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

Abstract

Pretrained segmentation models for cardiac magnetic resonance imaging (MRI) struggle to generalize across different imaging sequences due to significant variations in image contrast. These variations arise from changes in imaging protocols, yet the same fundamental spin properties, including proton density, T1, and T2 values, govern all acquired images. With this core principle, we introduce Reverse Imaging, a novel physics-driven method for cardiac MRI data augmentation and domain adaptation to fundamentally solve the generalization problem. Our method reversely infers the underlying spin properties from observed cardiac MRI images, by solving ill-posed nonlinear inverse problems regularized by the prior distribution of spin properties. We acquire this “spin prior” by learning a generative diffusion model from the multiparametric SAturation-recovery single-SHot acquisition sequence (mSASHA) dataset, which offers joint cardiac T1 and T2 maps. Our method enables approximate but meaningful spin-property estimates from MR images, which provide interpretable “latent variable” that lead to highly flexible image synthesis of arbitrary novel sequences. We show that Reverse Imaging enables highly accurate segmentation across vastly different image contrasts and imaging protocols, realizing wide-spectrum generalization of cardiac MRI segmentation.

Links to Paper and Supplementary Materials

Main Paper (Open Access Version): https://papers.miccai.org/miccai-2025/paper/2605_paper.pdf

SharedIt Link: Not yet available

SpringerLink (DOI): Not yet available

Supplementary Material: Not Submitted

Link to the Code Repository

https://github.com/Ido-zh/cmr_reverse

Link to the Dataset(s)

N/A

BibTex

@InProceedings{ZhaYid_Reverse_MICCAI2025,
        author = { Zhao, Yidong and Kellman, Peter and Xue, Hui and Yang, Tongyun and Zhang, Yi and Han, Yuchi and Simonetti, Orlando and Tao, Qian},
        title = { { Reverse Imaging for Wide-spectrum Generalization of Cardiac MRI Segmentation } },
        booktitle = {proceedings of Medical Image Computing and Computer Assisted Intervention -- MICCAI 2025},
        year = {2025},
        publisher = {Springer Nature Switzerland},
        volume = {LNCS 15962},
        month = {September},
        page = {563 -- 573}
}

Reviews

Review #1

  • Please describe the contribution of the paper

    The authors propose a protocol translation framework based on a probabilistic model of the intrinsic T1/T2 parameters in CMR imaging, allowing for the translation between “biased” domains or projections such as GRE and SSFP. This is then applied to the problem of generalisation of segmentation, showing significant improvement.

  • 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.

    This paper is excellent, albeit a bit niche. Really enjoyed the rationale for the development of the methodology, from medical physics through to segmentation. Simple, elegant and powerful.

    It’s clear and easy to follow, plus the experiments are well thought out. Metrics are appropriate and standard, and the comparison with other standard methods highlight the advantage of this approach.

  • 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.

    Perhaps in a future paper, the authors could quantify some of the approach’s superiority in more clinical terms (e.g. volumes) which should be a direct translation of the improvements in segmentations. This is important for longitudinal studies where more advanced protocols might not have been available, e.g.: https://www.sciencedirect.com/science/article/pii/S1097664723007962

  • 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.

  • 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/2025/en/REVIEWER-GUIDELINES.html

    Some minor typos / style issues in the References, check for “mr” etc.

  • 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 — could be accepted, dependent on rebuttal

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

    I think the paper has a lot of merit but would perhaps thrive more in a more specialised medical physics conference. On the other hand, protocol conversion is something that many struggle with in the MIC community so I see value in having this here too, especially in that it can incentivise further secondary research from older datasets and recycling of valuable data.

  • 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



Review #2

  • Please describe the contribution of the paper

    The authors present a groundbreaking innovation in CMR segmentation which helps to mitigate the difference between different spin properties due to sequence design of MRI images. It enables accurate segmentation across vastly different imaging contrasts and imaging protocols.

  • 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 strength of this paper is that the authors has solid understanding of both CMR and GenAI. They designed this solution from fundamental CMR physics, infer the physics from images for a better generalization in image segmentation. With Reverse Imaging, they achieve high-quality zero-shot generalization of cardiac MRI segmentation, for wide spectrum of image types. Explicit mathematical expressions, benchmarks are given.

  • 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.

    One thing that the author might ignore or should improve is that - in many medical centers, there might not have mSASHA sequences installed in MRI scanners. Under such circumstances, could reverse imaging be applied in routine CMR sequences such as T1 mapping, T2 mapping, T2 star mapping etc?

  • 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.

  • 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/2025/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.

    (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?
    1. Approach innovation, good understanding of both CMR and AI
    2. Good writing including mathematical formulus, chart demonstration etc.
  • 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



Review #3

  • Please describe the contribution of the paper

    Contribution:

    1. This paper proposed a novel image translation framework that contains (1) a physics-based translation model to convert a MRI scan from source domain into tissue spin properties (PD, T1 and T2) and (2) a reverse imaging model, implemented with diffusion model, to convert the tissue spin properties into images of the target domain, and preform the modality translation from the source domain to target domain
    2. Experiments on cardiac image segmentation work is done. Using the proposed model the authors added the synthesized data as augmentation, and the results show using the synthesis data the performance get greatly improved.
  • 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 framework is very innovative and clear. The experimental results are also quite impressive
    2. The strong physics-based rules also decrease the demand for very large dataset.
  • 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.

    This paper lacks the comparison with other image translation works. Currently the baselines don’t contain any other image translation / synthesis method.

  • 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 authors claimed to release the source code and/or dataset upon acceptance of the submission.

  • 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/2025/en/REVIEWER-GUIDELINES.html
    1. It would make the paper even more clear if you can explain the tissue spin properties (PD, T1 and T2) a little bit more when you first mention them. I thought they were some scalar parameters and realize they are also image-like data when I saw Fig 2.
    2. It would be great to elaborate more on the RI-T2S approach. I don’t quite get it why using translated bSSFP images improve the performance compared with the baseline which use the untranslated bSSFP data.
  • 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.

    (6) Strong Accept — must be accepted due to excellence

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

    The approach is very innovative and clear with impressive experimental results

  • 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



Author Feedback

We sincerely thank the reviewers for their insightful and encouraging feedback. We are pleased that the innovation, clarity, and strong integration of physics-based principles in our framework were well received. We’re also grateful for the recognition of the experimental design, appropriate metrics, and clear presentation. Below, we address the questions and concerns of each reviewer.

  • Reviewer 1 *
  • We thank the reviewer for your suggestion regarding the explanation of PD/T1/T2. In the camera-ready version, we will clarify that these are voxel-wise quantitative maps.
  • The reviewer is correct that the baseline does not consider other image translation methods. MRI contrast varies widely across protocols and acquisition settings. Prior translation approaches often require separate models for each contrast or sequence, making them less flexible and dependent on extensive target-domain data. We view these as intrinsic limitations of those methods and therefore chose not to include direct comparisons.

  • Regarding the confusion of RI-T2S: The baseline model is trained on original, untranslated bSSFP cine images (the source domain). As such, it may not generalize well to test inputs acquired with different sequences (e.g., MOLLI). RI-T2S addresses this by translating test inputs from the target domain to the bSSFP cine contrast of the source domain. This reduces the domain gap and improves the baseline model’s performance, as the translated inputs more closely resemble the training data. An example of this effect is shown in Fig. 3(c). We will clarify it in the finalized version.

  • Reviewer 2 *

  • We thank the reviewer for highlighting a compelling application of our method—studying the bias in clinical terms across imaging protocols—which has significant impact in mitigating the variation of segmentation-derived parameters. We will explore this direction in future work. We will also address minor errors in the reference.

  • Reviewer 3 *
  • Regarding the availability of mSASHA scans: While our method leverages the advanced mSASHA sequence, we recognize it may not be widely available. To address this, we will release the reverse imaging results, namely, the inferred PD/T1/T2 maps of the ACDC dataset—which can serve as a prior on spin properties without requiring mSASHA acquisitions.

We appreciate the reviewers’ valuable feedback and constructive suggestions. We hope our responses have addressed your concerns and clarified the key aspects of our approach.



Meta-Review

Meta-review #1

  • Your recommendation

    Provisional Accept

  • If your recommendation is “Provisional Reject”, then summarize the factors that went into this decision. In case you deviate from the reviewers’ recommendations, explain in detail the reasons why. You do not need to provide a justification for a recommendation of “Provisional Accept” or “Invite for Rebuttal”.

    N/A



back to top