List of Papers Browse by Subject Areas Author List
Abstract
Registering coronary CT angiography (CCTA) and intravascular ultrasound (IVUS) enables comprehensive coronary analysis that neither modality can provide alone, yet their fusion remains challenging due to differences in imaging geometry, resolution, and artifact profiles. Existing methods depend on pre-computed lumen or vessel wall segmentations that are unreliable under IVUS acoustic shadowing from calcifications, limiting their clinical applicability. We propose TG-OT, a fully automatic CCTA-IVUS registration framework that eliminates this dependency by integrating trained feature detectors directly into the registration pipeline. Lightweight CNNs are trained to predict calcifications, bifurcations, and lumen radii on the topological (theta, z) cylinder, encouraging topologically coherent detections without requiring explicit segmentation. Registration is formulated as an optimization over centerline warping parameters, driven by an unbalanced Sinkhorn optimal transport loss on the cylindrical geometry that provides spatially informative gradients even for spatially disjoint predictions, complemented by a lumen matching term. Evaluated on N=47 paired CCTA-IVUS cases in a 5-fold cross-validation setup, TG-OT achieves strong longitudinal (Dice_ctl = 0.99), rotational (S_=0.96), and lumen alignment (Dice_L=0.69) without manual interaction or prior segmentation, marking a meaningful step toward clinical integration of automatic CCTA-IVUS fusion.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026/paper/1931_paper.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: Not Submitted
Link to the Code Repository
https://github.com/RoelvH97/TG_OT
Link to the Dataset(s)
N/A
BibTex
@InProceedings{vanRud_TGOT_MICCAI2026,
author = { van Herten, Rudolf L. M. AND Henriques, José P. AND Planken, R. Nils AND Daemen, Joost AND Hartman, Eline M. J. AND Wentzel, Jolanda J. AND Paetzold, Johannes C. AND Išgum, Ivana},
title = { { TG-OT: Topology-Guided CCTA-IVUS Registration via Optimal Transport Matching } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 16889},
month = {September},
page = {pending}
}
Reviews
Review #1
- Please describe the contribution of the paper
The paper describes a novel approach to aligning pullback IVUS images of the coronary arteries with CCTA images. To achieve this, the paper uses unbalanced optimal transport to align segmentation masks in both modalities. The identified transform is then used to locally deform the centerline and rotate cross-sectional views. The method is developed and validated using a unique dataset of IVUS + CCTA images of 47 patients.
- 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.
- Using optimal transport for registration is very interesting, in particular in scenarios where little overlap is expected between two masks.
- The paper shows good results on challenging data.
- The paper is of very high quality, explains all concepts clearly and understandably, and presents a clear ablation study. Figures illustrate the method well.
- 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.
- While the method is interesting, results do not entirely convincingly show superiority over alternative approaches. Authors compare their OT scheme (the main contribution) with a Dice-based optimization scheme in Table 2.Results indicate that registration with a Dice loss leads to very similar results. In fact, there’s a bit of a mismatch between the single result shown in Fig. 3 and the aggregate results shown in Table 2.In Table 2, I barely see a difference between Dice and OT (also not in terms of Sc), while in Fig. 3, Dice demonstrates an Sc of 0.12 and OT an Sc of 0.96.This example seems to be an outlier.
- The Discussion is a bit confusing. It says that ‘the dependency on error-prone intermediate segmentations’ is removed, but as far as I can see, the method directly depends on intermediate segmentations. The difference is not clear to me.
- Please rate the clarity and organization of this paper
Good
- 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 has provided an anonymized link to the source code, dataset, or any other dependencies.
- 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
- The feature-extraction CNNs are frozen (Fig. 1), but the abstract states that feature detection is integrated directly into the registration pipeline.
- One of the parameters optimized is the per-frame local rotation \theta_p. I can imagine there is a strong correlation between this parameter across adjacent slices. Is this exploited?
- Why was the guidewire not used to mask the OT loss, but only the image losses?
- 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.
(5) Accept — should be accepted, independent of rebuttal
- Please justify your recommendation. What were the major factors that led you to your overall score for this paper?
This is a high-quality paper that solves a clinically and technically challenging problem using a technically sound method, and it would definitely be of interest to the community. The paper is very well-prepared.
- Reviewer confidence
Very confident (4)
- [Post rebuttal] After reading the authors’ rebuttal, please state your final opinion of the paper.
Accept
- [Post rebuttal] Please justify your final decision from above.
The few concerns I had were adequately addressed in the rebuttal. I remain of the opinion that this paper is of high quality and will be of interest to the MICCAI community.
Review #2
- Please describe the contribution of the paper
This paper addresses the registration problem between coronary intravascular ultrasound (IVUS) and coronary computed tomography angiography (CCTA). The proposed method integrates feature detection directly into the registration workflow. The authors formulate the registration task as a centerline deformation parameter estimation problem and solve it using optimal transport. The method is validated on 47 paired datasets.
- 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 primary strength of this work is the introduction of a novel approach to coronary image registration.
- 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.
However, a key limitation is the lack of clarity in the methodological description, particularly regarding notation. Many symbols used in the solution process are not explicitly defined. For instance, the parameters \theta and z, which appear in the abstract, are never introduced. Similarly, the meanings of M and F in Equation (2), presumably denoting the moving and fixed images, are left to the reader’s inference. It is also unclear what A represents in Equation (2), and the bold symbols \boldsymbol{\mu} and \boldsymbol{v} in Equation (9) are not explained.
- Please rate the clarity and organization of this paper
Poor
- 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 provide sufficient information for 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
All equations in this paper should be revised and clarified; otherwise, they are difficult for readers to understand.
- 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.
(1) Strong Reject — must be rejected due to major flaws
- Please justify your recommendation. What were the major factors that led you to your overall score for this paper?
The methodological desription is confusing for readers to understand.
- Reviewer confidence
Very confident (4)
- [Post rebuttal] After reading the authors’ rebuttal, please state your final opinion of the paper.
Reject
- [Post rebuttal] Please justify your final decision from above.
M_w (warped moving) and F (fixed) are introduced at the start of Sec. 2.2.->I mean M and F in the first term of Eq.(2) instead of M_w and F in the second term of Eq.(2), they are confusing. (θ, z) are angular and longitudinal coordinates on the unwound cylinder (Sec. 2.1, Fig. 2 caption).->The term “polar coordinates” is too general, and the authors do not explicitly state here that these are “angular and longitudinal coordinates”. A in Eq. 2 is the lumen cross-sectional area, described directly after the equation.-> The readers can not find “A” directly after the equation. The authors should check the submitted version. In Eq. 8, μ and ν are the marginal probability vectors from Eq. 4, where u and v are the Sinkhorn scaling vectors described after Eq. 7.->The term “marginal probability vectors” does not appear where the author mentions it, and explicit definitions of μ and ν are not found around Eq. 4.The authors should check the submitted version.
Review #3
- Please describe the contribution of the paper
Registration of coronary CCTA and IVUS images is a crucial step in coronary artery disease assessment. This work presents a novel topology-guided CCTA-IVUS registration approach, termed TG-OT. Method: We incorporate feature detection into the registration framework. An unbalanced Sinkhorn optimal transport loss is utilized to regularize the feature matching. The architecture inherently enforces topological coherency, thus enabling robust registration even for CCTA images suffering from severe geometric deformation, calcifications, and poor lumen segmentation. The TF-OT framework is evaluated on 47 pairs of CCTA-IVUS cases. Results: Comparable or better Dice scores for both lumen and calcification segmentation are obtained in comparison to other state-of-the-art methods. The fully automatic approach achieves results without manual intervention or prior CCTA segmentation.
- 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.Novelty in Registration Framework / Method: Novel methodology that integrates feature detection within a registration framework. Specifically, novelty is found in using an unbalanced Sinkhorn optimal transport loss to enable registration under large spatial displacement and extensive topological change between modalities.
2.Multimodal Integration: addressing the challenging task of image co-registration of CCTA and IVUS. Integrating information from both modality improves the ability to assess and diagnose coronary artery disease; and allows for a more complete analysis of both vessel morphology and plaque characteristics. 3.Robustness to Challenges: The method addresses challenges in CCTA-IVUS registration, specifically geometric deformation, calcifications, and lumen segmentation. The framework leverages topology-guided constraints and an optimal transport loss to enhance robustness in dealing with these inherently difficult aspects, which traditional registration approaches often struggle with. 4.Clinical Relevance: This improved registration accuracy and robustness has direct clinical relevance for CCTA-IVUS fusion images.
- 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.Limited Generalizability of Dataset: Although results were quantitatively validated on 47 pairs of CCTA-IVUS images, it would be highly desirable to validate results on a much larger, more diverse dataset from various centers. The number and the dataset of the images used in the study, although sufficient for the proof-of-concept study, may not cover all possible clinical scenarios in real clinical practice. 2.Computational Complexity and Efficiency: The optimization of the algorithm for multi-RTX 2080 Ti GPU calculations resulted in the method taking approximately 90 seconds per case on a single RTX 2080 Ti GPU, which, while feasible for research purposes, does not represent real-time clinical efficiency. 3.Lack of Ablation Studies on Specific Components: Although the full pipeline is described, it would give more strength to the paper if more extensive ablation studies could be performed to quantify the contribution of individual components such as the topology-guided constraints and unbalanced Sinkhorn optimal transport loss.
- 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?
1.A highly novel and methodically sound approach to a critical problem in medical image analysis. 2.The results are also supported by robust qualitative results on CC and LAD coronary arteries. 3.Though there is room for improvement, particularly regarding the diversity of the training datasets and runtime efficiency, the strength of this work lies in its contributions. 4.The paper is written clearly, and the algorithm is thoroughly explained.
- Reviewer confidence
Confident but not absolutely certain (3)
- [Post rebuttal] After reading the authors’ rebuttal, please state your final opinion of the paper.
Accept
- [Post rebuttal] Please justify your final decision from above.
I believe the rebuttal of this article can solve my problem, so I accept it.
Author Feedback
We thank the reviewers and AC for their constructive feedback, and are encouraged that all reviewers recognize the novelty and clinical relevance of TG-OT. Concerns are grouped below.
1.Apparent similarity of OT and Dice in Table 2 (R1, AC) R1 correctly notes that medians appear close in Table 2, while Fig. 3 shows a large gap. Aggregated metrics mask the relevant signal as cosine similarity has a heavily skewed distribution, and OT’s benefit is concentrated in outlier cases with severe rotational drift. Once predicted plaque/bifurcation locations stop overlapping, Dice’s gradient vanishes, and alignment cannot recover. OT supplies spatially informed gradients via the cylindrical geodesic cost (Eq. 6). This is reflected in (i) Wilcoxon signed-rank significance p<0.01 (Sec. 4), (ii) the Sc Q1 shift (0.92 vs. 0.90), and (iii) Fig. 3, which is a representative drift case.
2.”Removing dependency on intermediate segmentations” (R1) We thank R1 for catching this ambiguity. We refer to volumetric lumen/wall segmentations in the 3D MPR domain, which prior methods (Kadry et al., Li et al.) rely on and which are unreliable under IVUS acoustic shadowing. Our CNN instead predicts only feature presence along radial rays on the (θ, z) cylinder, a well-posed task even under shadowing, and a strictly weaker form of supervision than full volumetric segmentation. We will clarify this in the Discussion.
3.Notation and equation clarity (R2, AC) We note that the highlighted symbols are defined in the following places in the manuscript: M_w (warped moving) and F (fixed) are introduced at the start of Sec. 2.2.(θ, z) are angular and longitudinal coordinates on the unwound cylinder (Sec. 2.1, Fig. 2 caption). A in Eq. 2 is the lumen cross-sectional area, described directly after the equation. In Eq. 8, μ and ν are the marginal probability vectors from Eq. 4, where u and v are the Sinkhorn scaling vectors described after Eq. 7.We agree that A should be labeled explicitly and the kernel iteration made fully self-contained. We thank R2 for their notes on clarity and are committed to improving these passages in a camera-ready version.
4.Reproducibility (R2, R3) Code availability is noted in Sec. 2 of the manuscript. The reference appears as plain text rather than a hyperlink, which may have caused confusion. We will highlight this.
5.Dataset, runtime, ablations (R3) Dataset: We agree that the cohort is relatively small, and we aim to evaluate performance on a larger scale in future work. Nonetheless, the dataset contained considerable variation in plaque burden, pullback length and rotational drift, which motivated our OT formulation. Runtime: Our hardware resources were unfortunately limited to an RTX 2080Ti GPU. Modern hardware would yield runtime speedups, increasing feasibility of time-critical interventions. We would also like to note that for coronary analysis, CCTA-IVUS fusion is typically diagnostic-only with a longer time horizon, allowing for more processing time. Ablations: During method development, sanity checks on synthetic data with non-overlapping and mismatched topologies motivated the unbalanced OT formulation. The 8-page limit led us to prioritize comparison against established baselines on real data. Although we agree that a comprehensive ablation on this would benefit the paper, it cannot be included per the MICCAI rebuttal guidelines.
Optional concerns (R1) Frozen CNNs: CNNs are indeed pre-trained, and subsequently employed during the registration phase, which is why we note that feature detection is ‘integrated’ into the registration pipeline. We will gladly make this more explicit. Per-frame rotation correlation: This is exploited via a cumulative B-spline parameterization of θ (Sec. 2.2, Transformation model), enforcing smooth pullback twist informed by global rotational accumulation. Guidewire masking: Feature presence on (θ, z) remains detectable under shadowing, whereas NCC/NMI require pixel-level reliability.
Meta-Review
Meta-review #1
- Your recommendation
Invite for Rebuttal
- 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 proposes TG-OT, a topology-guided framework for CCTA–IVUS registration that formulates cross-modality alignment as an optimal transport problem, integrating feature detection and using an unbalanced Sinkhorn formulation to handle large spatial discrepancies and topological variations. Reviewers agree that the paper tackles an important and clinically relevant problem, and that the use of optimal transport for this challenging multimodal registration task is novel and technically interesting, with promising results demonstrated on a paired dataset of 47 patients. However, the reviews are highly conflicting. While one reviewer finds the paper to be high quality and clearly presented, another raises serious concerns about the clarity of the methodological description, particularly incomplete or undefined notation and insufficient explanation of key equations, which significantly limits readability and reproducibility. Additional concerns include limited evidence of clear superiority over simpler alternatives (e.g., Dice-based alignment), modest dataset size, lack of detailed ablation studies, and relatively high computational cost.
- After you have reviewed the rebuttal and updated reviews, please provide your recommendation based on all reviews and the authors’ rebuttal.
Reject
- Please justify your recommendation.
The author’s response still failed to address the reviewers’ concerns.
Meta-review #2
- After you have reviewed the rebuttal and updated reviews, please provide your recommendation based on all reviews and the authors’ rebuttal.
Accept
- Please justify your recommendation.
The three reviewers have acknowledged a highly novel approach and a challenging multi-modal registration problem. The method relies on an optimal transport formulation for registration, which I find worth discussing at MICCAI. The main remarks on R2 concern clarity. The authors should make an effort to ensure all clarity remarks are addressed in a revised version of the paper.
Meta-review #3
- After you have reviewed the rebuttal and updated reviews, please provide your recommendation based on all reviews and the authors’ rebuttal.
Accept
- Please justify your recommendation.
After carefully read the original submission, concerns from all reviewers, I share the opinion that this paper is of high quality and will be of interest to the MICCAI community. The I credited no new experiment, and the authors correctly declined the extra ablation R3 wanted, citing the rebuttal limit.
