List of Papers Browse by Subject Areas Author List
Abstract
Coronary guidewire endpoint localization is a fundamental capability for computer-assisted PCI, and its importance increases as robot-assisted PCI is progressively adopted to reduce operator radiation exposure. However, the scarcity of annotated CAG images with guidewires and the limited adaptability of existing guidewire synthesis models remain key bottlenecks for guidewire endpoint localization. To address this issue, we propose VDSB-GWSyn, a Diffusion Schrödinger Bridge (DSB) model-based framework, enabling synthesis of controllable, high-fidelity guidewire samples under complex anatomical backgrounds. VDSB-GWSyn first uses our shape prior algorithm to learn the basic guidewire geometry. It then generates guidewire masks under constraints imposed by the vessel segmentation masks and outputs the corresponding endpoint coordinates. Finally, it synthesizes realistic guidewire samples on real CAG images using DSB conditioned with SPADE. Experimental results show that the guidewire samples synthesized by VDSB-GWSyn achieve favorable ROI-FID and ROI-KID, as well as high IPR scores. In addition, incorporating our synthesized data for synthetic pre-training followed by real fine-tuning substantially improves downstream guidewire endpoint localization, reducing MPE from 16.01 px to 7.71 px and increasing PCK at 3 px from 52.63% to 86.27%, leading to more clinically reliable deployment of robot-assisted guidewire delivery systems. Moreover, the core design philosophy of controllable device synthesis with strict background preservation and anatomical feasibility constraints has the potential to transfer to other interventional device perception tasks where annotated data are scarce.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026/paper/2982_paper.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: Not Submitted
Link to the Code Repository
N/A
Link to the Dataset(s)
N/A
BibTex
@InProceedings{TanHao_VDSBGWSyn_MICCAI2026,
author = { Tang, Haoyuan AND Zhang, Zhuo AND Li, Jialin AND Xiao, Shuai AND Yang, Jiachen},
title = { { VDSB-GWSyn: Diffusion Schrödinger Bridge for Controllable and Anatomically Feasible Guidewire Synthesis in Coronary Angiography } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 16893},
month = {September},
page = {pending}
}
Reviews
Review #1
- Please describe the contribution of the paper
The manuscript proposes a method to simulate guidewires in fluoroscopic images constrained to vessels using a variant of a diffusion model (diffusion schrodinger bridge) modulated by spatially adaptive normalization. Evaluation metrics are presented against two general GAN and diffusion models.
- Please list the major strengths of the paper: you should highlight a novel formulation, an original way to use data, demonstration of clinical feasibility, a novel application, a particularly strong evaluation, or anything else that is a strong aspect of this work. Please provide details, for instance, if a method is novel, explain what aspect is novel and why this is interesting.
-The manuscript is well organized. -Synthetic generation of guidewire images is highly relevant to the field and can be useful for training models for downstream tasks. -The authors make a sound proposal to use existing methods in image synthesis to generate guidewire images with care to ensure that they are bound to vasculature, and show improved evaluation metrics over compared methods.
- Please list the major weaknesses of the paper. Please provide details: for instance, if you state that a formulation, way of using data, demonstration of clinical feasibility, or application is not novel, then you must provide specific references to prior work.
-In parts, the methods are described only at a high level. Without providing code I am not certain I would be able to reproduce this method. The main novelty methodology appears to be constraining the image generation to the vasculature, however details of this are not fully described such as how the wall hugging parameter functions or how a sequential set of images could be generated. -The figures are too small to appreciate whether or not the generated images are of good quality. -Details of the evaluation including the size of the ROI patches, how the ROI is selected around the guidewire, are missing.
- Please rate the clarity and organization of this paper
Satisfactory
- Please comment on the reproducibility of the paper. Please be aware that providing code and data is a plus, but not a requirement for acceptance.
The submission does not mention open access to source code or data, but provides a clear and detailed description of the algorithm to ensure reproducibility.
- Based on your review and your understanding of the MICCAI Scientific Code of Ethics, do you believe this submission may involve a potential ethics concern or violation?
N/A
- Optional: If you have any additional comments to share with the authors, please provide them here. Please also refer to our Reviewer’s guide on what makes a good review and pay specific attention to the different assessment criteria for the different paper categories: https://conferences.miccai.org/2026/en/REVIEWER-GUIDELINES.html
N/A
- Rate the paper on a scale of 1-6, 6 being the strongest (6-4: accept; 3-1: reject). Please use the entire range of the distribution. Spreading the score helps create a distribution for decision-making.
(4) Weak Accept — marginally above the acceptance threshold, but would not mind if rejected, dependent on rebuttal
- Please justify your recommendation. What were the major factors that led you to your overall score for this paper?
-The manuscript is well written and provides a sound proposal to apply existing methods to an important problem. -However, the methodology is described without relevant details and the results are difficult to evaluate qualitatively.
- Reviewer confidence
Confident but not absolutely certain (3)
- [Post rebuttal] After reading the authors’ rebuttal, please state your final opinion of the paper.
N/A
- [Post rebuttal] Please justify your final decision from above.
N/A
Review #2
- Please describe the contribution of the paper
This paper proposes VDSB-GWSyn, a controllable guidewire synthesis framework for coronary angiography. The main contribution is to improve downstream guidewire endpoint localization by generating synthetic guidewire samples with better anatomical fidelity and reliability through vessel-constrained mask generation, targeted DSB-based inpainting, and sparse-structure-aware optimization.
- Please list the major strengths of the paper: you should highlight a novel formulation, an original way to use data, demonstration of clinical feasibility, a novel application, a particularly strong evaluation, or anything else that is a strong aspect of this work. Please provide details, for instance, if a method is novel, explain what aspect is novel and why this is interesting.
1.The paper addresses a practical and clinically relevant problem: limited annotated guidewire data for endpoint localization in coronary angiography, which is important for computer-assisted and robot-assisted PCI. 2.The framework is well aligned with the task, combining vessel-constrained mask generation with localized diffusion rendering to synthesize plausible guidewires while preserving the original anatomical background. 3.The paper evaluates downstream utility rather than only reporting generative metrics, and synthetic pre-training leads to clear gains in endpoint localization.
- Please list the major weaknesses of the paper. Please provide details: for instance, if you state that a formulation, way of using data, demonstration of clinical feasibility, or application is not novel, then you must provide specific references to prior work.
1.Figure 2 shows substantial variation in the synthesized guidewires across left and right coronary cases. However, reporting a large number of CAG frames alone is not sufficient to support generalization, as the manuscript does not specify how many patients are included or how many frames are collected per patient. 2.It should be clarified whether pix2pixHD and ControlNet are also configured for localized guidewire synthesis. If they are used for global image generation instead, the poorer guidewire quality would be expected, making the comparison less fair. 3.Some minor errors are present. In Table 2, the best results for PCK@3, PCK@6, and TIP MPE appear to be achieved with 2,000 pre-training samples, but the bold formatting is instead assigned to the 3,000-sample setting. In Section 3.4, “PCK@3 [25] px” should be corrected to “PCK@3 px [25]”.
- Please rate the clarity and organization of this paper
Satisfactory
- Please comment on the reproducibility of the paper. Please be aware that providing code and data is a plus, but not a requirement for acceptance.
The submission does not mention open access to source code or data, but provides a clear and detailed description of the algorithm to ensure reproducibility.
- Based on your review and your understanding of the MICCAI Scientific Code of Ethics, do you believe this submission may involve a potential ethics concern or violation?
N/A
- Optional: If you have any additional comments to share with the authors, please provide them here. Please also refer to our Reviewer’s guide on what makes a good review and pay specific attention to the different assessment criteria for the different paper categories: https://conferences.miccai.org/2026/en/REVIEWER-GUIDELINES.html
N/A
- Rate the paper on a scale of 1-6, 6 being the strongest (6-4: accept; 3-1: reject). Please use the entire range of the distribution. Spreading the score helps create a distribution for decision-making.
(4) Weak Accept — marginally above the acceptance threshold, but would not mind if rejected, dependent on rebuttal
- Please justify your recommendation. What were the major factors that led you to your overall score for this paper?
This paper addresses an important problem and presents a coherent, task-specific framework with clear practical value. The combination of anatomically constrained synthesis and localized diffusion rendering is sensible, and the downstream endpoint localization results are encouraging. However, the evidence for generalization remains limited, the baseline comparison could be clarified further, and there are a few presentation issues in the experimental section, which together prevent a higher score.
- Reviewer confidence
Somewhat confident (2)
- [Post rebuttal] After reading the authors’ rebuttal, please state your final opinion of the paper.
N/A
- [Post rebuttal] Please justify your final decision from above.
N/A
Review #3
- Please describe the contribution of the paper
This paper presents VDSB-GWSyn, a novel framework for generating high-fidelity, anatomically plausible guidewire images in coronary angiography by leveraging a Diffusion Schrödinger Bridge model. Its key contributions include an anatomically constrained path sampling algorithm to ensure vessel-aware guidewire trajectories and a targeted inpainting strategy with a region-normalized objective to seamlessly integrate synthetic guidewires into real backgrounds. The framework’s primary impact is demonstrated through its ability to significantly enhance downstream guidewire tip localization performance, effectively mitigating the challenge of data scarcity in interventional cardiology.
- Please list the major strengths of the paper: you should highlight a novel formulation, an original way to use data, demonstration of clinical feasibility, a novel application, a particularly strong evaluation, or anything else that is a strong aspect of this work. Please provide details, for instance, if a method is novel, explain what aspect is novel and why this is interesting.
Novel Integration of Schrodinger Bridge: The paper innovatively applies the Diffusion Schrodinger Bridge (DSB) model to guidewire synthesis. This is a significant methodological advance over standard diffusion models, as it offers a more flexible probabilistic framework for modeling the complex, non-linear dynamics of guidewire manipulation. Anatomically Plausible Generation: A key strength is the Vessel-Constrained Path Sampling algorithm. By strictly enforcing vascular segmentation masks as priors, the method guarantees that generated guidewires remain within the vessel lumen, ensuring the synthetic data is anatomically feasible and clinically relevant. High-Fidelity Structure Preservation: The framework effectively addresses the “sparse pixel” problem of thin instruments through SPADE-based targeted inpainting and a Region-Normalized Objective. This technical design ensures seamless background integration and prevents the structural fragmentation often seen in generative models. Proven Practical Utility: The paper provides strong empirical evidence of its value. The proposed “synthetic pre-training + real fine-tuning” strategy significantly boosts downstream performance, reducing the Mean Pixel Error (MPE) by over 50%, which validates its effectiveness in solving the data scarcity challenge.
- Please list the major weaknesses of the paper. Please provide details: for instance, if you state that a formulation, way of using data, demonstration of clinical feasibility, or application is not novel, then you must provide specific references to prior work.
Despite its novel generative framework, the paper exhibits several notable limitations, primarily concerning the lack of validation using real physician manipulation data, which raises concerns about the behavioral realism of the synthetic guidewires compared to actual clinical operations. The method demonstrates a high dependency on the accuracy of upstream vessel segmentation, making it potentially vulnerable to errors in anatomical extraction, while its reliance on 2D projection constraints fails to fully account for the inherent ambiguities of 3D spatial structures. Furthermore, the study lacks comprehensive comparisons with other specialized generative models designed for tubular structures, leaving the specific contribution of the proposed Schrödinger Bridge approach somewhat ambiguous.
- Please rate the clarity and organization of this paper
Satisfactory
- Please comment on the reproducibility of the paper. Please be aware that providing code and data is a plus, but not a requirement for acceptance.
The submission does not mention open access to source code or data, but provides a clear and detailed description of the algorithm to ensure reproducibility.
- Based on your review and your understanding of the MICCAI Scientific Code of Ethics, do you believe this submission may involve a potential ethics concern or violation?
N/A
- Optional: If you have any additional comments to share with the authors, please provide them here. Please also refer to our Reviewer’s guide on what makes a good review and pay specific attention to the different assessment criteria for the different paper categories: https://conferences.miccai.org/2026/en/REVIEWER-GUIDELINES.html
N/A
- Rate the paper on a scale of 1-6, 6 being the strongest (6-4: accept; 3-1: reject). Please use the entire range of the distribution. Spreading the score helps create a distribution for decision-making.
(4) Weak Accept — marginally above the acceptance threshold, but would not mind if rejected, dependent on rebuttal
- Please justify your recommendation. What were the major factors that led you to your overall score for this paper?
I assign a Weak Accept. The paper’s strength lies in its novel use of Schrodinger Bridge formulations for medical synthesis, offering a fresh perspective on data generation. However, the evaluation is limited by the absence of real physician manipulation data and a dependency on idealized segmentation masks. While the contribution is valuable enough for inclusion, the concerns about anatomical robustness and 3D spatial ambiguity prevent a stronger recommendation.
- Reviewer confidence
Very confident (4)
- [Post rebuttal] After reading the authors’ rebuttal, please state your final opinion of the paper.
N/A
- [Post rebuttal] Please justify your final decision from above.
N/A
Author Feedback
We thank the reviewers, Area Chairs, and Program Chairs for their constructive comments and for recognizing the clinical relevance of guidewire synthesis in coronary angiography. Since the paper has received supportive reviews, we mainly use this response to clarify several points. Regarding the concern that some methodological details were described at a high level, we clarify that our vessel-constrained path sampling is performed on the skeleton graph of the vessel mask. Vessel junctions are treated as graph nodes and vessel segments as edges. A shortest-path search with penalties on high-degree junctions is used to reduce unrealistic branch switching. The sampled path is then projected into the valid vessel lumen using the distance-transform field. The wall-hugging parameter controls the lateral position of the guidewire within the vessel lumen: smaller values keep the guidewire closer to the centerline, while larger values move it closer to the vessel wall without leaving the valid vascular region. Regarding sequential image generation, the submitted work focuses on controllable frame-level guidewire synthesis for data augmentation and downstream endpoint localization, rather than full physical simulation of guidewire manipulation. Nevertheless, temporally ordered masks can be generated by progressively updating the guidewire tip along the vessel graph with small step-size constraints, enabling advancement or retraction while preserving vessel feasibility. Regarding patient-level diversity, we have clarified the dataset statistics in the camera-ready version. Our internal CAG dataset consists of 1,125 frames collected from 84 patients, with 10–25 frames per patient and an average of 13.4 frames per patient. In addition, the synthesis pool includes 832 high-quality frames selected from the public ARCADE dataset, which provides additional anatomical diversity across coronary views and vessel morphologies. ARCADE frames and vessel masks are used only for vessel-constrained guidewire mask sampling and are not used as inputs to the downstream endpoint localizer. Regarding ROI-based evaluation, the ROI is centered around the synthesized guidewire region so that ROI-FID and ROI-KID emphasize the clinically relevant thin guidewire structure rather than the largely unchanged angiographic background. The same ROI extraction protocol is applied to all compared methods. Regarding baseline fairness, pix2pixHD and ControlNet were evaluated under the same synthetic pre-training and real fine-tuning protocol for downstream endpoint localization. The comparison is intended to evaluate whether each synthesis method provides useful training signal for the same downstream task. We acknowledge that these baselines are general-purpose image synthesis models and are not specifically optimized for sparse guidewire structures, which motivates our vessel-constrained and region-normalized design. We also acknowledge the reviewers’ comments on dependence on vessel segmentation quality, lack of real physician manipulation trajectories, and 2D projection ambiguity. These are important limitations and future directions. However, they do not change the main conclusion of the submitted manuscript. Anatomically constrained guidewire synthesis provides useful synthetic data and substantially improves downstream guidewire endpoint localization compared with real-only training and general synthesis baselines.
Meta-Review
Meta-review #1
- Your recommendation
Provisional Accept
- Please justify your decision. In case you deviate from the reviewers’ recommendations, explain in detail the reasons why. In case of an invitation for rebuttal, clarify which points are important to address in the rebuttal.
This paper addresses a clinically relevant and practically important problem, namely the scarcity of annotated guidewire data for coronary angiography analysis. Reviewers were positive about the overall motivation and found the proposed synthesis framework well aligned with the task. In particular, the vessel-constrained generation strategy was considered a meaningful and effective way to ensure anatomical plausibility, and the use of localized rendering to preserve the original anatomical background was viewed as a strength. Reviewers also appreciated that the paper evaluates downstream utility rather than relying only on generative quality metrics, with synthetic pre-training showing clear benefit for endpoint localization.
The main concerns raised by the reviewers relate primarily to clarity and completeness rather than to the overall validity of the contribution. In particular, some methodological details remain under-specified, including aspects of the vessel-constrained generation process, sequential synthesis, ROI selection, and certain implementation details needed for reproducibility. Reviewers also asked for clearer reporting of dataset composition at the patient level, more explicit clarification of the fairness of the baseline comparisons, and improved figure readability. One reviewer further noted that stronger discussion of limitations—such as dependence on vessel segmentation quality and the lack of validation against real physician manipulation trajectories—would strengthen the paper.
Overall, the reviewers found the work novel, clinically relevant, and empirically convincing, with a clear practical contribution to data augmentation for guidewire analysis. While several clarifications would improve the final version, the concerns appear addressable through revision and do not outweigh the strengths of the paper