OncoReg: Medical Image Registration for Oncological Challenges
Authors: Wiebke Heyer, Yannic Elser, Lennart Berkel, Xinrui Song, Xuanang Xu, Pingkun Yan, Xi Jia, Jinming Duan, Zi Li, Tony C. W. Mok, Bowen Li, Tim Hable, Christian Staackmann, Christoph Großbröhmer, Lasse Hansen, Alessa Hering, Malte M. Sieren, Mattias P. Heinrich
Organizations: Institute of Medical Informatics, University of Lübeck, Lübeck, Germany · Institute of Radiology and Nuclear Medicine, University Hospital Schleswig-Holstein, Lübeck, Germany · Department of Biomedical Engineering and Center for Biotechnology and Interdisciplinary Studies, Rensselaer Polytechnic Institute, Troy, NY, USA · School of Computer Science, University of Birmingham, Birmingham, UK · Division of Informatics, Imaging and Data Sciences, University of Manchester, UK · DAMO Academy, Alibaba Group, Hangzhou, China · Hangzhou Shengshi Technology Co., Ltd, Hangzhou, China · Department of Radiation Oncology, University Hospital Schleswig-Holstein, Lübeck, Germany · EchoScout GmbH, Lübeck, Germany · Radboud University Medical Center, Nijmegen, Netherlands · Institute of Interventional Radiology, University Hospital Schleswig-Holstein, Lübeck, Germany
In modern cancer research, the vast volume of medical data generated is often underutilised due to challenges related to patient privacy. The OncoReg Challenge addresses this issue by enabling researchers to develop and validate image registration methods through a two-phase framework that ensures patient privacy while fostering the development of more generalisable AI models. Phase one involves working with a publicly available dataset, while phase two focuses on training models on a private dataset within secure hospital networks. OncoReg builds upon the foundation established by the Learn2Reg Challenge by incorporating the registration of interventional cone-beam computed tomography with standard planning fan-beam CT images in radiotherapy. Accurate image registration is crucial in oncology, particularly for dynamic treatment adjustments in image-guided radiotherapy, where precise alignment is necessary to minimise radiation exposure to healthy tissues while effectively targeting tumours. This work details the methodology and data behind the OncoReg Challenge and provides a comprehensive analysis of the competition entries and results. Findings reveal that feature extraction plays a pivotal role in this registration task. A new method emerging from this challenge demonstrated its versatility, while established approaches continue to perform comparably to newer techniques. Both deep learning and classical approaches still play significant roles in image registration, with the combination of methods, particularly in feature extraction, proving most effective.
Figures & tables
Dataset
ThoraxCBCT
OncoReg
Modality
Fan-beam CT & Cone-beam CT
Fan-beam CT & Cone-beam CT
4D (max. Insp./Exp.)
✓
×
# Cases (2 Pairs/Case)
19 (1xFBCT, 2xCBCT per Case)
27 (1xFBCT, 2xCBCT per Case)
Split
11 Train, 3 Val., 5 Test
3-Fold CV (15 Train, 3 Val., 9 Test)
Image Size
390, 280, 300
256, 192, 192
Spacing
1 mm isotropic
1.5 mm isotropic
Table 1: Overview of the two datasets for the ThoraxCBCT and OncoReg tasks.
Figure 1: Exemplary image pair from the ThoraxCBCT dataset including semi-automatic label mask annotations. Left: Fan-Beam CT. Right: Cone-Beam CT.
Figure 2: Semi-automatic label mask annotations for an exemplary Fan-Beam CT from the OncoReg dataset. 3D rendering generated with 3D Slicer ( Fedorov et al., 2012 ).
Metric
Description
Purpose
Unit
DSC
Dice Similarity Coefficient
Overlap between segmentations
%
DSC30
30th Percentile of DSC
Measures robustness of overlap
%
HD95
95th Percentile Hausdorff Dist.
Robust surface distance
mm
TRE
Target Registration Error
Landmark misalignment
mm
TRE30
70th Percentile of TRE
Robustness of landmark error
mm
SDlogJ
Standard Deviation of log(Jac.)
Local volume changes
-
Table 2: Summary of Metrics used throughout the challenge for evaluation of registration performance.
Figure 3: SynDeeds: Fixed and moving images are registered using Deeds after the fixed image went through the image restoration module.
Figure 4: Overview of the image restoration module.
Figure 5: Exemplary ThoraxCBCT image pair before registration (left: moving FBCT; middle: fixed CBCT) and after registration with ConvexAdam (right: warped FBCT).
TRE ↓{\color[rgb]{1,1,1}a}_{mm}
TRE30 ↓{\color[rgb]{1,1,1}a}_{mm}
DSC ↑{\color[rgb]{1,1,1}a}_{\%}
HD95 ↓{\color[rgb]{1,1,1}a}_{mm}
SDlogJ ↓
Rank ↑
Initial
11.91
16.26
29.04
48.13
ConvexAdam
4.93
4.98
68.40
38.00
0.10
0.710
SynDeeds
6.01
5.67
70.16
22.97
0.15
0.705
NiftyReg
4.90
4.51
62.11
43.41
0.05
0.577
DeedsBCV
8.76
7.38
69.00
22.38
0.16
0.537
Fourier-Net
4.03
3.85
61.91
47.80
0.07
0.535
Table 3: Quantitative results of the ThoraxCBCT task.
Figure 6: ThoraxCBCT: Dice coefficients for each anatomical label and registration method, averaged across all test images.
Figure 10
TRE ↓{\color[rgb]{1,1,1}a}_{mm}
TRE30 ↓{\color[rgb]{1,1,1}a}_{mm}
DSC ↑{\color[rgb]{1,1,1}a}_{\%}
HD95 ↓{\color[rgb]{1,1,1}a}_{mm}
SDlogJ ↓
RT ↓{\color[rgb]{1,1,1}a}_{s}
Rank ↑
Initial
6.41
11.43
43.38
32.68
TimH
5.30
6.31
63.21
6.92
0.039
N/A
0.73
DINO-Reg Ens.
3.94
7.63
63.89
31.12
0.062
>300
0.67
VoxelMorph++
4.66
8.45
62.36
26.86
0.081
<60
0.60
ConvexAdam
4.16
7.89
60.13
31.03
0.066
(<5)
0.60
DINO-Reg
4.53
8.46
57.64
81.83
0.039
>300
0.57
Table 4: Quantitative Results of the OncoReg task. Runtime measurements were only recorded for submissions evaluated through the original challenge pipeline. For methods submitted after the official challenge evaluation, runtime information is not available.
Figure 9: OncoReg: Dice coefficients for ten anatomical labels and each registration method, averaged across all images.
Appendix figures & tables4 assets
Supplementary material from the paper’s appendix.
Appendix
Figure 10: OncoReg: Relationship between mean Target Registration Error and mean Dice Similarity Coefficient across all registration methods. Each point represents one method, with TRE averaged across all landmark correspondences and DSC averaged across all images and anatomical structures. A moderate negative correlation (Spearman ρ=−0.52 ) is observed.
Figure 11: OncoReg: Relationship between Dice Similarity Coefficient and 95th percentile Hausdorff Distance for five clinically relevant anatomical structures across all images and registration methods.
Figure 12: OncoReg: Dice coefficients for individual vertebrae-labels and each registration method, averaged across all images.
Figure 13: OncoReg: Dice coefficients for individual rib-labels and each registration method, averaged across all images.
Medical image registration aligns scans across time, patients, or imaging modalities to track anatomy and support diagnosis, treatment, and image-guided procedures.
Medical image registration is a fundamental task in medical image analysis, enabling the alignment of images from different modalities or time points. However, intensity inconsistencies and nonlinear tissue deformations pose significant challenges to the robustness of registration methods. Recent approaches leveraging self-supervised representation learning show promise by pre-training feature extractors to generate robust anatomical embeddings, that farther used for the registration. In this work, we propose a novel framework that integrates equivariant contrastive learning directly into the registration model. Our approach leverages the power of contrastive learning to learn robust feature representations that are invariant to tissue deformations. By jointly optimizing the contrastive and registration objectives, we ensure that the learned representations are not only informative but also suitable for the registration task. We evaluate our method on abdominal and thoracic image registration tasks, including both intra-patient and inter-patient scenarios. Experimental results demonstrate that the integration of contrastive learning directly into the registration framework significantly improves performance, surpassing strong baseline methods.
Eytan Kats, Christoph Grossbroehmer, Ziad Al-Haj Hemidi +3
Insitute of Medical Informatics, University of Luebeck, Luebeck, Germany
Multi-modal image registration plays a critical role in precision medicine but faces challenges from non-linear intensity relationships and local optima. While deep learning models enable rapid inference, they often suffer from generalization collapse on unseen modalities. To address this, we propose Search-MIND, a training-free, iterative optimization framework for instance-specific registration. Our pipeline utilizes a coarse-to-fine strategy: a hierarchical coarse alignment stage followed by deformable refinement. We introduce two novel loss functions: Variance-Weighted Mutual Information (VWMI), which prioritizes informative tissue regions to shield global alignment from background noise and uniform regions, and Search-MIND (S-MIND), which broadens the convergence basin of structural descriptors by considering larger local search range. Evaluations on CARE Liver 2025 and CHAOS Challenge datasets show that Search-MIND consistently outperforms classical baselines like ANTs and foundation model-based approaches like DINO-reg, offering superior stability across diverse modalities.
Boya Wang, Ruizhe Li, Chao Chen +1
Intelligent Modelling & Analysis Group (IMA), School of Computer Science, University of Nottingham, UK · Lab for Uncertainty in Data and Decision Making (LUCID), Nottingham Biomedical Research Centre (BRC), School of Medicine, University of Nottingham, UK
Raw image intensities are a poor surrogate for similarity across images in correspondence problems. Further, while auxiliary segmentation labels can provide semantic identification and improve matching, they can assign only coarse identities to fine-grained anatomy and often cover only a subset of structures, particularly in whole-body settings. We present an approach that instead establishes correspondence using complementary representations of anatomy and demonstrate its utility for longitudinal whole-body PET/CT registration in the Learn2Reg 2026 PSMAReg Challenge. We use our 3D foundation model, anatomix, to derive fine-grained, per-voxel semantic representations that are consistent across volumes and further employ off-the-shelf segmentation models to obtain coarse anatomical identities. Together, these representations drive a GPU-accelerated iterative registration solver (FireANTs) in a hybrid multistage framework: the first stage estimates alignment from dense multi-channel features, after which a second refines the deformation using both anatomix- and segmentation-based losses. Lastly, because registration should not introduce non-biological tumor volume changes across time, we develop lesion-specific, component-wise volume preservation schemes. Without any task-specific training, our system placed first on the PSMAReg Challenge leaderboard. Code and interactive visualizations are available at https://www.neeldey.com/reps-for-reg.
Neel Dey, Adrian V. Dalca
A. A. Martinos Center for Biomedical Imaging, Massachusetts General Hospital · Department of Radiology, Harvard Medical School · Computer Science and AI Lab, Massachusetts Institute of Technology