Medical image registration is a critical component of clinical imaging workflows, enabling accurate longitudinal assessment, multi-modal data fusion, and image-guided interventions. Intensity-based approaches often struggle with interscanner variability and complex anatomical deformations, whereas feature-based methods offer improved robustness by leveraging semantically informed representations. In this work, we investigate DINO-style self-supervised pretraining directly on 3D medical imaging data, aiming to learn dense volumetric features well suited for deformable registration. We assess the resulting representations on challenging interpatient abdominal registration task across both MRI and CT modalities. Our domain-specialized pretraining outperforms the DINOv2 model trained on a large-scale collection of natural images, while requiring substantially lower computational resources at inference time. Moreover, it surpasses established registration models under out-of-domain evaluation, demonstrating the value of task-agnostic yet medical imaging-focused pretraining for robust and efficient 3D image registration.
Figures & tables
Figure 1: Zero-shot registration framework. Features from the moving and fixed images are first extracted using the pretrained transformer and then projected into a low-dimensional space using PCA. ConvexAdam leverages these semantically rich representations to estimate an accurate displacement field.
Figure 2: Comparison of registration performance across different feature extractors. The results highlight the advantage of domain-specific DINO-style pretraining over DINOv2 and MIND features.
patch size
DSC
SDlogJ
∣J∣<0
12×12×12
49.09±15.72
0.264±0.049
0.01±0.05
8×8×8
51.02±17.21
0.269±0.059
0.01±0.04
4×4×4
57.74±20.14
0.388±0.099
0.11±0.2
Table 1: Impact of token resolution on downstream registration task. Increasing token resolution yields consistent performance gains, highlighting its role in capturing fine anatomical details.
AbdomenCTCT
NAKO
method
DSC
SDlogJ
∣J∣<0
DSC
SDlogJ
∣J∣<0
Initial
25.31±6.89
-
-
29.57±14.66
-
-
MIND [ 6 ]
37.08±9.1
0.451±0.092
0.29±0.54
50.55±20.6
0.398±0.112
0.11±0.3
DINOv2 [ 8 ]
34.73±7.99
0.485±0.106
0.67±1.25
54.95±16.07
0.326±0.085
0.01±0.03
TotalRegistrator [ 9 ]
36.2
-
-
-
-
-
UniGradIcon [ 13 ]
34.1
-
0.02
-
-
-
Table 2: Registration results on the AbdomenCTCT and NAKO datasets. The findings demonstrate the benefit of domain-specific DINO-style pretraining for dense deformable registration. Reported values for TotalRegistrator and UniGradIcon are taken from their respective publications and correspond to out-of-domain evaluations, where the models were not trained on AbdomenCTCT.
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
Although DINOv3 has demonstrated remarkable semantic discrimination in natural imagery, its direct application to volumetric medical segmentation is hindered by inherent dimension and domain disparities. To resolve these issues, we propose DINO-Med3D, a two-stage progressive framework that repurpose the pre-trained DINOv3 encoder for 3D medical tasks. In the first stage, we mitigate the dimension gap by introducing a multi-slice embedding module that incorporates pseudo-3D context, while simultaneously employing a segmentation proxy task to adapt representations learned from natural scenes to the medical domain. Subsequently, we further enhance volumetric understanding by adding lightweight 3D adapters into the frozen backbone to enforce global inter-slice continuity. Finally, to compensate for the spatial information loss inherent in the embedding process, we design a parallel detail recovery stream to explicitly preserve high-frequency boundary cues. Extensive experiments on five public datasets demonstrate that our approach successfully adapts DINOv3 to the medical domain and significantly outperforms state-of-the-art baselines.
Haoyu Hu, Xiyao Ma, Shiqi Liu +4
University of Chinese Academy of Sciences, Beijing, China · Institute of Automation, Chinese Academy of Sciences, Beijing, China
3D human registration has historically been treated as a regression task, assuming a unique ground-truth alignment exists between the template and an input point cloud. In reality, acquisition noise, occlusions, and unknown soft tissue dynamics introduce inherent ambiguity into human scans. Regression-based methods consequently converge to an average prediction, often failing to represent a plausible geometry. In our work, we embrace such uncertainty by modeling the registration as a distribution of alignments. We propose ODin, which formulates registration as a 3D diffusion process that generates a point cloud aligned with the target geometry while preserving template semantics through consistent point ordering. To achieve this, ODin relies on global, local, and positional conditioning, guiding each point to its correct location. Our experiments demonstrate that such a generative formulation not only outperforms its regression-based baseline, but also establishes a new state of the art, surpassing highly engineered methods while reducing the registration time by two-thirds. Pre-trained models and code are available at https://riccardomarin.github.io/odin/.
Mattia Masiero, Ilya A. Petrov, Daniel Cremers +2
University of Tübingen, Germany · Tübingen AI Center, Germany · Technical University of Munich, Germany +2