Abstract

Microvascular obstruction (MVO) is a key prognostic factor in acute myocardial infarction, with affected patients experiencing higher mortality rates. Currently, late gadolinium enhancement cardiac magnetic resonance (CMR) is the gold standard for MVO identification. However, it is unsuitable for patients with renal impairment, who make up 20\% of all patients. Recent studies have demonstrated the feasibility of using non-contrast cine CMR to identify MVO. Despite this, existing methods struggle to effectively learn crucial motion features, as they implicitly model motion dynamics while overlooking regional wall motion abnormalities, which are important for MVO identification. To this end, we introduce a Dual Correlation-aware Mamba, which includes an Adjacent Frame Correlation (AFC) module and a Diastolic Frame Correlation (DFC) module to address these limitations. The AFC module calculates the correlations through adjacent frames to explicitly model the motion dynamics. The DFC module learns correlations between the diastolic frame and others. Leveraging the diastolic frame as a reference, this module highlights regional abnormalities and guides motion learning. Experimental results demonstrate that our method outperforms competing methods, potentially providing a non-contrast tool for MVO identification. The code is available at https://github.com/code-koukai/Dual-Correlation-Mamba.

Links to Paper and Supplementary Materials

Main Paper (Open Access Version): https://papers.miccai.org/miccai-2025/paper/0386_paper.pdf

SharedIt Link: Not yet available

SpringerLink (DOI): Not yet available

Supplementary Material: Not Submitted

Link to the Code Repository

https://github.com/code-koukai/Dual-Correlation-Mamba

Link to the Dataset(s)

N/A

BibTex

@InProceedings{YanYig_Dual_MICCAI2025,
        author = { Yan, Yige and Cheng, Jun and Yang, Xulei and Leng, Shuang and Tan, Ru San and Zhong, Liang and Rajapakse, Jagath C.},
        title = { { Dual Correlation-aware Mamba for Microvascular Obstruction Identification in Non-contrast Cine Cardiac Magnetic Resonance } },
        booktitle = {proceedings of Medical Image Computing and Computer Assisted Intervention -- MICCAI 2025},
        year = {2025},
        publisher = {Springer Nature Switzerland},
        volume = {LNCS 15960},
        month = {September},
        page = {177 -- 187}
}


Reviews

Review #1

  • Please describe the contribution of the paper

    The paper proposes a novel method, Dual Correlation-aware Mamba (DCM), comprising two modules—Adjacent Frame Correlation (AFC) and Diastolic Frame Correlation (DFC)—to explicitly model motion dynamics and highlight regional wall motion abnormalities (RWMA) for improved identification of Microvascular Obstruction (MVO) in non-contrast cine cardiac magnetic resonance (CMR).

  • 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 Innovative Framework: The introduction of AFC and DFC modules explicitly addresses limitations in capturing motion dynamics and RWMA, which are critical for identifying MVO. 2 Clinical Relevance: Addresses a significant clinical issue by providing a viable alternative for patients unable to undergo contrast-enhanced CMR due to renal impairment. 3 Strong Experimental Validation: Demonstrates clear performance improvements over state-of-the-art methods through comprehensive quantitative experiments, validated on a relevant clinical dataset. 4 Good Reproducibility Potential: Clearly outlines methods and promises code release upon acceptance, facilitating reproducibility.

  • 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. Clinical Generalizability: Although promising, the approach’s performance variability across different patient populations or imaging conditions remains unclear, potentially affecting broad clinical applicability.
    2. Computational Complexity: The computational load and inference time are not explicitly discussed, raising concerns regarding the practical deployment in clinical environments.
    3. Detailed Comparison: While the method surpasses existing approaches, a more nuanced analysis on specific failure cases would strengthen understanding of the proposed model’s limitations.
  • 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. Consider discussing the clinical integration feasibility explicitly, such as addressing real-time processing capabilities or computational requirements.
    2. Additional insights into why certain cases fail and potential avenues for improvement could significantly enhance the manuscript’s value.
  • 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?

    The paper presents a strong methodological innovation addressing a clinically relevant challenge. The approach is novel, well-motivated, and validated thoroughly. The main reservations include questions regarding computational feasibility in clinical settings, potential generalization issues, and the detailed examination of failure cases. Addressing these points during rebuttal could significantly reinforce acceptance.

  • 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

    The authors propose a machine learning model for microvascular obstruction (MVO) identification in non-contrast cine cardiac magnetic resonance, based on Mamba. They benchmark their approach against competing baselines and find superior performance across all metrics.

  • 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 authors address a relevant problem with state-of-the-art methodology.

  • 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.

    To explain their approach, the authors opt for graphical explanation rather than formulae. Due to this, I found it difficult to understand the approach but this might be personal preference.

  • 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 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
    • Introduction: The contributions paragraph makes it seem like adjacent and diastolic frame correlation modules are off-the-shelf methods. Please confirm that you came up with the modules yourself.
    • 2.1 Dual correlation-aware Mamba: In (1), what does the notation of parentheses following the $X_i$ mean? I assume PyTorch-style indexing, while $(\cdot)$ means “[:]” (explaining also the inner product) but please clarify this very non-standard notation.
    • 3 Experimental results: Please formally introduce the task of MVO identification: is it binary classification or also location (yellow boxes in Fig. 4)?
  • 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?
    • the authors benchmark their approach against competing baselines and find superior performance across all metrics
    • the authors address a relevant problem with state-of-the-art methodology
    • I found it difficult to understand the approach from graphical explanation alone
    • code – that the authors promise to publish – will clarify the approach
  • 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

    The authors propose a Dual Correlation-aware Mamba, which includes an Adjacent Frame Correlation (AFC) module and a Diastolic Frame Correlation (DFC) module, for MVO identification in cine CMR.

  • 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. MVO identification is both an important and challenging problem.
    2. The writing is fluent, and the motivation is clearly articulated.
    3. Using the diastolic frame as a global reference is reasonable.
    4. The ablation experiments are comprehensive and effectively demonstrate the effectiveness of the proposed modules.
  • 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. The authors analyze some hyperparameters, but it would be beneficial to also study others, such as the offset scale factor \alpha in the Offset Calculation module and the \lambda weight in the loss function.
    2. The paper shows the offsets learned by the AFC module and it would be better to present the interpolated images, which may further illuminate the method’s insights.
    3. How is the end diastolic frame identified in the dataset?
  • 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

    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?

    Please refer to the strengths and weaknesses sections.

  • 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 thank all reviewers for their constructive comments and thoughtful feedback. We address the main concerns below.


Reviewer #1

We appreciate the reviewer’s valuable suggestions regarding generalizability, computational complexity, and failure case analysis. These are important considerations for future work. Our current study was designed as an initial validation of the approach, and we acknowledge these limitations.


Reviewer #2

We thank the reviewer for the helpful feedback.

  • Regarding the clarity of the method, we have provided mathematical formulations in Section 2.1 (Eq. 1–3) for the AFC module, and described the DFC module in detail through text and illustrations. We will improve the clarity of the notation explanation in the final version.
  • The AFC and DFC modules are first proposed in this work. We will ensure this is clearly stated to avoid ambiguity.
  • The notation in Eq. (1) follows standard indexing conventions, where Xi(b,⋅,h,w)X_i(b, \cdot, h, w)Xi(b,⋅,h,w) refers to the feature vector at position (h, w). We will consider adding explanations for clarity.
  • The task is binary classification. The yellow bounding boxes in Figure 4 serve as visual markers to indicate the LGE-derived MVO regions at these locations, not as model predictions. The MVO regions themselves are not visually discernible in the displayed images.

Reviewer #3

We thank the reviewer for the detailed and constructive comments.

  • We focused on selected hyperparameters for ablation due to space limitations, but agree that further analysis could be informative.
  • We acknowledge that interpolated feature visualizations could offer additional insight into the AFC module’s effect and may consider including such illustrations.
  • These end-diastolic frames are automatically labeled during standard clinical cine CMR acquisition protocols. Our method leverages this readily available information as the reference frame, requiring no additional manual annotation. We will clarify this in the final version.




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