pith. sign in

arxiv: 2605.20543 · v1 · pith:HLECSZL7new · submitted 2026-05-19 · 💻 cs.CV

Uncertainty-Guided Conservative Propagation for Structured Inference in Vessel Segmentation

Huan Huang , Michele Esposito , Chen Zhao This is my paper

Pith reviewed 2026-05-21 06:32 UTC · model grok-4.3

classification 💻 cs.CV
keywords vessel segmentationuncertainty guidancestructured inferencemedical imagingdeep learning refinementconservative propagationvascular patterns
0
0 comments X

The pith

Uncertainty-guided conservative propagation refines vessel segmentations by letting reliable regions support ambiguous ones in a few update steps.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper proposes Uncertainty-Guided Conservative Propagation (UGCP) as a general plug-in module that refines initial segmentation predictions for vessels. It does this through a small number of logit-space update steps where predictive uncertainty guides interactions between reliable and ambiguous regions, while structure-aware modulation prevents unreliable changes. Experiments across four public datasets for retinal, coronary, and cerebral vessels demonstrate consistent gains in Dice scores, centerline Dice, and reduced Hausdorff distances when added to both CNN and Transformer models. This approach addresses challenges from complex vascular patterns and imaging ambiguity by improving structural consistency with limited added computation. The module is differentiable and trainable end-to-end with segmentation networks.

Core claim

UGCP performs a small number of logit-space update steps to refine the segmentation through local predictions interaction, with predictive uncertainty guiding reliable regions to support ambiguous regions, while structure-aware modulation and source-based stabilization reduce unreliable propagation and excessive drift.

What carries the argument

Uncertainty-Guided Conservative Propagation (UGCP), a differentiable plug-in module for iterative logit updates guided by uncertainty maps and local interactions.

If this is right

  • Consistent improvements in Dice similarity coefficient, centerline Dice, and 95th percentile Hausdorff distance on four vessel segmentation datasets.
  • Reduction in vessel disconnections and better structural consistency.
  • Compatibility with both convolutional neural network-based and Transformer-based backbones.
  • End-to-end training with different segmentation networks and limited additional computation.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Similar propagation strategies could apply to other medical segmentation tasks involving thin or branched structures.
  • The method suggests that inference-time refinement can address limitations in single-pass predictions for ambiguous images.
  • Integration into clinical workflows might decrease reliance on manual corrections for vessel maps.

Load-bearing premise

The initial model's uncertainty estimates are sufficiently calibrated and spatially informative to guide reliable propagation without introducing new errors or excessive drift.

What would settle it

Observing that UGCP degrades segmentation performance on datasets where the base model's uncertainty is poorly calibrated or uninformative would falsify the method's reliability.

Figures

Figures reproduced from arXiv: 2605.20543 by Chen Zhao, Huan Huang, Michele Esposito.

Figure 1
Figure 1. Figure 1: Motivation of the proposed UGCP framework. (a) Representative invasive coro [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Overview of the proposed UGCP framework. The backbone feature [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Qualitative comparison on the 2D vessel segmentation datasets. Representative [PITH_FULL_IMAGE:figures/full_fig_p020_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Qualitative comparison on the 3D vessel segmentation datasets. Representative [PITH_FULL_IMAGE:figures/full_fig_p021_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Effect of uncertainty-guided propagation on representative FIVES examples. [PITH_FULL_IMAGE:figures/full_fig_p025_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Effect of uncertainty-guided propagation on representative COSTA examples. [PITH_FULL_IMAGE:figures/full_fig_p026_6.png] view at source ↗
read the original abstract

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.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 2 minor

Summary. The manuscript proposes Uncertainty-Guided Conservative Propagation (UGCP), a differentiable plug-in module for vessel segmentation networks. UGCP performs a small number of logit-space update steps at inference time in which predictive uncertainty modulates the support from reliable regions to ambiguous ones, augmented by structure-aware modulation and source-based stabilization to limit drift. The module can be trained end-to-end with CNN or Transformer backbones and is evaluated on four public 2D/3D vessel datasets (retinal, coronary, cerebral), reporting consistent gains in Dice, centerline Dice, and 95th-percentile Hausdorff distance together with reduced vessel disconnections.

Significance. If the gains can be attributed specifically to the uncertainty-guided mechanism rather than generic refinement, UGCP would supply a lightweight, architecture-agnostic inference-time tool for improving structural consistency in medical segmentation where connectivity and boundary precision matter. Positive aspects include end-to-end differentiability, evaluation across two backbone families and four datasets, and the stated intention to release code.

major comments (3)
  1. [Abstract] Abstract: the claim of 'consistent improvements' in Dice, clDice and HD95 is stated without any quantitative calibration diagnostics (ECE, reliability diagrams), component-wise ablations, or statistical significance tests, leaving open whether the reported metric gains are driven by the uncertainty-guided propagation or by other factors.
  2. [Method] Method section (paragraph describing UGCP components): the mechanism assumes that the initial model's uncertainty estimates are both calibrated and spatially informative enough to guide safe logit propagation without introducing new errors or excessive drift; no verification of this assumption (e.g., ablation replacing uncertainty maps with constant or random fields, or analysis of failure cases on thin/disconnected vessels) is provided.
  3. [Experiments] Experiments section: while results with CNN and Transformer backbones are mentioned, the manuscript does not report controls that isolate the contribution of uncertainty modulation versus the number of propagation steps or generic smoothing, which is load-bearing for the central claim that UGCP provides structured inference benefits.
minor comments (2)
  1. [Method] The notation used for the logit update rule and the structure-aware modulation term would benefit from an explicit equation with all variables defined in one place.
  2. [Figures] Figure captions could more clearly indicate which backbone and dataset each panel corresponds to.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive and detailed review of our manuscript. The comments highlight important aspects that can strengthen the presentation of our results and the validation of our method's assumptions. We respond to each major comment below and indicate the changes we will make in the revised version.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the claim of 'consistent improvements' in Dice, clDice and HD95 is stated without any quantitative calibration diagnostics (ECE, reliability diagrams), component-wise ablations, or statistical significance tests, leaving open whether the reported metric gains are driven by the uncertainty-guided propagation or by other factors.

    Authors: We agree that additional evidence would better support the claim of consistent improvements. In the revised manuscript, we will add statistical significance tests using paired t-tests to report p-values for the metric gains. We will also include component-wise ablations in the main paper (currently referenced in supplementary) and qualify the abstract claim accordingly. For calibration diagnostics, we will add a short discussion noting that we rely on the base model's uncertainty estimates and plan to explore explicit calibration in future work. revision: partial

  2. Referee: [Method] Method section (paragraph describing UGCP components): the mechanism assumes that the initial model's uncertainty estimates are both calibrated and spatially informative enough to guide safe logit propagation without introducing new errors or excessive drift; no verification of this assumption (e.g., ablation replacing uncertainty maps with constant or random fields, or analysis of failure cases on thin/disconnected vessels) is provided.

    Authors: This is a valid concern regarding the core assumption of our method. We will revise the method section to include an ablation where uncertainty maps are substituted with constant or random fields, demonstrating the importance of informative uncertainty estimates. We will also add an analysis of failure cases, focusing on scenarios involving thin or disconnected vessels, to provide a more complete picture of the method's behavior and limitations. revision: yes

  3. Referee: [Experiments] Experiments section: while results with CNN and Transformer backbones are mentioned, the manuscript does not report controls that isolate the contribution of uncertainty modulation versus the number of propagation steps or generic smoothing, which is load-bearing for the central claim that UGCP provides structured inference benefits.

    Authors: We appreciate this suggestion to better isolate the contributions. In the updated experiments section, we will report results for different numbers of propagation steps and include a comparison with a generic smoothing approach that applies uniform updates without uncertainty guidance. This will help substantiate that the structured inference benefits stem specifically from the uncertainty modulation component of UGCP. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical method with external validation

full rationale

The paper introduces UGCP as a differentiable plug-in module performing a fixed number of logit-space updates modulated by the base model's uncertainty estimates. No equations or derivations are presented that reduce the claimed metric gains to quantities defined by construction from the inputs. All reported improvements (Dice, clDice, HD95) are measured on four independent public datasets using standard metrics, with the method trained end-to-end but evaluated separately. The central assumption that uncertainty is calibrated and spatially informative is stated but not derived; it remains an empirical precondition rather than a self-referential fit. No self-citation chains, uniqueness theorems, or ansatz smuggling appear in the provided description.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The method relies on the existence of a usable uncertainty signal from the base network and on the assumption that local logit interactions can be stabilized without additional learned parameters beyond those described.

axioms (1)
  • domain assumption Predictive uncertainty from the base segmentation network provides a reliable spatial signal for directing propagation.
    Invoked in the description of how reliable regions support ambiguous regions.

pith-pipeline@v0.9.0 · 5747 in / 1163 out tokens · 28868 ms · 2026-05-21T06:32:19.398205+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

  • IndisputableMonolith/Cost/FunctionalEquation.lean washburn_uniqueness_aczel unclear
    ?
    unclear

    Relation between the paper passage and the cited Recognition theorem.

    We formulate prediction update as a state-change modeling problem rather than a value assignment process. This leads to a conservation-inspired flux-balance formulation with a source term. In the continuous form, the evolution of the state variable can be expressed in the partial differential equation (PDE) form ∂s/∂t = −∇·F(s) + R(s).

  • IndisputableMonolith/Foundation/RealityFromDistinction.lean reality_from_one_distinction unclear
    ?
    unclear

    Relation between the paper passage and the cited Recognition theorem.

    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.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

51 extracted references · 51 canonical work pages · 2 internal anchors

  1. [1]

    Zhao, 3d vessel-like structure segmentation in medical images by an edge-reinforced network, Medical Image Analysis 82 (2022) 102581

    L.Xia, H.Zhang, Y.Wu, R.Song, Y.Ma, L.Mou, J.Liu, Y.Xie, M.Ma, Y. Zhao, 3d vessel-like structure segmentation in medical images by an edge-reinforced network, Medical Image Analysis 82 (2022) 102581

    work page 2022

  2. [2]

    Y. Meng, Z. Du, C. Zhao, M. Dong, D. Pienta, J. Tang, W. Zhou, Automatic extraction of coronary arteries using deep learning in invasive coronary angiograms, Technology and Health Care 31 (6) (2023) 2303– 2317. 28

    work page 2023

  3. [3]

    Y. Li, Y. Wu, J. He, W. Jiang, J. Wang, Y. Peng, Y. Jia, T. Xiong, K. Jia, Z. Yi, et al., Automatic coronary artery segmentation and di- agnosis of stenosis by deep learning based on computed tomographic coronary angiography, European Radiology 32 (9) (2022) 6037–6045

    work page 2022

  4. [4]

    Peper, L

    J. Peper, L. M. Becker, H. van den Berg, W. L. Bor, J. Brouwer, V. J. Nijenhuis, D.-J. Van Ginkel, B. J. Rensing, J. M. Ten Berg, L. Timmers, et al., Diagnostic performance of ccta and ct-ffr for the detection of cad in tavr work-up, Cardiovascular Interventions 15 (11) (2022) 1140–1149

    work page 2022

  5. [5]

    G. Wang, P. Zhou, H. Gao, Z. Qin, S. Wang, J. Sun, H. Yu, Coronary vesselsegmentationincoronaryangiographywithamulti-scaleu-shaped transformer incorporating boundary aggregation and topology preserva- tion, Physics in Medicine & Biology 69 (2) (2024) 025012

    work page 2024

  6. [6]

    O. O. Sule, A survey of deep learning for retinal blood vessel segmenta- tion methods: taxonomy, trends, challenges and future directions, IEEE Access 10 (2022) 38202–38236

    work page 2022

  7. [7]

    M. R. Goni, N. I. R. Ruhaiyem, M. Mustapha, A. Achuthan, C. M. N. C. M. Nassir, Brain vessel segmentation using deep learning—a review, IEEE access 10 (2022) 111322–111336

    work page 2022

  8. [9]

    Khandouzi, A

    A. Khandouzi, A. Ariafar, Z. Mashayekhpour, M. Pazira, Y. Baleghi, Retinal vessel segmentation, a review of classic and deep methods, An- nals of Biomedical Engineering 50 (10) (2022) 1292–1314

    work page 2022

  9. [10]

    Radha, Y

    K. Radha, Y. Karuna, Modified depthwise parallel attention unet for retinal vessel segmentation, IEEE Access 11 (2023) 102572–102588

    work page 2023

  10. [11]

    D.Chen, W.Yang, L.Wang, S.Tan, J.Lin, W.Bu, Pcat-unet: Unet-like network fused convolution and transformer for retinal vessel segmenta- tion, PloS one 17 (1) (2022) e0262689. 29

    work page 2022

  11. [12]

    J. Cao, J. Chen, Y. Gu, J. Liu, Mfa-unet: A vessel segmentation method based on multi-scale feature fusion and attention module, Frontiers in Neuroscience 17 (2023) 1249331

    work page 2023

  12. [13]

    Attari, N

    M. Attari, N. P. Nguyen, K. Palaniappan, F. Bunyak, Multi-loss topology-aware deep learning network for segmentation of vessels in mi- croscopy images, in: 2023 IEEE Applied Imagery Pattern Recognition Workshop (AIPR), IEEE, 2023, pp. 1–7

    work page 2023

  13. [14]

    Tan, K.-F

    Y. Tan, K.-F. Yang, S.-X. Zhao, Y.-J. Li, Retinal vessel segmentation with skeletal prior and contrastive loss, IEEE Transactions on Medical Imaging 41 (9) (2022) 2238–2251

    work page 2022

  14. [15]

    W. Li, Y. Xiao, H. Hu, C. Zhu, H. Wang, Z. Liu, A. K. Sangaiah, Retinal vessel segmentation based on b-cosfire filters in fundus images, Frontiers in Public Health 10 (2022) 914973

    work page 2022

  15. [16]

    M. Xia, H. Yang, Y. Huang, Y. Qu, G. Zhou, F. Zhang, Y. Wang, Y. Guo, 3d pyramidal densely connected network with cross-frame un- certainty guidance for intravascular ultrasound sequence segmentation, Physics in Medicine & Biology 68 (5) (2023) 055001

    work page 2023

  16. [17]

    Galdran, A

    A. Galdran, A. Anjos, J. Dolz, H. Chakor, H. Lombaert, I. B. Ayed, State-of-the-art retinal vessel segmentation with minimalistic models, Scientific Reports 12 (1) (2022) 6174

    work page 2022

  17. [18]

    Payandeh, K

    A. Payandeh, K. T. Baghaei, P. Fayyazsanavi, S. B. Ramezani, Z. Chen, S. Rahimi, Deep representation learning: Fundamentals, technologies, applications, and open challenges, IEEE Access 11 (2023) 137621– 137659

    work page 2023

  18. [19]

    X.Zhao, L.Wang, Y.Zhang, X.Han, M.Deveci, M.Parmar, Areviewof convolutional neural networks in computer vision, Artificial Intelligence Review 57 (4) (2024) 99

    work page 2024

  19. [20]

    Ronneberger, P

    O. Ronneberger, P. Fischer, T. Brox, U-net: Convolutional networks for biomedicalimagesegmentation, in: InternationalConferenceonMedical image computing and computer-assisted intervention, Springer, 2015, pp. 234–241. 30

    work page 2015

  20. [21]

    M. K. Kar, D. R. Neog, M. K. Nath, Retinal vessel segmentation using multi-scale residual convolutional neural network (msr-net) combined with generative adversarial networks, Circuits, Systems, and Signal Pro- cessing 42 (2) (2023) 1206–1235

    work page 2023

  21. [22]

    A. Song, L. Xu, L. Wang, B. Wang, X. Yang, B. Xu, B. Yang, S. E. Greenwald, Automatic coronary artery segmentation of ccta images with an efficient feature-fusion-and-rectification 3d-unet, IEEE journal of biomedical and health informatics 26 (8) (2022) 4044–4055

    work page 2022

  22. [23]

    Vaswani, N

    A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, Ł. Kaiser, I. Polosukhin, Attention is all you need, Advances in neural information processing systems 30 (2017)

    work page 2017

  23. [24]

    An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale

    A. Dosovitskiy, L. Beyer, A. Kolesnikov, D. Weissenborn, X. Zhai, T. Unterthiner, M. Dehghani, M. Minderer, G. Heigold, S. Gelly, et al., An image is worth 16x16 words: Transformers for image recognition at scale, arXiv preprint arXiv:2010.11929 (2020)

    work page internal anchor Pith review Pith/arXiv arXiv 2010

  24. [25]

    Huang, L

    H. Huang, L. Qiu, S. Yang, L. Li, J. Nan, Y. Li, C. Han, F. Zhu, C. Zhao, W. Zhou, 3d lymphoma segmentation on pet/ct images via multi-scale information fusion with cross-attention, Medical Physics 52 (6) (2025) 4371–4389

    work page 2025

  25. [26]

    J. Chen, Y. Lu, Q. Yu, X. Luo, E. Adeli, Y. Wang, L. Lu, A. L. Yuille, Y. Zhou, Transunet: Transformers make strong encoders for medical image segmentation, arXiv preprint arXiv:2102.04306 (2021)

    work page internal anchor Pith review Pith/arXiv arXiv 2021

  26. [27]

    C. Pan, B. Qi, G. Zhao, J. Liu, C. Fang, D. Zhang, J. Li, Deep 3d vessel segmentation based on cross transformer network, in: 2022 IEEE inter- national conference on bioinformatics and biomedicine (BIBM), IEEE, 2022, pp. 1115–1120

    work page 2022

  27. [28]

    Y. Wu, S. Qi, M. Wang, S. Zhao, H. Pang, J. Xu, L. Bai, H. Ren, Transformer-based 3d u-net for pulmonary vessel segmentation and artery-vein separation from ct images, Medical & Biological Engineering & Computing 61 (10) (2023) 2649–2663

    work page 2023

  28. [29]

    Khandouzi, A

    A. Khandouzi, A. Ariafar, Z. Mashayekhpour, M. Pazira, Y. Baleghi, Retinal vessel segmentation, a review of classic and deep methods, An- nals of Biomedical Engineering 50 (10) (2022) 1292–1314. 31

    work page 2022

  29. [30]

    Y. Jin, A. Pepe, J. Li, C. Gsaxner, Y. Chen, B. Puladi, F.-h. Zhao, K. Pomykala, J. Kleesiek, A. F. Frangi, et al., Aortic vessel tree segmen- tation for cardiovascular diseases treatment: Status quo, ACM Comput- ing Surveys 57 (9) (2025) 1–35

    work page 2025

  30. [31]

    Banerjee, D

    S. Banerjee, D. Toumpanakis, A. K. Dhara, J. Wikström, R. Strand, Topology-aware learning for volumetric cerebrovascular segmentation, in: 2022 IEEE 19th International Symposium on Biomedical Imaging (ISBI), IEEE, 2022, pp. 1–4

    work page 2022

  31. [32]

    X. Li, R. Bala, V. Monga, Robust deep 3d blood vessel segmentation us- ing structural priors, IEEE Transactions on Image Processing 31 (2022) 1271–1284

    work page 2022

  32. [33]

    Z. Zhao, Z. Zhang, Y. Liu, Z. Zhang, H. Yu, D. Wang, L. Wang, De- formcl: Learning deformable centerline representation for vessel extrac- tion in 3d medical image, in: Proceedings of the Computer Vision and Pattern Recognition Conference, 2025, pp. 30896–30905

    work page 2025

  33. [34]

    L. Yao, F. Shi, S. Wang, X. Zhang, Z. Xue, X. Cao, Y. Zhan, L. Chen, Y. Chen, B. Song, et al., Tag-net: topology-aware graph network for centerline-based vessel labeling, IEEE transactions on medical imaging 42 (11) (2023) 3155–3166

    work page 2023

  34. [35]

    Krähenbühl, V

    P. Krähenbühl, V. Koltun, Efficient inference in fully connected crfs withgaussianedgepotentials, Advancesinneuralinformationprocessing systems 24 (2011)

    work page 2011

  35. [36]

    Dulau, B

    I. Dulau, B. Recur, C. Helmer, C. Delcourt, M. Beurton-Aimar, Vnr-av: structural post-processing for retinal arteries and veins segmentation, in: International workshop on ophthalmic medical image analysis, Springer, 2024, pp. 22–31

    work page 2024

  36. [37]

    W. He, Z. Jiang, T. Xiao, Z. Xu, Y. Li, A survey on uncertainty quan- tification methods for deep learning, ACM Computing Surveys 58 (7) (2026) 1–35

    work page 2026

  37. [38]

    Faghani, M

    S. Faghani, M. Moassefi, P. Rouzrokh, B. Khosravi, F. I. Baffour, M. D. Ringler, B. J. Erickson, Quantifying uncertainty in deep learning of radiologic images, Radiology 308 (2) (2023) e222217. 32

    work page 2023

  38. [39]

    P. Tang, P. Yang, D. Nie, X. Wu, J. Zhou, Y. Wang, Unified medi- cal image segmentation by learning from uncertainty in an end-to-end manner, Knowledge-Based Systems 241 (2022) 108215

    work page 2022

  39. [40]

    C. Li, W. Ma, L. Sun, X. Ding, Y. Huang, G. Wang, Y. Yu, Hierarchical deep network with uncertainty-aware semi-supervised learning for vessel segmentation, Neural Computing and Applications 34 (4) (2022) 3151– 3164

    work page 2022

  40. [41]

    Huang, S

    L. Huang, S. Ruan, P. Decazes, T. Denœux, Deep evidential fusion with uncertainty quantification and reliability learning for multimodal medical image segmentation, Information Fusion 113 (2025) 102648

    work page 2025

  41. [42]

    Zheng, S

    S. Zheng, S. Jayasumana, B. Romera-Paredes, V. Vineet, Z. Su, D. Du, C. Huang, P. H. Torr, Conditional random fields as recurrent neural networks, in: Proceedings of the IEEE international conference on com- puter vision, 2015, pp. 1529–1537

    work page 2015

  42. [43]

    Ghoshal, A

    B. Ghoshal, A. Tucker, B. Sanghera, W. Lup Wong, Estimating uncer- tainty in deep learning for reporting confidence to clinicians in medical image segmentation and diseases detection, Computational Intelligence 37 (2) (2021) 701–734

    work page 2021

  43. [44]

    A. Sagar, Uncertainty quantification using variational inference for biomedical image segmentation, in: Proceedings of the IEEE/CVF Win- ter Conference on Applications of Computer Vision, 2022, pp. 44–51

    work page 2022

  44. [45]

    M. C. Krygier, T. LaBonte, C. Martinez, C. Norris, K. Sharma, L. N. Collins, P. P. Mukherjee, S. A. Roberts, Quantifying the unknown im- pact of segmentation uncertainty on image-based simulations, Nature communications 12 (1) (2021) 5414

    work page 2021

  45. [46]

    H.Li, Y.Nan, J.DelSer, G.Yang, Region-basedevidentialdeeplearning to quantify uncertainty and improve robustness of brain tumor segmen- tation, Neural Computing and Applications 35 (30) (2023) 22071–22085

    work page 2023

  46. [47]

    Sensoy, L

    M. Sensoy, L. Kaplan, M. Kandemir, Evidential deep learning to quan- tify classification uncertainty, Advances in neural information processing systems 31 (2018). 33

    work page 2018

  47. [48]

    K. Jin, X. Huang, J. Zhou, Y. Li, Y. Yan, Y. Sun, Q. Zhang, Y. Wang, J. Ye, Fives: A fundus image dataset for artificial intelligence based vessel segmentation, Scientific data 9 (1) (2022) 475

    work page 2022

  48. [49]

    C. Zhao, A. Vij, S. Malhotra, J. Tang, H. Tang, D. Pienta, Z. Xu, W. Zhou, Automatic extraction and stenosis evaluation of coronary arteries in invasive coronary angiograms, Computers in biology and medicine 136 (2021) 104667

    work page 2021

  49. [50]

    A. Zeng, C. Wu, G. Lin, W. Xie, J. Hong, M. Huang, J. Zhuang, S. Bi, D. Pan, N. Ullah, et al., Imagecas: A large-scale dataset and bench- mark for coronary artery segmentation based on computed tomography angiography images, Computerized Medical Imaging and Graphics 109 (2023) 102287

    work page 2023

  50. [51]

    L. Mou, J. Lin, Y. Zhao, Y. Liu, S. Ma, J. Zhang, W. Lv, T. Zhou, J. Liu, A. F. Frangi, et al., Costa: A multi-center tof-mra dataset and a style self-consistency network for cerebrovascular segmentation, IEEE transactions on medical imaging 43 (12) (2024) 4442–4456

    work page 2024

  51. [52]

    Y. He, V. Nath, D. Yang, Y. Tang, A. Myronenko, D. Xu, Swinunetr- v2: Stronger swin transformers with stagewise convolutions for 3d med- ical image segmentation, in: International Conference on Medical Im- ageComputingandComputer-AssistedIntervention, Springer, 2023, pp. 416–426. 34

    work page 2023