, Electronics and Information, Guangxi University, Nanning 530004, ChinaSchool of Computer, Electronics and Information, Guangxi University, Nanning 530004, ChinaSchool of Computer, Electronics and Information, Guangxi University, Nanning 530004, ChinaSchool of Computer, Electronics and Information, Guangxi University, Nanning 530004, ChinaDepartment of General Surgery, The Fourth Affiliated Hospital of Guangxi Medical University, Liuzhou 545005, ChinaDepartment of General Surgery, The Fourth Affiliated Hospital of Guangxi Medical University, Liuzhou 545005, ChinaDepartment of Electrical, Electronic and Computer Engineering, Gifu University, Gifu 501-1193, JapanDepartment of General Surgery, The Fourth Affiliated Hospital of Guangxi Medical University, Liuzhou 545005, ChinaMedical image registration is essential in clinical practices such as surgical navigation and image-guided diagnosis. The Transformer architecture of TransMorph demonstrates better accuracy in non-rigid registration tasks. However, its weaker spatial locality priors necessitate large-scale training datasets and a heavy number of parameters, which conflict with the limited annotated data and real-time demands of clinical workflows. Moreover, traditional downsampling and upsampling always degrade high-frequency anatomical features such as tissue boundaries or small lesions. We proposed WaveMorph, a wavelet-guided multi-scale ConvNeXt method for unsupervised medical image registration. A novel multi-scale wavelet feature fusion downsampling module is proposed by integrating the ConvNeXt architecture with Haar wavelet lossless decomposition to extract and fuse features from eight frequency sub-images using multi-scale convolution kernels. Additionally, a lightweight dynamic upsampling module is introduced in the decoder to reconstruct fine-grained anatomical structures. WaveMorph integrates the inductive bias of CNNs with the advantages of Transformers, effectively mitigating topological distortions caused by spatial information loss while supporting real-time inference. In both atlas-to-patient (IXI) and inter-patient (OASIS) registration tasks, WaveMorph demonstrates state-of-the-art performance, achieving Dice scores of 0.779 ± 0.015 and 0.824 ± 0.021, respectively, and real-time inference (0.072 s/image), validating the effectiveness of our model in medical image registration.https://www.mdpi.com/2306-5354/12/4/406non-rigid medical image registrationunsupervised deep learningHaar waveletConvNeXt
spellingShingle Xuejun Zhang
Aobo Xu
Ganxin Ouyang
Zhengrong Xu
Shaofei Shen
Wenkang Chen
Mingxian Liang
Guiqi Zhang
Jiashun Wei
Xiangrong Zhou
Dongbo Wu
Wavelet-Guided Multi-Scale ConvNeXt for Unsupervised Medical Image Registration
Bioengineering
non-rigid medical image registration
unsupervised deep learning
Haar wavelet
ConvNeXt
title Wavelet-Guided Multi-Scale ConvNeXt for Unsupervised Medical Image Registration
title_full Wavelet-Guided Multi-Scale ConvNeXt for Unsupervised Medical Image Registration
title_fullStr Wavelet-Guided Multi-Scale ConvNeXt for Unsupervised Medical Image Registration
title_full_unstemmed Wavelet-Guided Multi-Scale ConvNeXt for Unsupervised Medical Image Registration
title_short Wavelet-Guided Multi-Scale ConvNeXt for Unsupervised Medical Image Registration
title_sort wavelet guided multi scale convnext for unsupervised medical image registration
topic non-rigid medical image registration
unsupervised deep learning
Haar wavelet
ConvNeXt
url https://www.mdpi.com/2306-5354/12/4/406
work_keys_str_mv AT xuejunzhang waveletguidedmultiscaleconvnextforunsupervisedmedicalimageregistration
AT aoboxu waveletguidedmultiscaleconvnextforunsupervisedmedicalimageregistration
AT ganxinouyang waveletguidedmultiscaleconvnextforunsupervisedmedicalimageregistration
AT zhengrongxu waveletguidedmultiscaleconvnextforunsupervisedmedicalimageregistration
AT shaofeishen waveletguidedmultiscaleconvnextforunsupervisedmedicalimageregistration
AT wenkangchen waveletguidedmultiscaleconvnextforunsupervisedmedicalimageregistration
AT mingxianliang waveletguidedmultiscaleconvnextforunsupervisedmedicalimageregistration
AT guiqizhang waveletguidedmultiscaleconvnextforunsupervisedmedicalimageregistration
AT jiashunwei waveletguidedmultiscaleconvnextforunsupervisedmedicalimageregistration
AT xiangrongzhou waveletguidedmultiscaleconvnextforunsupervisedmedicalimageregistration
AT dongbowu waveletguidedmultiscaleconvnextforunsupervisedmedicalimageregistration