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

Abstract

Head and neck squamous cell carcinoma has one of the highest rates of recurrence. Recurrence rates can be reduced by accurate localization of positive margins. While frozen section analysis of resected specimens provides accurate intraoperative margin assessment, complex 3D anatomy and significant shrinkage of resected specimens complicate margin relocation from the specimen back to the post-resection cavity. We propose a novel deformable registration framework that uses both the pre-resection external surface and the post-resection cavity of the specimen to incorporate thickness information. In tongue specimens, the proposed framework improved the target registration error (TRE) by up to 33% as compared to using the post-resection cavity alone. We found distinct deformation behaviors in skin, buccal, and tongue specimens, highlighting the need for tailored deformation strategies. Notably, tongue specimens hold the highest clinical need for improvement among head and neck specimens. To further aid intraoperative visualization, we also integrated this framework into an augmented reality-based guidance system. This system can automatically overlay the deformed 3D specimen mesh with positive margin annotation onto the post-resection cavity. The integrated system improved a surgeon and a trainee’s average relocation error from 9.8 mm to 4.8 mm in a pilot study. Our implementation code for AR guidance and generating the target point cloud is available at https://github.com/vu-maple-lab/Head-and-Neck-Tumor-Resection-Guidance.

Links to Paper and Supplementary Materials

Main Paper (Open Access Version): https://papers.miccai.org/miccai-2025/paper/5025_paper.pdf

SharedIt Link: Not yet available

SpringerLink (DOI): Not yet available

Supplementary Material: Not Submitted

Link to the Code Repository

https://github.com/vu-maple-lab/Head-and-Neck-Tumor-Resection-Guidance

Link to the Dataset(s)

N/A

BibTex

@InProceedings{YanQin_Augmented_MICCAI2025,
        author = { Yang, Qingyun and Li, Fangjie and Xu, Jiayi and Liu, Zixuan and Sridhar, Sindhura and Jin, Whitney and Du, Jennifer and Heiselman, Jon and Miga, Michael and Topf, Michael and Wu, Jie Ying},
        title = { { Augmented Reality-based Guidance with Deformable Registration in Head and Neck Tumor Resection } },
        booktitle = {proceedings of Medical Image Computing and Computer Assisted Intervention -- MICCAI 2025},
        year = {2025},
        publisher = {Springer Nature Switzerland},
        volume = {LNCS 15968},
        month = {September},

}

Reviews

Review #1

  • Please describe the contribution of the paper

    This paper presents an integrated augmented reality surgical guidance system for head and neck tumor resection that addresses a significant clinical challenge in oncological surgery. The core contribution is a novel Kelvinlet-based deformable registration framework that maps 3D models of resected tumor specimens to point clouds of the surgical resection site. This approach enables direct visualization of margin status overlaid on the surgical field via AR, potentially improving surgical precision compared to the current standard of verbal guidance. The complete system integrates a 3D scanner, RGBD camera, and HoloLens headset to provide surgeons with critical intraoperative information about margin status from frozen section analysis in an intuitive visual format. The authors also present a pilot study on a fully integrated AR guidance system including the deformable registration to demonstrate proof-of-concept of the feasibility and potential benefits of this system for accurate resection.

  • 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 addresses a genuine clinical need in oncological surgery, where margin status assessment is critical for patient outcomes. The introduction of AR visualization to replace verbal guidance represents a meaningful advance in surgical information delivery.
    2. The technical approach incorporates physically-motivated deformation models through Kelvinlet-based registration with strain energy regularization, which aligns well with the properties of biological tissues.
    3. Their comprehensive evaluation includes both technical validation of registration accuracy through leave-one-out cross-validation and a preliminary clinical evaluation of the end-to-end AR guidance system for which the registration process is designed.
    4. The authors demonstrate statistically significant improvements over rigid and similarity registration, and although they note no statistical significance compared to an existing state-of-the-art deformable registration approach, the results indicate a small improvement to both mean and maximum error. The end-to-end pilot study is a good inclusion to demonstrate the feasibility of the registration approach in an integrated system, as well as a proof of concept for the benefit of visual AR guidance over verbal instructions.
  • 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 lack of statistically significant improvement over existing deformable registration methods diminishes the technical novelty claim, despite reported reductions in mean and maximum errors.
    2. The pilot AR guidance study involves only two specimens and two surgeons, which severely limits the strength of conclusions about clinical utility and generalizability. Additionally, the task measured by the pilot study is a target localization task, so it is difficult to draw direct conclusions on the benefit of AR guidance to clinical use in terms of resection margin or procedure time. The authors thoroughly discuss the technical aspects of registration, but it would be helpful to include some context for the results shown in the pilot study to strengthen the conclusion and contributions of the paper.
    3. The paper appears to lack sufficient discussion of computational performance and real-time feasibility, which are critical considerations for intraoperative use.
    4. The evaluation focuses primarily on technical accuracy rather than workflow integration or usability considerations that would affect clinical adoption. There seems to be limited discussion of how the system handles registration uncertainties and how these uncertainties are communicated to the surgeon in the AR visualization.
  • 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

    While the authors present a novel deformable registration framework, given the lack of statistically significant improvement over existing methods the main contribution of the paper appears to lie more in the presentation of the first system for head and neck surgery with integrated deformable registration and AR guidance. Given this, the limited pilot study and lack of discussion on how the performance improvement on the task performed in the study may relate to surgical outcomes limits the conclusions that can be drawn. It would be helpful to include a focus in the discussion relating the outcomes of the work back to the original goal of improving resection margins, as this is not entirely clear in the manuscript as it is presented.

    Also, it is unclear what the need or benefit of the novel registration framework is if it does not significantly outperform existing methods. I would like further justification of this need, whether it be reliability, runtime performance, etc. to fully understand the value of the contributions made on the methodology side.

  • 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 — could be rejected, dependent on rebuttal

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

    This paper presents an innovative application of deformable registration in a potentially impactful AR system for oncological surgery. The technical approach is sound, incorporating physically-motivated deformation models appropriate for biological tissues. While improvements over alternative deformable methods are not statistically significant, the preliminary evaluation shows promise in improving target localization in a small pilot study. The work bridges technical innovation and clinical application, with evaluation that includes both technical validation and preliminary surgical assessment. However, the clinical evaluation is limited to a small pilot study focused only on fiducial localization rather than actual tumor resection, limiting any conclusions that can be made on the impact of AR guidance on resection margins. Additionally, the paper would benefit from addressing computational performance for real-time use and elaborating on uncertainty visualization. This research represents a promising step toward applying AR in oncological surgery, with a well-conceived registration approach tailored to the specific challenges of head and neck tumor resection. However, further justification of the need for the novel registration method over existing methods, as well as additional clinically relevant discussion on the pilot study results are needed to better establish the contributions made in this paper.

  • 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 rebuttal effectively addresses my concerns and justifies acceptance, particularly within the CAI track. The authors clarify that while the registration algorithm is based on prior work, its adaptation to tongue specimens constitutes a meaningful contribution. Although statistical improvements were modest, the authors commit to the inclusion of qualitative results, such as deformed mesh visualizations to demonstrate improved anatomical plausibility and alignment fidelity where standard methods fail.

    The authors rebuttal addresses my concerns over missing context of clinical validity from the limited user study presented here. I was looking for a stronger justification of how the task performed in the study related to potential clinical utility and feel the authors have addressed this in their rebuttal to a sufficient degree.

    The authors also address workflow concerns, noting minimal intraoperative time and compatibility with standard hardware. Their discussion of fiducial placement acknowledges current limitations while pointing to straightforward paths for automation. The open-sourcing of key components further supports reproducibility.

    Overall, the rebuttal reinforces the paper’s value as a system integration effort with practical clinical implications. I now recommend acceptance.



Review #2

  • Please describe the contribution of the paper

    This paper addresses an important and interesting problem for head and neck tumor resection. It proposes a registration approach, using the surface of the pre-resection and the post-resection location of the specimen. The main contribution relies in the experimental setup and study for this specific application in an augmented reality context.

  • 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.
    • A new way to consider tumor resection issues.
    • Experimental set-up
    • Experimental results
  • 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 general pipeline is interesting and shows some novelty. It relies mostly on a judicious choice of existing methods and the methodology is therefore not really original. Furthermore, it does not take into account the deformation between pre and post resection surfaces, which seems to be a strong limitations (mentioned by the authors in the discussion). However, thus study could be considered as a first step.

    The registration relies on four fiducial markers. Although this is theoretically enough to estimate simple transformations, it might not be robust to imprecision in the definition of these markers nor to outliers. This should be discussed. Furthermore these markers are defined manually. Could this step be potentially automated?

    The way thickness is taken into account is not clearly explained. Results are promising, yet preliminary. They would benefit from a deeper analysis to better explain them, in particular to better understand their varied quality depending on the specimens. Illustrations using images and meshes could be added for a qualitative assessment. Table 2 lacks comments as well.

    The way the proposed approach would help reducing recurrence rates should be better explained.

  • 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.

  • 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

    The reference list could be presented in a more homogeneous way.

  • 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 — could be rejected, dependent on rebuttal

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

    See detailed comments

  • 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.

    While the authors partially answered the reviewers’ concerns, some important issues remain, such as the number of landmarks and the potential deformations. While the application is interesting, the methodological novelty is limited. The paper still presents an interesting pipeline for a dedicated application.



Review #3

  • Please describe the contribution of the paper

    The paper presents a deformable registration framework tailored for head and neck tumor resection, specifically addressing the challenge of accurately relocating intraoperative positive margins from resected specimens back to the surgical site. By incorporating both pre-resection and post-resection surface scans, the method provides improved constraints for tissue deformation modeling. The authors further integrate this framework with an AR-based guidance system, enabling automatic overlay of the deformed specimen mesh onto the patient using marker-based registration. Experiments on cadaver heads demonstrate improved registration accuracy, particularly for anatomically complex specimens like the tongue, and show promise for enhancing intraoperative margin localization.

  • 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 addresses a practical and underexplored challenge in surgical AR—accurately relocating and visualizing positive margins from resected specimens during head and neck cancer surgery.

    2. The proposed registration framework is well-motivated, introducing the use of both pre- and post-resection intraoperative surfaces as dual constraints for deformation, which improves robustness over prior methods that rely on post-resection geometry alone.

    3. The integration with an AR guidance system is well-executed and fully automatic once fiducials are defined, showing clear clinical relevance and real-world applicability in cadaver experiments.

  • 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 method relies on manually identified fiducial landmarks, which introduces user dependence and limits automation.

    The core registration framework builds on existing elastic deformation models, and the methodological novelty is relatively limited.

    The experiments are performed on a small number of cadaver cases, without validation under live surgical conditions.

    The paper does not provide code or implementation details, making it difficult to reproduce or adapt the method for future research.

  • 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?

    I gave a weak accept to this paper. While the methodological novelty is somewhat limited, the proposed framework addresses a practical and clinically relevant challenge in surgical margin relocation and visualization. The use of dual intraoperative surfaces and AR integration is well-motivated and demonstrates promise, particularly in anatomically complex regions like the tongue. However, the work would be stronger with the inclusion of code or implementation resources to support reproducibility and future adoption.

  • 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.

    This work addresses a clinically important problem of positive margin relocation in head and neck surgery using AR and deformable registration. Despite limited sample size, the approach shows promise, particularly in tongue specimens, and the integration of AR into the surgical workflow demonstrates innovation and practical feasibility.



Author Feedback

We introduce an innovative (R1) and well-motivated (R2) application of deformable registration, which “aligns well with the properties of biological tissues” (R1). We demonstrate its utility as an augmented reality (AR)–based guidance system. Our system is the first of its kind, tailored to the underexplored challenges of head and neck oncological surgery (R3), addressing a critical need (R1, R2, R3). In this rebuttal, we will focus on novelty, clinical significance of the user study, workflow, and reproducibility.
1) Novelty Although our pipeline uses existing registration methods (R2, R3), it is tailored for the clinically challenging tongue specimens, evidenced by our proposed method outperforming the existing one in all tongue cases, with reduced error variance. This supports the need for a novel method for determining the boundary conditions for tongue specimens and the significance of our registration method despite the lack of statistical improvements (R1). We will also add more qualitative results (e.g. deformed mesh figures) (R3) to support our analysis. Moreover, we invite the reviewers to view our overall system integration as a novel clinical application (R1, R2, R3) in AR-guided tumor resection. Its plug-and-play nature allows seamless integration with various deformable registration algorithms. Ultimately, our study shows that augmented reality paired with deformable registration facilitates more accurate relocation and can aid complete tumor resection. 2) Clinical significance of the user study We will add a discussion of the user study task results (e.g. table 2) (R3), particularly its relevance to tumor resection (R1) and recurrence rates reduction (R3). Briefly, the recurrence rate is inversely correlated with re-resection performance [1], which is dependent on margin relocation accuracy [2]. In our task, a fiducial target replaces the positive margin finding by pathology to be relocated. Although the number of cases is limited (R1, R2), the pilot study demonstrates “real-world applicability in cadaver experiments” (R2) and provides promising results to power a larger clinical trial. 3) Workflow We will clarify concerns about our system’s integration into clinical workflows (R1). Our pipeline, running on a regular laptop, takes 30 minutes. Only 10 minutes of direct access to the specimen and surgical site are needed for 3D scanning with mobile scanners. The remaining process can occur remotely while pathology analyses the specimen, minimizing workflow interruption. R2 and R3 noted that manually defined fiducial points can lead to user dependence (R2), imprecision (R3), and limited pipeline automation. Manual fiducial localization can have low error and inter-user variance. [3] While we used manual fiducial detection, our framework readily supports automation (e.g., reflective surgical clips detectable via mature algorithms [4]). These were expensive for our proof-of-concept study but are commonly available in clinical settings. Thus, the current manual approach does not detract the pipeline’s overall contribution. R3 also noted the use of only four fiducials, which, while minimal, was necessitated by the limited space on specimens. 4) Reproducibility Lastly, we will open-source our implementation for AR guidance and generating the target point cloud for reproducibility upon acceptance of this paper. The implementation details of the Kelvinlets Model can be found in [5]. [1] Positive Surgical Margins in the 10 Most Common Solid Cancers,2018 [2] How far are we off? Analyzing the accuracy of surgical margin relocation in the head and neck,2024 [3] The effect of repetitive manual fiducial localization on target localization in image space,2007 [4] Detection and correction of specular reflections for automatic surgical tool segmentation in thoracoscopic images,2011 [5] Comparing regularized Kelvinlet functions and the finite element method for registration of medical images to sparse organ data,2024



Meta-Review

Meta-review #1

  • Your recommendation

    Invite for Rebuttal

  • 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

  • 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. You may optionally write justifications for ‘accepts’, but are expected to write a justification for ‘rejects’

    N/A



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. You may optionally write justifications for ‘accepts’, but are expected to write a justification for ‘rejects’

    Despite the concerns raised by R3 post-rebuttal, this AC has reached the decision of Accept. For a CAI-focused paper, the noted “innovation” should be a minor concern as long as clinical significance is demonstrated.



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. You may optionally write justifications for ‘accepts’, but are expected to write a justification for ‘rejects’

    N/A



back to top