Abstract

Multimodal medical imaging provides critical data for the early diagnosis and clinical management of clear cell renal cell carcinoma (ccRCC). However, early prediction primarily relies on computed tomography (CT), while whole-slide images (WSI) are often unavailable. Consequently, developing a model that can be trained on multimodal data and make predictions using single-modality data is essential. In this paper, we propose a multimodal hypergraph guide learning framework for non-invasive ccRCC survival prediction. First, we propose a patch-aware global hypergraph computation (PAGHC) module, including a hypergraph diffusion step for capturing correlational structure information and a control step to generate stable WSI semantic embeddings. These WSI semantic embeddings are then used to guide a cross-view fusion method, forming the hypergraph WSI-guided cross-view fusion (HWCVF) to generate CT semantic embeddings, improving single-modality performance in inference. We validate our proposed method on three ccRCC datasets, and quantitative results demonstrate a significant improvement in C-index, outperforming state-of-the-art methods. The source code is available in https://github.com/iMoonLab/PAGHC.

Links to Paper and Supplementary Materials

Main Paper (Open Access Version): https://papers.miccai.org/miccai-2025/paper/1586_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{YanJie_Multimodal_MICCAI2025,
        author = { Yan, Jielong and Han, Xiangmin and Zhao, Jieyi and Gao, Yue},
        title = { { Multimodal Hypergraph Guide Learning for Non-Invasive ccRCC Survival Prediction } },
        booktitle = {proceedings of Medical Image Computing and Computer Assisted Intervention -- MICCAI 2025},
        year = {2025},
        publisher = {Springer Nature Switzerland},
        volume = {LNCS 15971},
        month = {September},
        page = {510 -- 519}
}


Reviews

Review #1

  • Please describe the contribution of the paper
    • Introduces a patch-aware global hypergraph computation module to generate stable whole slide image (WSI) embeddings.
    • Proposes a cross-view fusion method that, during training, is guided by WSI data to enable accurate survival prediction based solely on CT data during inference.
  • 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.
    • Effectively leverages multimodal information during training while allowing for non-invasive CT-based survival prediction, which has clear clinical significance.
    • Presents an innovative approach that is both conceptually and practically straightforward.
    • The manuscript is well-organized and visually appealing, with high-quality figures.
  • 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.
    • It remains unclear whether variations in the number of sampled patches significantly affect the results, and what approach should be taken if a WSI contains fewer than 2000 patches.
    • The specifics of the five-fold cross-validation process are not sufficiently detailed, particularly regarding whether the best validation checkpoint or the final checkpoint is used.
    • While the paper highlights the use of pathological data to guide the CT model during training, there is a lack of experimental comparison with a CT model that does not incorporate this multimodal alignment, which could better demonstrate the effectiveness of the proposed guidance.
    • Figure 2 omits the Kaplan–Meier (KM) curve for the H2 group. Including KM curves with median survival times would offer a clearer depiction of differences in patient outcomes, which is crucial for clinical interpretation.
    • In Table 4, the differences between the “angle” and “distance” settings in the guided loss configuration are not clearly explained.
  • 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.

  • 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.

    (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 recommendation is based on the strengths and weaknesses outlined above, with the expectation that the authors will address and improve the identified weaknesses.

  • 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 proposed a multimodal hypergraph guide learning framework for non-invasive clear cell renal cell carcinoma (ccRCC) survival prediction. A patch-aware global hypergraph computation (PAGHC) module is proposed to capture correlational structure information and semantic embeddings. The proposed model achieved state-of-the-art performance on one public datasets and two private 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.
    1. The paper is easy to read and follow. The framework details are well described.
    2. A novel patch-aware global hypergraph computation module and a cross-view fusion method are proposed in this paper. 3.The proposed model achieves state-of-the-art (SOTA) segmentation performance among one public and two private datasets.
  • 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. Large-scale paired WSI and CT data are not easily obtained. Is there more analysis for missing data scenarios?
    2. More experiments and analysis can be provided in the Ablation Studies section to evaluate the effectiveness and robustness of proposed model. For example, is it possible to evaluate confidence intervals or significance testing?
    3. Some qualitative results could be shown. For example, how the whole slide images and ct scan images quality will influence the performance.
    4. More datasets details can be showed for private datasets.
  • 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.

  • 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.

    (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?

    My overall score is based on the above major strengths and major weaknesses.

  • 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 #3

  • Please describe the contribution of the paper

    This paper proposes a multimodal hypergraph-guided learning framework for noninvasive ccRCC survival prediction and introduces a hybrid approach. Firstly, a patch-aware global hypergraph is created to generate stable WSI semantic embeddings for accurate survival prediction. Then, CT-based survival prediction is performed using WSI semantic embeddings. This enables WSI image-less ccRCC survival prediction from CT image. Hypergraph-guided approach is used here. Experimental results show very promising results in survival prediction.

  • 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’s strong point is training the model from a WSI image and using it for CT image-based prediction. The hyper-graph approach is utilized for stable WSI semantic embeddings for accurate survival prediction. WSI-based model is used to train the CT-based model for enabling CT-based survival prediction. Knowledge distillation from WSI to CT is achieved.

  • 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.

    It is hard to see the contribution of the hypergraph in the final result. This may appear in ablation study, but all methods are represented in the method name, not by actual method.

    In this paper, the authors say, “CT is a non-invasive 3D imaging modality”. However, X-ray is not invasive procedure due to radiation, especially in 3D X-ray CT scan.

  • 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 provide sufficient information for reproducibility.

  • 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.

    (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?

    It is good idea to use the model trained by WSI for prediction for CT image.

  • 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

Thank you to the reviewers for your valuable feedback.

  1. The number of patches affects WSI sampling coverage. Generally, more patches improve performance at the cost of increased computation. When fewer than 2,000 patches are contained in the WSI, we allow them to overlap to ensure adequate coverage.
  2. In Table 2, PAGHC and HGSurvNet employ hypergraph-based methods, whereas DeepGraphSurv and Patch-GCN use graph-based approaches. Since each hyperedge can connect two or more vertices, hypergraphs naturally generalize graphs and capture higher-order relationships among patches, demonstrating their experimental advantage.
  3. Due to space constraints in MICCAI, we will expand our presentation of the five-fold cross-validation setup, clinical KM curves, ablation studies, missing-data analyses, image-quality impact assessments, and dataset details to more fully showcase our results. Again, we appreciate the reviewers’ valuable time and constructive suggestions.




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