Accurate 3D segmentation of cone-beam CT (CBCT) is critical for interventional and radiation therapy applications, yet it remains limited by two compounding challenges: the scarcity of annotated CBCT data and the large domain shift from diagnostic CT. Interventional CBCT exhibits fundamental modality differences from conventional CT, driven by acquisition and physics effects as well as contrast-specific vascular content, thereby limiting effective cross-modality model transfer. We propose a novel unsupervised domain adaptation (UDA) framework based on redundancy-reducing feature alignment, enabling 3D CBCT segmentation with no target-domain annotations or inference-time adaptation. Our framework is architecture-agnostic, seamlessly adapting both CNN-based and ViT-based foundation models. We evaluate our method on two challenging CT-CBCT liver segmentation benchmarks: one for interventional vascular procedures and one for radiation therapy, demonstrating that even large-scale pretrained segmentation networks require explicit feature-space bridging to generalize across acquisition modalities, and that our approach consistently outperforms existing pretrained foundation model and UDA strategies. To support reproducibility and benchmarking, we release the liver segmentations for a public CBCT dataset, along with the code, trained models, and weights.
Figures & tables
Figure 1: Overview of the proposed UDA framework. Independently of the underlying backbone architecture, which can be either CNN- or ViT-based, the original network is decomposed into a shared feature extractor ψ , a representation head f , its adversarial counterpart f′ (i.e., a copy of f at initialization), and a task-specific prediction head g . Given source-domain inputs xS (3D CT patches) and target-domain inputs xT (3D CBCT patches), the extracted features z=ψ(x) are encouraged to become domain-invariant through a redundancy-reduction adversarial strategy acting between f and f′ . The prediction head g is kept separate to allow adaptation to the target task, while f′ is discarded at inference time.
Type
Method
Year
Mode
F1 ↑
Σp (M) ↓
DR
DI
Source Only
nnUNet [ 2 ]
66.5
80.1
Unsupervised Domain Adaptation
DA-nnUNet [ 4 ]
2024
Auto
73.3
84.6
6.62
MDD-UNet [ 5 ]
2023
Auto
66.7
80.6
0.09
SIFA-3D [ 6 ]
2023
Auto
55.2
64.7
37.69
MAPSeg [ 7 ]
2024
Auto
59.9
70.2
210.96
Table 1: F1 performance evaluation of UDA strategies and zero-shot foundation models for CBCT liver segmentation on two datasets. Source: CT, Target: radiation therapy CBCT (DR) or interventional CBCT (DI). Best results (per category) in bold . Σp (M) denotes the total number of additional trainable parameters introduced by the UDA method (in millions).
Figure 4: Qualitative comparison of liver segmentation on the DI (top two rows) and DR (bottom row) test sets. Left to right: input image, ground truth (GT), source-only model, MedSAM2 (zero-shot, 10 prompts), VISTA-3D (zero-shot, 10 prompts), and our UDA-adapted model. Predicted masks are shown in blue.
In interventional radiology, Cone-Beam Computed Tomography (CBCT) is a helpful imaging modality that provides guidance to practicians during minimally invasive procedures. CBCT differs from traditional Computed Tomography (CT) due to its limited reconstructed field of view, specific artefacts, and the intra-arterial administration of contrast medium. While CT benefits from abundant publicly available annotated datasets, interventional CBCT data remain scarce and largely unannotated, with existing datasets focused primarily on radiotherapy applications. To address this limitation, we leverage a proprietary collection of unannotated interventional CBCT scans in conjunction with annotated CT data, employing domain adaptation techniques to bridge the modality gap and enhance liver segmentation performance on CBCT. We propose a novel unsupervised domain adaptation (UDA) framework based on the formalism of Margin Disparity Discrepancy (MDD), which improves target domain performance through a reformulation of the original MDD optimization framework. Experimental results on CT and CBCT datasets for liver segmentation demonstrate that our method achieves state-of-the-art performance in UDA, as well as in the few-shot setting.
Gauthier Miralles, Loïc Le Folgoc, Vincent Jugnon +1
LTCI, Télécom Paris, Institut Polytechnique de Paris, Palaiseau, France · GE Healthcare, Buc, France
State-of-the-art vessel segmentation methods typically require large-scale annotated datasets and suffer from severe performance degradation under domain shifts. In clinical practice, however, acquiring extensive annotations for every new scanner or protocol is unfeasible. To address this, we propose VesselBridge3D, a foundation model adaptation framework that bridges frozen vision foundation models and volumetric vessel segmentation through lightweight 3D adaptation modules. The framework combines a lightweight 3D Adapter, a multi-scale 3D Aggregator, and Z-channel embedding for efficient adaptation to volumetric medical images. We instantiate VesselBridge3D with three frozen foundation encoders (DINOv3, MedSAM, and MedGemma) and evaluate it on the TopCoW (ID) and Lausanne (OOD) datasets. In the extreme low-data regime with 5 training samples, our method achieved a Dice score of 43.42%, marking a 30% relative improvement over the state-of-the-art nnU-Net (33.41%) and outperforming other Transformer-based baselines by up to 45%. The proposed framework was effective across all evaluated frozen foundation encoders, with DINOv3 yielding the best performance in the most label-efficient settings. Furthermore, in the out-of-distribution setting, our model demonstrated superior robustness, achieving a 50% relative improvement over nnU-Net (21.37% vs. 14.22%), which suffered from severe domain overfitting. Ablation studies confirmed the effectiveness of the proposed 3D adaptation modules. Our results demonstrate that VesselBridge3D is an effective framework for label-efficient 3D vessel segmentation under data scarcity and domain shifts.
Kirato Yoshihara, Yohei Sugawara, Yuta Tokuoka +1
The University of Osaka · Preferred Networks, Inc.
3D unsupervised domain adaptive (UDA) segmentation mitigates the high cost of manual annotations of the new domain data. Self-training has emerged as the dominant approach in this area, where its success heavily depends on a well-initialized warm-up model to generate reliable pseudo labels. However, existing methods often depend on source supervision or output-level adversarial alignment to obtain the warm-up model, which suffer from limited generalization and training instability due to the large domain gap between domains. Constructing domain-similar representations is an effective way to bridge this gap. In this work, we propose CVKD-UDA, which revisits voxel size as a core design factor to construct domain-similar representations and leverages cross-view complementary cues to balance transferability and discriminability of the warm-up model. First, we generate two complementary views by varying voxel sizes and introduce a cross-view knowledge distillation (CVKD) to enhance generalization and target perception of the model. Second, to balance transferability and discriminability, we design a lightweight Decouple-Adapter and an auxiliary imitation classifier to decouple cross-view knowledge transfer. Extensive experiments on two benchmarks demonstrate that CVKD-UDA effectively improves the performance of self-training methods and provides a new perspective for 3D UDA segmentation. Our code will be available at GitHub.
Zhimin Yuan, Ming Cheng, Shangshu Yu +4
School of Artificial Intelligence, Nanyang Normal University, Nanyang 473061, China · Fujian Key Laboratory of Sensing and Computing for Smart Cities, School of Informatics, Xiamen University, Xiamen 361005, China · School of Computer Science and Engineering, Northeastern University, Shenyang 110819, China