Ideal Registration? Segmentation is All You Need

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Chen, Xiang, Zhang, Fengting, Liu, Qinghao, Liu, Min, Wu, Kun, Wang, Yaonan, Zhang, Hang
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866911163462189056
author Chen, Xiang
Zhang, Fengting
Liu, Qinghao
Liu, Min
Wu, Kun
Wang, Yaonan
Zhang, Hang
author_facet Chen, Xiang
Zhang, Fengting
Liu, Qinghao
Liu, Min
Wu, Kun
Wang, Yaonan
Zhang, Hang
contents Deep learning has revolutionized image registration by its ability to handle diverse tasks while achieving significant speed advantages over conventional approaches. Current approaches, however, often employ globally uniform smoothness constraints that fail to accommodate the complex, regionally varying deformations characteristic of anatomical motion. To address this limitation, we propose SegReg, a Segmentation-driven Registration framework that implements anatomically adaptive regularization by exploiting region-specific deformation patterns. Our SegReg first decomposes input moving and fixed images into anatomically coherent subregions through segmentation. These localized domains are then processed by the same registration backbone to compute optimized partial deformation fields, which are subsequently integrated into a global deformation field. SegReg achieves near-perfect structural alignment (98.23% Dice on critical anatomies) using ground-truth segmentation, and outperforms existing methods by 2-12% across three clinical registration scenarios (cardiac, abdominal, and lung images) even with automatic segmentation. Our SegReg demonstrates a near-linear dependence of registration accuracy on segmentation quality, transforming the registration challenge into a segmentation problem. The source code will be released upon manuscript acceptance.
format Preprint
id arxiv_https___arxiv_org_abs_2509_15784
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Ideal Registration? Segmentation is All You Need
Chen, Xiang
Zhang, Fengting
Liu, Qinghao
Liu, Min
Wu, Kun
Wang, Yaonan
Zhang, Hang
Computer Vision and Pattern Recognition
Artificial Intelligence
Deep learning has revolutionized image registration by its ability to handle diverse tasks while achieving significant speed advantages over conventional approaches. Current approaches, however, often employ globally uniform smoothness constraints that fail to accommodate the complex, regionally varying deformations characteristic of anatomical motion. To address this limitation, we propose SegReg, a Segmentation-driven Registration framework that implements anatomically adaptive regularization by exploiting region-specific deformation patterns. Our SegReg first decomposes input moving and fixed images into anatomically coherent subregions through segmentation. These localized domains are then processed by the same registration backbone to compute optimized partial deformation fields, which are subsequently integrated into a global deformation field. SegReg achieves near-perfect structural alignment (98.23% Dice on critical anatomies) using ground-truth segmentation, and outperforms existing methods by 2-12% across three clinical registration scenarios (cardiac, abdominal, and lung images) even with automatic segmentation. Our SegReg demonstrates a near-linear dependence of registration accuracy on segmentation quality, transforming the registration challenge into a segmentation problem. The source code will be released upon manuscript acceptance.
title Ideal Registration? Segmentation is All You Need
topic Computer Vision and Pattern Recognition
Artificial Intelligence
url https://arxiv.org/abs/2509.15784