Paper Info Reviews Author Feedback Meta-Review Back to top
List of Papers Browse by Subject Areas Author List

Abstract

Dynamic late gadolinium enhancement MRI (dLGE-MRI) synthesized from cine-MRI is clinically valuable for dynamic myocardial lesion assessment. Integrating physics information into LGE synthesis has become a research focus; however, existing methods overlook the potential physics information gap induced by inaccurate estimation of distorted myocardial deformation, leading the model to deviate toward erroneous physics losses. To address this, we propose a method that transfers motion fields extracted from cine-MRI to static LGE image for dynamic sequence generation with noise-adaptive weighted physics constraints. Unlike existing approaches that treat physics loss as a hard constraint the model must fully comply with, we first convert the physics loss into a probabilistic constraint loss based on virtual likelihood. When motion errors are substantial, we appropriately reduce the weight of the physics loss to prevent model non-convergence caused by excessive physics noise; meanwhile, when deformation estimation stabilizes in local regions, the physics loss further guides the deformation field (driven by cine-MRI pixel data) to align with the physics priors characterized in the target static LGE image. Experiments on myocardial infarction datasets demonstrate that our method enhances lesion fidelity while maintaining motion tracking accuracy. Reverse validation confirms high consistency between static LGE-MRI generated by a generator trained on the synthesized dLGE-MRI and real static LGE-MRI, verifying the effectiveness of the synthesized images.

Links to Paper and Supplementary Materials

Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026/paper/2889_paper.pdf

SharedIt Link: Not yet available

SpringerLink (DOI): Not yet available

Supplementary Material: Not Submitted

Link to the Code Repository

https://github.com/Happy1ending/Cardiac-Motion-Transfer-Video

Link to the Dataset(s)

N/A

BibTex

@InProceedings{YuChe_Cardiac_MICCAI2026,
        author = { Yu, Chengjin AND Xu, Puneng AND Pu, Cailing AND Xi, Mengmei AND Xie, Ruonan AND Yan, Yuanting AND Ruan, Dongsheng AND Liu, Huafeng},
        title = { { Cardiac Motion Transfer with Adaptive Physics Constraint } },
        booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026},
        year = {2026},
        publisher = {Springer Nature Switzerland},
        volume = {LNCS 16890},
        month = {September},
        page = {pending}
}

Reviews

Review #1

  • Please describe the contribution of the paper

    The paper addresses a clinically interesting synthesis problem and proposes a plausible adaptive physics-weighting mechanism. The core idea is moderately novel, but the contribution is closer to an incremental methodological improvement than to a major conceptual advance.

  • 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 problem is clinically meaningful. Combining lesion visibility and motion dynamics is a useful direction for infarction assessment.
    • The adaptive weighting idea is technically sensible and better motivated than a fixed hard physics penalty.
    • The method is clearly structured and easy to follow at a high level.
    • The paper includes multiple baseline comparisons and ablations, and the reported gains are consistent across several metrics.
  • 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.
    • The novelty is moderate. The contribution is mainly an adaptive weighting of a physics-guided loss, not a fundamentally new synthesis framework.
    • The evaluation protocol is not sufficiently clear. In particular, the paper does not explain well how direct image-quality metrics are computed without real dynamic LGE ground truth.
    • Reverse validation is interesting but indirect, and it cannot fully replace direct validation of dynamic lesion fidelity.
    • The paper uses a large private clinical dataset but does not visibly report ethics approval, patient consent, or governance details.
    • The study is limited to 2D slices, with no discussion of through-plane motion or 3D consistency.
  • 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

    This is a thoughtful and technically solid paper on an interesting synthesis problem. I found the adaptive weighting of the physics constraint well motivated, and the reported gains are encouraging. I suggest strengthening the final version in three areas: first, clarify the evaluation protocol for the image-quality metrics and make the available reference targets more explicit; second, discuss the limitations of the current 2D setting, especially with respect to through-plane motion and 3D consistency; third, add a clear statement on ethics approval, consent, and private data handling. These additions would make the paper stronger and easier to assess.

  • 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 paper presents a meaningful methodological contribution to cardiac motion transfer by introducing an adaptive probabilistic physics constraint that is better aligned with noisy clinical motion estimates than a fixed hard penalty. The method is well structured, clinically relevant, and supported by a fairly broad experimental comparison with useful ablations. I find the adaptive weighting mechanism technically sound and the empirical improvements credible. At the same time, the contribution is more incremental than groundbreaking, and some aspects of the evaluation and dataset reporting should be clarified, especially the image-quality protocol and the ethical handling of the private data. Overall, I view the paper as marginally above the acceptance threshold because the idea is useful, the implementation appears careful, and the reported results are promising, even though some presentation and validation issues remain to be addressed.

  • 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.

    My comments have been addressed.



Review #2

  • Please describe the contribution of the paper

    This paper targets dynamic lesion assessment by transferring motion information from cine MRI to static LGE MRI. The authors propose a noise-adaptive physics loss that adjusts the weight of the physics constraint according to estimated motion errors. The method is evaluated on a private myocardial infarction dataset of 850 subjects, with comparisons against baseline methods and ablation studies. Based on the reported results, the framework appears to improve motion tracking relative to state-of-the-art methods. The authors also perform a reverse validation study showing high consistency between synthesized dLGE-MRI and real LGE-MRI images.

  • 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.
    • Clear motivation and clinical relevance: The paper addresses a clinically meaningful problem in dynamic lesion assessment by connecting cine MRI and LGE MRI.
    • Technical innovation: The proposed dLGE-MRI framework introduces a noise-adaptive physics weighting mechanism that adjusts the physics constraint using a virtual residual likelihood-based strategy.
    • Comprehensive evaluation: The paper includes comparisons against nine methods, ablation studies, and evaluation on a private myocardial infarction dataset of 850 subjects.
    • Good presentation: The manuscript is well organized and visually clear, with supporting figures and tables that make the experimental setup and results easy to interpret.
  • 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.
    • Missing justification: Since the LGE image already provides structural/anatomical prior to adapting cine-derived deformation to LGE deformation learning, the manuscript should better explain the role of myocardial mask. If feasible, an ablation study without the mask would help clarify its contribution.
    • Lack of limitations discussion: The manuscript would be stronger if it included a brief discussion of the framework’s limitations and assumptions. The ablation results also need more interpretation. For example, please provide intuition for why removing the physics constraint degrades image quality. There is no discussion section.
    • Limited temporal visualization of results: Since this is a dynamic LGE MRI problem, showing only selected frames makes it harder to assess motion preservation over the full cardiac cycle. If feasible within space constraints, additional temporal visualization or supplemental dLGE sequences would strengthen the evidence that the method preserves lesion-wall motion.
  • 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 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
    • Provide justification for using the myocardial mask, for example by citing relevant literature, or include an ablation study demonstrating its contribution.
    • Add a brief paragraph discussing the study limitations.
    • Better explain how temporal motion preservation is achieved. Include supplemental dLGE visualizations.
    • Minor: In Figure 2, use consistent font colors for both loss terms. Make “(I) L_data” red as well, consistent with the physics loss term.
  • 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 paper presents a technically interesting and clinically relevant approach to cardiac motion transfer from cine MRI to LGE MRI. Its main strengths are the adaptive physics loss formulation and the breadth of experimental evaluation. My main concerns relate to missing justification for design choices and the lack of discussion/interpretation. These concerns do not necessarily undermine the central contribution, but they do affect how easy it is to assess and interpret the method.

  • 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 authors’ rebuttal addresses my initial concerns by justifying the methodological choices and acknowledging the missing details of limitations that need to be included in the final version. Since these issues appear resolvable through the promised revisions and do not undermine the paper’s main contribution, I maintain my decision to accept.



Review #3

  • Please describe the contribution of the paper

    The paper introduces a method for transferring motion from Cine-MRI to (static) LGE-MRI to generate a dynamic sequence. The method enforces a probabilistic physics constraint that adjusts to motion errors using an adaptive weighting strategy. Experiments are conducted on one private dataset, and several state-of-the-art motion estimation and registration methods are compared. The results show quantitative and qualitative improvements.

  • 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 proposed method is compared to a wide range of methods from the literature.

  • 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.

    • Presentation clarity: Section 2 would benefit from improved clarity. The notation is often confusing, especially in Section 2.3, where several symbols appear without being properly introduced, making the methodology difficult to follow. Figure 2 is also unclear and does not help clarify the approach.

    • Experimental validation: The experimental validation relies solely on a private dataset, which makes it difficult to assess the practical relevance and generalizability of the proposed methodology. The ablation study for the weighting strategy shows only marginal qualitative and quantitative improvements; evaluating this component on additional datasets would help substantiate its impact. Furthermore, comparisons with other methods that incorporate physics‑based losses would strengthen the evidence supporting the method’s relevance.

  • 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 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

    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.

    (2) Reject — should be rejected, independent of rebuttal

  • Please justify your recommendation. What were the major factors that led you to your overall score for this paper?

    Conducting experiments on additional datasets would significantly strengthen the maturity of the work. In its current form, the validation is too limited, and the paper does not yet appear ready for publication.

  • 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 #4

  • Please describe the contribution of the paper

    This paper proposes a physics-constrained method for creating dynamic LGE slices by extracting motion from Cine-CMR and transferring it to a static LGE image. To overcome the limitations of standard physics-informed losses, the paper introduce a novel adaptive probabilistic weighting strategy that adjusts these constraints based on the reliability of the estimated motion. This treats the physics constraints as soft guidance by combining cine-derived motion, warping error, and physiological priors. Extensive ablation studies were conducted on the proposed optimization components, and a reverse-validation protocol was used to show that models trained on generated dynamic LGE CMR achieve similar performance to those trained on real static LGE-CMR data.

  • 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 soft weighting in motion optimization helps account for noisy motion and resulting training instability often seen in Cine-CMR.

    • The method is robust in estimating scar-involved myocardial deformation by embedding strain-tensor information into the constraint, outperforming several state-of-the-art methods.

    • Introducing the physics loss helps ensure biological plausibility of the generated motion, and the adaptive weighting shows that soft incorporation of these priors is important, as supported by ablation studies and qualitative results showing structural distortions and unstable training convergence when these components are removed.

  • 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.

    • Since each training iteration involves computing the stress, it would be useful to know how much additional computational overhead this method adds and whether it is easily adaptable to clinical settings.

    • The method is trained on 2D Cine motion sequences, which fail to capture the full 3D cardiac dynamics that are often affected by the presence and extent of scar tissue. Moreover, heart motion is inherently anisotropic, but the method assumes isotropy and a linear-elastic model.

    • The method focuses on the correlation between cardiac motion and scar alone, ignoring other pathologies such as inflammation versus infarction, where the motion is mechanically different.

  • 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
    • The reverse validation method could be explained better. The approach is lacking in several details and cannot be replicated with the current information.
  • 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.

    (3) Weak Reject — marginally below the acceptance threshold, but would not mind if accepted, dependent on rebuttal

  • Please justify your recommendation. What were the major factors that led you to your overall score for this paper?

    Although this paper proposes a new physics-informed framework to generate dynamic LGE CMR from Cine sequences, the downstream clinical applicability and ease of adaption is unclear. Moreover, the method would benefit from downstream analysis such as showing that the dynamic LGE sequences help in better disease detection or scar segmentation.

  • Reviewer confidence

    Confident but not absolutely certain (3)

  • [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.

    Although the rebuttal addresses concerns on computational time, the limitation of 2D modeling and linear elastic assumption could be backed by further analysis on the resulting error. More information on the “disease agnostic” loss and its performance on a wider range of pathologies along with the framework’s generalizability could make the paper’s claim stronger.



Author Feedback

We thank the meta-reviewer and three reviewers for their valuable feedback. Upon closer reading, we note that the meta-review may be more relevant to another work, and we kindly seek your further guidance. We appreciate the reviewers’ encouragement: “thoughtful and technically solid paper”[R1], “clinically meaningful”[R1,R2], “novel strategy”[R4], recognition of our experiments: “comprehensive evaluation”[R2], “compared to a wide range of methods”[R3], and presentation guidance: “easy to follow at a high level”[R1], “well organized”[R2], “would benefit from improved clarity” [R3]. The main concerns cover five aspects: validation, limitation discussion, method design, experimental details and presentation. Validation: 1)“evaluation protocol is not sufficiently clear”[R1]; 2)“reverse validation”[R1,R4]; 3)“private dataset”[R3]; “downstream clinical applicability”[R4]. A: To our knowledge, this is an early exploration of dynamic LGE synthesis, providing a potential means to assess dynamic motion of LGE lesions. But given the absence of real dynamic LGE in practice, direct evaluation metrics are unsuitable. 1) Referring to the evaluation protocols in motion transfer works LIA [25] (TPAMI) and LIA-X [26], we adopt the IS metric for identity consistency and the IQ metric for perceptual image quality based on content and local distortions. 2) We design a reverse validation strategy: a generative model is trained on the synthesized dynamic LGE, and its outputs are compared with real static LGE during testing. 3) We also thank R3, R4 for their insightful suggestions: -Public dataset: Existing public datasets (e.g., ACDC, MM) contain only cine-MRI for segmentation and motion tracking, but not direct validation of dynamic LGE generation. Even so, we have validated our physics-constrained module on motion tracking tasks using these datasets with better performance. -Downstream clinical value: Our clinical collaborators have similarly encouraged this direction. Preliminary results indicate that dynamic deformation features better assess infarct damage and remodeling, enabling more precise risk stratification. We will report these findings in the journal version. Limitations:1)“limited to 2D slices”[R1,R2], “a linear-elastic model”[R4]; 2)“temporal visualization”[R2]; 3)“ignoring other pathologies”[R4]. A: 1) We have added discussion: the current model is limited by 2D slices and the linear-elastic assumption for computational efficiency, while 3D hyperelastic modeling can better capture the physiological features of the heart. 2) We have uploaded the LGE videos to GitHub for release after acceptance. 3) We agree that motion patterns vary across cardiac pathologies. Our physics loss is grounded in classic mechanical laws and remains disease agnostic. Following R4’s suggestion, extending our method to other conditions in future work would be clinically meaningful. Method design: 1)“the role of myocardial mask”, “why removing the physics constraint degrades image quality”[R2]; 2)“comparisons with methods incorporate physics losses” [R3]. A: 1) We have added interpretation: The myocardial mask enhances boundary clarity in dynamic LGE synthesis. Without the physics constraint, some physically implausible noisy motions are difficult to avoid. 2) We apologize that our comparisons focus on image generation methods, overlooking those with physics-based losses. Nonetheless, the “w/o AdaW” setting with lower performance in our ablation study corresponds to fixed-weight methods. Experimental details: 1)“ethics approval”[R1]; 2)“additional computational”[R4]. A: 1) We have added the information on ethics approval, patient consent, and data governance details. 2) Compared with baseline, the additional training cost comes from PINN, leading to about 1.5× longer training time, and 1.2 s extra inference time. Presentation: “use consistent font colors”[R2], “benefit from improved clarity”[R3]. A: We have clarified the notations, figures and related issues.



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.

    Reviewers agreed that the paper addresses a clinically relevant and challenging problem in AF-centric multi-label ECG classification, with reported improvements over multiple baselines on both private and public datasets. Strengths include the imbalance/correlation-aware loss, adaptive multi-scale fusion, and generally supportive ablation results. However, concerns remain regarding incremental novelty, unclear positioning against prior imbalance-aware and multi-scale ECG methods, insufficient implementation details, and missing analyses on patient-level splits, uncertainty, and hyperparameter sensitivity. In the rebuttal, the authors should clarify these experimental details, strengthen comparison to task-specific imbalance/correlation methods, and better justify the novelty of each component.

  • 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 rebuttal provides reasonable clarifications on the use of indirect/reverse validation, explains why direct metrics are difficult in this setting, adds ethics and governance details, and acknowledges the 2D and pathology-generalization limitations.

    Limitations remain, especially the indirect nature of validation and limited evidence for broader clinical applicability. However, these are partly inherent to this emerging task. Therefore, I support acceptance.



Meta-review #2

  • 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.

    Key concerns persisted after rebuttal, including reliance on a single private dataset with no direct dynamic-LGE ground truth, only indirect reverse validation, marginal ablation gains, and unexamined effects of the 2D linear-elastic assumption. Given these unresolved limitations, the paper is not yet ready for publication and is recommended for rejection.



Meta-review #3

  • 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 paper addresses an important cardiac MRI problem and proposes a reasonable adaptive physics-weighted motion transfer method, with encouraging results on a large private dataset. However, the evidence remains insufficient for acceptance because the main dynamic LGE synthesis claim is validated only through indirect proxy metrics, without direct dynamic LGE ground truth, external validation, or comparison to relevant physics-informed cardiac motion methods. The rebuttal clarifies several points, but the remaining concerns require additional experiments and stronger validation rather than only camera-ready text changes.



back to top