RECORD
DBF-UNet: A Two-Stage Framework for Carotid Artery Segmentation with Pseudo-Label Generation
Computer Vision and Pattern Recognition · Submitted 1 Apr 2025
abstract compiled · 2026-09-23 · no language-model review
The authors' main claim Experimental validation on public datasets demonstrates that our proposed method effectively addresses the sparse annotation challenge in carotid artery segmentation while achieving superior performance compared to existing approaches.
The compiled record
The paper's own sentences, arranged. Each label is drawn from a closed vocabulary; the number beside it is the probability the classifier gave that label. Numbers in the tables are the authors', checked by code. No sentence in this record was written by a model.
Open the compiled record
The abstract, sentence by sentence
Medical image analysis faces significant challenges due to limited annotation data, particularly in three-dimensional carotid artery segmentation tasks, where existing datasets exhibit spatially discontinuous slice annotations with only a small portion of expert-labeled slices in complete 3D volumetric data. To address this challenge, we propose a two-stage segmentation framework. First, we construct continuous vessel centerlines by interpolating between annotated slice centroids and propagate labels along these centerlines to generate interpolated annotations for unlabeled slices. The slices with expert annotations are used for fine-tuning SAM-Med2D, while the interpolated labels on unlabeled slices serve as prompts to guide segmentation during inference. In the second stage, we propose a novel Dense Bidirectional Feature Fusion UNet (DBF-UNet). This lightweight architecture achieves precise segmentation of complete 3D vascular structures. The network incorporates bidirectional feature fusion in the encoder and integrates multi-scale feature aggregation with dense connectivity for effective feature reuse. Experimental validation on public datasets demonstrates that our proposed method effectively addresses the sparse annotation challenge in carotid artery segmentation while achieving superior performance compared to existing approaches. The source code is available at https://github.com/Haoxuanli-Thu/DBF-UNet.
background contribution method result interpretation main claim
| # | sentence | role | the authors' own work | states the main conclusion | strength |
|---|---|---|---|---|---|
| s0 | Medical image analysis faces significant challenges due to limited annotation data, particularly in three-dimensional carotid artery segmentation tasks, where existing datasets exhibit spatially discontinuous slice annotations with only a small portion of expert-labeled slices in complete 3D volumetric data. | background 0.94 | 0.07 | 0.03 | asserted 1.00 |
| s1 | To address this challenge, we propose a two-stage segmentation framework. | contribution 0.84 | 0.73 | 0.03 | asserted 0.86 |
| s2 | First, we construct continuous vessel centerlines by interpolating between annotated slice centroids and propagate labels along these centerlines to generate interpolated annotations for unlabeled slices. | method 0.99 | 0.95 | 0.03 | asserted 0.96 |
| s3 | The slices with expert annotations are used for fine-tuning SAM-Med2D, while the interpolated labels on unlabeled slices serve as prompts to guide segmentation during inference. | method 1.00 | 0.90 | 0.03 | asserted 0.99 |
| s4 | In the second stage, we propose a novel Dense Bidirectional Feature Fusion UNet (DBF-UNet). | method 0.61 | 0.91 | 0.03 | asserted 0.98 |
| s5 | This lightweight architecture achieves precise segmentation of complete 3D vascular structures. | method 0.47 | 0.77 | 0.11 | asserted 1.00 |
| s6 | The network incorporates bidirectional feature fusion in the encoder and integrates multi-scale feature aggregation with dense connectivity for effective feature reuse. | method 0.99 | 0.81 | 0.03 | asserted 1.00 |
| s7 | Experimental validation on public datasets demonstrates that our proposed method effectively addresses the sparse annotation challenge in carotid artery segmentation while achieving superior performance compared to existing approaches. | result 0.94 | 0.95 | 0.90 | asserted 1.00 |
| s8 | The source code is available at https://github.com/Haoxuanli-Thu/DBF-UNet. | contribution 0.56 | 0.35 | 0.02 | asserted 0.96 |
What was examined, and what was not
- arXiv listing abstract v1
- full text
- figures and tables
- references and citing sentences
- relations to other papers
- P abstract
- S sections
- T body sentences
- C citations
- R cross-paper relations
The paper
Original abstract
Medical image analysis faces significant challenges due to limited annotation data, particularly in three-dimensional carotid artery segmentation tasks, where existing datasets exhibit spatially discontinuous slice annotations with only a small portion of expert-labeled slices in complete 3D volumetric data. To address this challenge, we propose a two-stage segmentation framework. First, we construct continuous vessel centerlines by interpolating between annotated slice centroids and propagate labels along these centerlines to generate interpolated annotations for unlabeled slices. The slices with expert annotations are used for fine-tuning SAM-Med2D, while the interpolated labels on unlabeled slices serve as prompts to guide segmentation during inference. In the second stage, we propose a novel Dense Bidirectional Feature Fusion UNet (DBF-UNet). This lightweight architecture achieves precise segmentation of complete 3D vascular structures. The network incorporates bidirectional feature fusion in the encoder and integrates multi-scale feature aggregation with dense connectivity for effective feature reuse. Experimental validation on public datasets demonstrates that our proposed method effectively addresses the sparse annotation challenge in carotid artery segmentation while achieving superior performance compared to existing approaches. The source code is available at https://github.com/Haoxuanli-Thu/DBF-UNet.
Human review
How the literature uses it
What this paper does with the arXiv papers it cites
What it cites · 2 sentences without a confident label
Cites as background 2 papers
-
Subsequently, Hu et al. introduced a label propagation-based two-stage segmentation network, which initially obtains continuous 3D pseudo-labels through interpolation from limited 2D annotations, using these pseudo-labels to generate refined labels for subsequent nnUNet training [ref] .
Label Propagation for 3D Carotid Vessel Wall Segmentation and Atherosclerosis Diagnosis · Carotid Artery Segmentation · 0.89
-
To address this challenge, researchers have proposed various improved approaches: Med-SA [ref] introduces a lightweight adapter module that achieves performance improvement by updating only 2% of the parameters; SAM-Med2D [ref] adopts a more comprehensive strategy by constructing a large-scale 2D medical image dataset and fine-tuning SAM to support multiple interactive prompting modes; while for 3D medical images, SAM-Med3D [ref] abandons SAM's pre-trained weights and employs a fully 3D network architecture, achieving superior segmentation performance through a two-stage training strategy on an extensive 3D dataset (22K images and 143K masks).
SAM-Med2D · Segment Anything Model · 0.83
Discussion
Know something about this paper? Leave a comment, or highlight any passage above to comment on it.