Mind the Gap: Mesh-Guided Repair of Broken Vessels

Gniewosz Drwiega; Wojciech Szymanski; Marek Wodzinski

cs.CVSep 24, 2026

Mind the Gap: Mesh-Guided Repair of Broken Vessels

Authors: Gniewosz Drwiega, Wojciech Szymanski, Marek Wodzinski

Organizations: Sano – Centre for Computational Personalised Medicine International Research Foundation, Krakow, Poland · AGH University of Krakow, Krakow, Poland

Abstract

Vessel segmentation is commonly optimized as voxel-wise classification, but small local errors can strongly disrupt vascular connectivity while having little effect on overlap scores. This is particularly problematic for downstream analyses that rely on centerlines, branches, connected components, or graph structure. We propose a mesh-guided post-processing framework for repairing broken vessel segmentations produced by nnU-Net. For each predicted binary mask, a deformable template mesh is fitted to the mask surface in physical space and used as a case-specific geometric scaffold. The fitted mesh is not voxelized as the final segmentation; instead, it guides conservative reconnection of disconnected components by proposing or validating thin bridge candidates under foreground-growth constraints. We evaluated this approach in three vascular anatomies using AortaSeg24 and SEGA for the aorta, TopCoW for the Circle of Willis, and PARSE for the pulmonary arteries. Performance is measured using Dice, connected-component Dice (ccDice), and the Betti-0 number. Across these datasets, repair substantially improved connectivity while preserving overlap: Dice remained nearly unchanged, whereas ccDice increased from 0.596 to 0.992 for aorta, from 0.722 to 0.835 for TopCoW, and from 0.028 to 0.862 for PARSE. The FOMAML meta-initialization further accelerated the fitting per-case, supporting practical mesh-based repair of the vascular topology. These results suggest that explicit mesh representations can provide a useful geometric prior for correcting topological failures in otherwise accurate voxel segmentations.

Figures & tables

Figure 1: Examples of broken-vessel failure modes across vascular anatomies. Top row: reference vessel structures. Bottom row: broken nnU-Net predictions, with missing local connections highlighted. Such small local errors can fragment otherwise accurate masks.

Figure 2: Overview of the proposed mesh-guided nnU-Net repair pipeline. A CT volume is first processed by a CNN segmentation branch based on nnU-Net, producing segmented vessel sections and an initial nnU-Net mask. When the predicted mask is fragmented or topologically inconsistent, its surface is extracted and used as the target for mesh fitting. Then, a fixed template mesh is passed through a GCN deformation branch. The mesh is progressively refined over multiple deformation stages, where each stage predicts bounded vertex offsets and forwards the updated mesh to the next refinement stage. The final overfitted mesh approximates the surface geometry of the nnU-Net prediction while preserving the template mesh connectivity. This fitted mesh is then used as a geometric prior for topology-aware mask repair, producing a repaired nnU-Net mask with improved connectivity.

Dataset	Template	Edge	Lap.	Normal / Area
Aorta Unified	Cylinder, 10k	0.03125	1.25	$2.5{\times}10^{-4}$ / $2.5{\times}10^{-5}$
TopCoW	Torus-like, 10k	0.000625	0.025	$2.5{\times}10^{-4}$ / $2.5{\times}10^{-5}$
PARSE	Sphere, 20k	0.1	5.0	$1.0{\times}10^{-3}$ / $5.0{\times}10^{-4}$

Table 1: Dataset-specific mesh-fitting hyperparameters.

Parameter	Aorta Unified	TopCoW	PARSE
Repair variant	Mesh-graph path	Mesh-graph path	Endpoint + mesh validation
Artifact keep/remove (mm)	57.0 / 57.5	57.0 / 57.5	20.0 / 25.0
Bridge anchor/support	Anchor 3.0 mm	Anchor 3.0 mm	Gap 12.0 mm; support 3.0 mm; fraction 0.50
Tube radius base/min/max (mm)	1.0 / 2.0 / 5.0	0.25 / 0.35 / 1.0	0.4 / 0.7 / 1.8
Adaptive radius percentile/scale	80 / 1.0	80 / 0.7	70 / 0.8
Max added foreground fraction	0.03	0.15	0.06

Table 2: Dataset-specific mask-repair hyperparameters.

Figure 3: Qualitative examples of mesh-guided repair across vascular anatomies. The method reconnects fragmented vessel components using local mesh-guided bridges while preserving the original mask away from the repaired regions. Blue arrows indicate broken vessel regions repaired by the proposed method.

Dataset	Mask	Dice $\uparrow$	ccDice $\uparrow$	$\beta_{0}$ $\downarrow$	FB $\downarrow$
Aorta ( $n=145$ )	Raw	$0.934\pm 0.007$	$0.596\pm 0.045$	$3.35\pm 0.43$	$0.056\pm 0.009$
	Filtered	$\mathbf{0.935\pm 0.007}$	$0.695\pm 0.043$	$2.47\pm 0.30$	$\mathbf{0.044\pm 0.008}$
	Repaired	$0.934\pm 0.007$	$\mathbf{0.992\pm 0.009}$	$\mathbf{1.01\pm 0.02}$	$0.055\pm 0.009$
TopCoW ( $n=125$ )	Raw	$\mathbf{0.870\pm 0.010}$	$0.722\pm 0.043$	$2.58\pm 0.20$	$\mathbf{0.040\pm 0.006}$
	Filtered	$\mathbf{0.870\pm 0.010}$	$0.723\pm 0.043$	$2.58\pm 0.20$	$\mathbf{0.040\pm 0.006}$
	Repaired	$0.867\pm 0.010$	$\mathbf{0.835\pm 0.034}$	$\mathbf{1.02\pm 0.02}$	$0.063\pm 0.007$

Table 3: Segmentation and connectivity metrics before and after repair. Values are reported as mean

\pm

95% confidence interval. Best values within each dataset are shown in bold. FB denotes false-branch fraction.

Figure 4: Quantitative summary of repair effects and mesh-fitting convergence. Mesh-guided repair improves connectivity-oriented metrics while preserving Dice, and meta-initialization accelerates per-case mesh fitting.

Figure 5: Effect of FOMAML meta-initialization on per-case mesh-fitting convergence for 25 held-out TopCoW cases. Lines show mean Chamfer loss across cases, and shaded regions indicate

\pm

SEM. The black horizontal line indicates a matched Chamfer-distance level, allowing comparison of the number of optimization steps required by FOMAML and standard initialization to reach the same reconstruction error.

Explore similar work

May 1, 2026cs.CV

Vesselpose: Vessel Graph Reconstruction from Learned Voxel-wise Direction Vectors in 3D Vascular Images

Blood vessel segmentation and -tracing are essential tasks in many medical imaging applications. Although numerous methods exist, the prevailing segment-then-fix paradigm is fundamentally limited regarding its suitability for modeling the task of complete and topologically accurate vascular network reconstruction. Here, we propose an approach to extract topologically more accurate vascular graphs from 3D image data, building upon highly successful ideas from the related biomedical tasks of cell segmentation and -tracking. Our approach first predicts voxel-wise vessel direction vectors joint with standard vessel segmentation masks. Second, to extract the vascular graph from these predictions, we introduce a direction-vector-guided extension of the TEASAR algorithm. Our approach achieves state-of-the-art performance on three benchmark datasets, spanning both synthetic and real imagery. We further demonstrate the applicability of our approach to challenging 3D micro-CT scans of rat heart vasculature. Finally, we propose meaningful and interpretable measures of topological error, namely false splits and false merges for graphs. Overall, our approach substantially improves the topological accuracy of reconstructed vascular graphs, being able to separate closely apposed vessel segments and handle multiple vascular trees within a single volume.

Rajalakshmi Palaniappan, Christoph Karg, Nemesio Navarro-Arambula +4

Max-Delbrueck-Center for Molecular Medicine in the Helmholtz Association (MDC) · Humboldt University of Berlin · University of Potsdam +4

Vessel Segmentation 3D Medical Image Segmentation Voxel Vessels Topology

Ask PDF

Jun 23, 2026cs.CV

MorVess: Morphology-Aware Pulmonary Vessel Segmentation Network

Accurate pulmonary vessel segmentation remains challenging due to the sparse, tortuous, and multi-scale nature of vascular structures, where small branches are easily lost and topology integrity is difficult to preserve under voxel-wise supervision. Existing deep segmentation models primarily optimize binary masks, lacking explicit geometric constraints, thus struggling to recover continuous tubular morphology and fine vascular connectivity. In this study, we introduce MorVess, a morphology-aware segmentation framework that integrates differentiable geometric priors with large-scale foundation model adaptation to achieve fine-grained vascular parsing. MorVess jointly predicts vessel masks, distance maps, and thickness maps, providing explicit supervision for vascular boundaries, centerline consistency, and smooth diameter transitions. A lightweight 2.5D adapter bridges 3D spatial context and 2D SAM representations, while a global-local fusion block aggregates multi-level semantics and geometric cues for high-fidelity topology reconstruction. Across two challenging pulmonary CT benchmarks, MorVess delivers superior Dice, clDice, and HD95 scores, substantially improving small-vessel recovery and global connectivity. These results demonstrate that embedding geometric intelligence into pretrained vision models offers a principled and scalable pathway toward precise vessel analysis and clinically reliable structural quantification. Our source code is available at https://github.com/MaoFuyou/MorVess.

Fuyou Mao, Yifei Chen, Beining Wu +9

Central South University, 410083 Changsha, China · Tsinghua University, 100084 Beijing, China · Hangzhou Dianzi University, 310018 Hangzhou, China +1

Vessel Segmentation

Ask PDF

May 19, 2026cs.CV

Uncertainty-Guided Conservative Propagation for Structured Inference in Vessel Segmentation

Accurate vessel segmentation is essential for medical image analysis, yet remains challenging due to complex vascular patterns and imaging ambiguity. Most deep models rely on single-pass prediction, limiting their ability to refine uncertain or disconnected regions during inference. To address this limitation, we propose Uncertainty-Guided Conservative Propagation (UGCP), a general plug-in module for vessel segmentation. Instead of directly using a one-shot output as the final prediction, UGCP performs a small number of logit-space update steps to refine the segmentation through local predictions interaction. Predictive uncertainty guides reliable regions to support ambiguous regions, while structure-aware modulation and source-based stabilization reduce unreliable propagation and excessive drift. The module is differentiable and can be trained end-to-end with different segmentation networks. We evaluate UGCP on four public vessel segmentation datasets covering 2D and 3D tasks, including retinal vessel, coronary artery, and cerebral vessel segmentation. Experiments with convolutional neural network-based and Transformer-based backbones show consistent improvements in Dice similarity coefficient, centerline Dice, and 95th percentile Hausdorff distance. Further analysis demonstrates that UGCP reduces vessel disconnections and improves structural consistency with limited additional computation. The code will be made available at https://github.com/chenzhao2023/UGC_PR.

Huan Huang, Michele Esposito, Chen Zhao

Department of Computer Science, Kennesaw State University, Marietta, 30060, GA, USA · Department of Cardiology, Medical University of South Carolina, Charleston, SC, USA

Vessel Segmentation Semi-Supervised Medical Image Segmentation U-Net Convolutional

Ask PDF