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

Abstract

Liver surface landmark detection is a fundamental prerequisite for anatomical guidance in laparoscopic liver surgery. However, it remains unreliable in practice due to two pervasive challenges: illumination attenuation in underexposed regions and the structural mismatch between pixel-wise localization and continuous curvilinear geometry. To address these limitations, we propose A2ONet, an attenuation-resilient alternating optimization network for robust liver landmark detection. To mitigate illumination attenuation, A2ONet embraces an illumination field compensation (IFC) block that adaptively enhances dark regions while preserving structural consistency. Meanwhile, we introduce a lightweight frequency–orientation selective filter (FOSF) to suppress repetitive texture interference and preserve salient curvilinear cues. Building upon these resilient representations, we design an alternating seg–curve optimization (ASCO) decoder that iteratively couples dense segmentation with explicit curve modeling, enabling mutual guidance to optimize both structural continuity and endpoint localization. Extensive evaluations on L3D-2K, L3D, and P2ILF demonstrate consistent improvements over competitive methods, establishing a more reliable foundation for intraoperative anatomy guidance. Our code is available at https://github.com/hyperiondk115/A2ONet.

Links to Paper and Supplementary Materials

Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026/paper/1734_paper.pdf

SharedIt Link: Not yet available

SpringerLink (DOI): Not yet available

Supplementary Material: Not Submitted

Link to the Code Repository

https://github.com/hyperiondk115/A2ONet

Link to the Dataset(s)

P2ILF dataset: https://p2ilf.grand-challenge.org/ L3D dataset: https://github.com/PJLallen/D2GPLand L3D-2K dataset: https://github.com/cuiruize/D2GPLand-Plus

BibTex

@InProceedings{LiuLan_AttenuationResilient_MICCAI2026,
        author = { Liu, Lanqing AND Cui, Ruize AND Pei, Jialun AND Guo, Diandian AND So, Tiffany Y. AND Heng, Pheng-Ann AND Qin, Jing},
        title = { { Attenuation-Resilient Alternating Optimization for Laparoscopic Liver Landmark Detection } },
        booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026},
        year = {2026},
        publisher = {Springer Nature Switzerland},
        volume = {LNCS 16893},
        month = {September},
        page = {pending}
}

Reviews

Review #1

  • Please describe the contribution of the paper

    The paper proposes A2ONet, an attenuation-resilient framework for laparoscopic liver landmark detection. The method addresses two key challenges: (1) illumination degradation in laparoscopic images and (2) the structural mismatch between pixel-wise segmentation and continuous curvilinear landmarks.

    The approach integrates three main components:

    • an illumination field compensation (IFC) module to normalize low-light regions,
    • a frequency–orientation selective filter (FOSF) to enhance curvilinear structures and suppress texture noise,
    • an alternating segmentation–curve optimization (ASCO) decoder that jointly models dense segmentation and parametric curve representations.

    The method is evaluated on three datasets (L3D, L3D-2K, P2ILF) and shows consistent improvements over existing approaches.

  • 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 paper tackles an important and clinically relevant problem in laparoscopic liver surgery.
    • The motivation is clear and well-justified, particularly regarding illumination issues and structural continuity.
    • The proposed architecture is coherent and integrates complementary ideas (appearance enhancement + geometric modeling).
    • The alternating optimization between segmentation and curve modeling is interesting and addresses a known limitation of segmentation-based approaches.
    • The method is evaluated on multiple datasets with consistent improvements.
    • The ablation study is thorough and demonstrates the contribution of individual components.
    • The paper is technically sound and well-positioned within 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.
    • The performance gains over strong baselines (e. g. , D2GPLand+) are relatively modest (~1–2%), raising questions about practical significance.
    • The proposed framework is complex, combining multiple modules (IFC, FOSF, ASCO, depth estimation, SAM encoder), but lacks analysis of computational cost and inference time.
    • There is no discussion on real-time applicability, which is critical for intraoperative surgical use.
    • The evaluation is limited to segmentation metrics (DSC, IoU, ASSD), without demonstrating clinical impact or integration into AR navigation systems.
    • Curve-based methods are excluded from comparison, which is questionable given the explicit curve modeling contribution.
    • The illumination compensation component appears incremental relative to existing enhancement approaches.
    • The paper lacks analysis of failure cases and robustness under challenging conditions (e. g. , occlusion, bleeding, specular reflections).
  • 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.

  • 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 paper presents a well-structured and technically sound approach to liver landmark detection. However, to strengthen the contribution, the authors should provide a deeper analysis of computational efficiency and real-time feasibility, as well as a clearer demonstration of clinical relevance. Including comparisons with curve-based methods and providing failure case analysis would further improve the work.

  • 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 addresses a relevant problem and proposes a technically sound solution with consistent improvements across multiple datasets. The integration of illumination compensation and curve-based modeling is meaningful and well-motivated. However, the contribution appears somewhat incremental, and the performance gains over strong baselines are limited.

    More importantly, the lack of computational analysis and clinical validation reduces the practical impact of the work, especially in the context of real-time surgical applications. Overall, the paper is slightly above the acceptance threshold but would benefit from stronger evidence of impact and efficiency.

  • 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 paper proposes a method to detect liver landmarks in laparoscopic images. The method normalizes the intensity map as part of the IFC block to account for uneven illumination in input image. The proposed method leverages Gabor filters to learn the contours and suppress the texture details. The extracted features are used to estimate initial curves and then refine them further with the help of segmentation groundtruth.

  • 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 presents and explains the method clearly. 2.Ablations in Tab2, clearly show the importance of various components in the method. 3.The scope of the current work is stated well to a certain extent. 4.Comparison to piror-works and networks is well established in Tab1 and visually supposed in Fig1,3.

  • 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.

    There as such no weakness of the paper, however some sections need clarifications which are mentioned in the additional comments section below.

  • 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.

  • 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

    Following are few questions, clarifications and suggestions on the paper: 1.The paper emphasizes on the attenuation resilient feature. However, it would have been better if the impact of this feature was clearly shown in the paper. For example, Fig 1 already shows comparisons of predictions from various method and Fig 3 also show the same but for more image with less illumination. The reviewer thinks, in Fig 3, same image with different illumination settings could be simulated to show the impact of the IFC block visually where ideally the predictions should be the same for all illumination settings.

    2.It is unclear as to what a logit function is in section 2.3.Kindly clarify that or use the formula instead of the function name.

    3.There are parameters such as K (from top-K), S which are chosen as per user’s prerogative. However, ideally an ablation on the same should be done. It is encourages that these aspects are covered in the extended version of the paper, preferably for a journal.

    4.On the same point as before, as S increases ideally there would be an S where the curvature updates converge. It is an important aspect of the alternate optimization algorithm to show this convergence and parameters that influence this convergence.

    5.Although the paper slightly touches upon the limitations on the scope of this work, it is preferred to have a subsection explaining in detail the limitations of the method and the limitations in scope of the work. In that way, the readers would have a clear idea of the coverage of the work and what else is to be worked on in future.

  • 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 has all aspects covered well, the topic is novel, there is sufficient depth and clarity in presentation. However, there is one aspect in the claims related to illumination which is not well justified in the paper. It is understandable that this might be due to the paper outline and page restrictions. Hence, if this issue is addressed as limitation of the work in the paper by introducing a limitations section then the paper could be accepted in its current format based on its merits.

  • 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

    This paper proposes A2ONet for laparoscopic liver landmark detection, addressing two challenges: (1) illumination attenuation in underexposed surgical regions and (2) the structural mismatch between pixel-wise segmentation and continuous curvilinear landmark geometry. The architecture integrates three modules: an illumination field compensation (IFC) block based on reflectance-illumination decomposition with a learnable temperature parameter; a frequency–orientation selective filter (FOSF) using Haar wavelet decomposition and Gabor filtering to suppress texture interference and enhance curvilinear structures; and an alternating seg–curve optimization (ASCO) decoder that iteratively couples dense segmentation with Bezier curve modeling through cross-attention-based curve-to-mask fusion. The model uses ResNet-34 for RGB encoding and a frozen SAM-b encoder for depth features from Depth Anything V2.Evaluated on L3D-2K (2,000 images), L3D (1,152 images), and P2ILF (183 images), A2ONet reports DSC of 53.51% / 65.31% / 43.02% respectively, improving over the previous best (D2GPLand+) by 1.85% / 0.69% / 0.87% DSC.

  • 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.

    S1.Well-motivated and clearly structured problem formulation. The paper identifies two concrete, practically relevant failure modes — illumination attenuation and the localization-continuity tradeoff — and designs dedicated modules for each. The introduction provides clear visual examples (Fig. 1) of blue-circle attenuation errors, yellow-circle segmentation false positives, and pink-circle endpoint localization failures that motivate the approach.

    S2.Technically sound module design with principled foundations. Each component is technically well-motivated and clearly described: the IFC block builds on a reflectance-illumination decomposition assumption (akin to Retinex, though the paper does not use that term) with a learnable intensity collapse (Eq. 1); the FOSF module combines Haar wavelet frequency analysis with Gabor orientation filtering; and the ASCO decoder models landmarks as Bezier curves with soft rasterization (Eq. 4) and differentiable control-point updates (Eq. 5). The interplay is well-articulated: IFC addresses appearance, FOSF addresses feature-level texture suppression, and ASCO addresses the output-level geometry gap.

    S3.Thorough ablation study demonstrating component complementarity. Table 2 systematically isolates the contribution of each module (IFC, L_illum, FOSF, ASCO, L_curve) through six progressive variants, showing that each addition yields measurable improvement. Adding L_illum to IFC (V1→V2) yields the largest single-step DSC gain (+2.22%), and further combining with FOSF (V2→V4: +2.03%) produces a substantial cumulative improvement, validating the complementarity of illumination compensation and frequency filtering.

    S4.Comprehensive baseline comparison. The paper benchmarks against 16 methods across three categories (general segmentation, SAM-based adaptation, task-specific detection), providing a well-contextualised evaluation landscape. The consistent improvement across all three datasets and all three metrics (DSC, IoU, ASSD) lends credibility to the approach.

    S5.Clean mathematical exposition. The equations are well-defined and the notation is consistent throughout. The paper makes the training pipeline clear (end-to-end with joint supervision, Eq. 6) and specifies all key hyperparameters.

  • 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.

    W1.No uncertainty estimates or statistical significance tests. (General Review Consideration §5, §6) The paper reports single-point performance numbers throughout Tables 1 and 2 with no standard deviations, confidence intervals, or statistical tests. Given that the improvements over D2GPLand+ are modest — 1.85% DSC on L3D-2K, 0.69% DSC on L3D, and 0.87% DSC on P2ILF — it is not possible to assess whether these gains are statistically meaningful. The number of training runs, random seed policy, and variance across seeds are not disclosed. Without at least a paired test or multi-seed standard deviation, the evidence for superiority over D2GPLand+ is incomplete.

    W2.Notable validation–test performance gap unexplained. The ablation study (Table 2) is conducted on the L3D-2K validation set, where the full A2ONet achieves DSC 45.16% / IoU 33.76%. However, Table 1 reports DSC 53.51% / IoU 38.32% on the L3D-2K test set — an 8.35% DSC gap favouring the test set. This is unusual (typically validation and test performance are comparable or test is slightly lower) and warrants explanation. Possible causes include different data characteristics between splits or overfitting to validation-set early stopping, but the paper does not discuss this discrepancy.

    W3.No inference time or model complexity analysis. (Clinical significance — General Review Consideration §1) The paper motivates A2ONet for intraoperative anatomy guidance and AR-guided laparoscopic interventions, so reporting runtime, parameter count, or FLOPs would help assess practical deployability. Yet no inference speed (FPS), parameter count, or FLOPs are reported. The architecture includes a ResNet-34 encoder, a frozen SAM-b encoder, Depth Anything V2 for monocular depth estimation, Haar wavelet transforms, Gabor filtering, and a multi-stage ASCO decoder with cross-attention — these components together likely carry a non-trivial computational cost. Without runtime analysis, the practical applicability for the stated clinical use case remains unclear.

    W4.Minimal discussion of limitations and failure cases. (General Review Consideration §8) The conclusion mentions no limitations at all. A brief sentence states A2ONet “provides a stable and robust prerequisite for AR-guided navigation,” but no failure case analysis is provided. Given the moderate absolute performance on L3D-2K (53.51% DSC), understanding where and why the method fails — for example across landmark types, illumination conditions, or unusual anatomy — would be valuable.

    W5.P2ILF evaluation protocol is non-standard. The paper acknowledges that P2ILF’s official testing set is unavailable, so it subdivides the 167-image training set into 124 train / 43 test. This means (a) results are not directly comparable with future work that uses the official test set, and (b) 43 test images is very small, amplifying variance concerns (further supporting W1). The paper does not discuss this limitation.

    W6.Curve-based evaluation absent. The paper acknowledges excluding “curve-only methods” because their evaluation protocol “precludes fair benchmarking against mask-based predictors.” However, since a core contribution is the explicit Bezier curve modeling (ASCO), it would strengthen the paper to report curve-specific metrics (e.g., endpoint localization error, Frechet distance between predicted and ground-truth curves) that directly validate this contribution. Relying solely on mask-based metrics (DSC, IoU, ASSD) may not fully capture the benefit of curve modeling.

    W7.Data diversity and label quality not discussed. (General Review Considerations §3, §4) The three datasets are from the same clinical domain (laparoscopic liver surgery), but the paper does not discuss patient demographics, inter-annotator agreement for landmark labeling, or how annotation noise might affect the thin curvilinear structures being detected. Given that the paper does not report inter-rater agreement, annotation protocol details, or label uncertainty for thin curvilinear landmarks, this omission is notable.

    W8.Depth estimation dependency not ablated. The paper uses Depth Anything V2 for monocular depth estimation as input to the frozen SAM encoder. This is a significant design choice, yet no ablation examines A2ONet’s sensitivity to depth quality or compares against using no depth input. The absence of such analysis leaves the depth modality’s contribution and the method’s robustness to depth estimation errors unclear.

  • 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.

  • 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

    C1.The absence of multi-seed results (mean ± std) and statistical tests (e.g., Wilcoxon signed-rank or paired t-test on per-image DSC) limits confidence in the reported improvements over D2GPLand+, which I weight in my recommendation.

    C2.The validation–test DSC gap on L3D-2K (45.16% vs. 53.51%) is notable. An analysis of the composition of the val/test splits (e.g., per-landmark-type breakdown, illumination distribution) would help explain this discrepancy and increase confidence in the results.

    C3.The absence of runtime analysis (FPS, parameter count, FLOPs) limits assessment of the claimed intraoperative applicability, particularly given the multi-encoder architecture. Without these figures — ideally also for the D2GPLand+ baseline — readers cannot judge whether A2ONet meets a practical latency budget.

    C4.Consider reporting curve-specific evaluation metrics (endpoint error, Frechet distance) to directly validate the ASCO contribution. The current evaluation via mask-based metrics alone may understate the benefit of explicit curve modeling.

    C5.The values of λ_curve and λ_illum in Eq. 6 and the top-K value for the curve proposer are not specified, which hinders reproducibility. The sensitivity of performance to these weights is also unclear.

    C6.The paper omits a limitations discussion entirely. Notable unaddressed gaps include: (a) dependence on Depth Anything V2 quality for laparoscopic images, (b) absolute performance levels (DSC ~ 53% on L3D-2K suggests substantial room for improvement), and (c) computational cost considerations for real-time surgical guidance. This omission limits the reader’s ability to contextualise the contribution.

    C7.Without a depth-ablation (e.g., using only RGB features), the contribution of the depth modality and robustness to depth estimation errors remain unclear, which limits the strength of the depth-integration claim.

  • 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 well-designed architecture that addresses genuine challenges in laparoscopic liver landmark detection. The three proposed modules (IFC, FOSF, ASCO) are technically principled and their complementarity is convincingly demonstrated through systematic ablation. The consistent improvement across three benchmarks and all metrics supports the method’s effectiveness. However, the absence of uncertainty estimates and statistical tests (W1) is a significant weakness given the modest improvement margins (0.69–1.85% DSC over D2GPLand+). Without statistical validation, the claimed SOTA status is not firmly established. The unexplained validation–test gap (W2), missing computational analysis for the claimed intraoperative application (W3), and lack of limitations discussion (W4) are further notable omissions. I lean toward accept because: (a) the problem is clinically relevant and well-motivated, (b) the architectural design is sound with clear ablation support, (c) improvements are consistent across all three datasets and metrics (not cherry-picked), and (d) the ASSD improvements (3.65–7.79 pixels across the three datasets) are more substantial in relative terms than the DSC gains, suggesting real improvement in boundary quality. However, this rating is contingent on the authors addressing W1 (statistical tests) and W3 (runtime) in the rebuttal.

  • 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



Author Feedback

We thank all reviewers for recognizing the clinical relevance and motivation (R1, R3), clear presentation (R2), and consistent multi-dataset improvements (R1, R3) of our work. We address the remaining concerns below: Q1.Statistical significance (R1, R3): We have performed significance tests against D2GPLand+ on all three datasets, confirming our statistically significant improvements. We will report the mean±std results. Beyond DSC and IoU, A2ONet also substantially reduces ASSD, supporting improved boundary localization. Q2.Model complexity and efficiency (R1, R3): Although A2ONet integrates multiple modules, its trainable components are lightweight, relying mainly on convolutions and gating operations. The computational cost and inference speed are comparable to other SAM-based methods, reaching real-time inference. Q3.Curve-based evaluation (R1, R3): Our current comparison focuses on mask-based metrics to ensure consistency with existing segmentation benchmarks, but we agree that they may not fully reflect the benefit of explicit curve modeling. In future work, we will develop more comprehensive landmark evaluation metrics, including curve-specific metrics, to enable fairer comparison between mask- and curve-based methods. Q4.Design of IFC block and visualization (R1, R2): Unlike general low-light enhancement, our setting mainly suffers from spatially non-uniform laparoscopic illumination, with central saturation and peripheral attenuation. IFC is therefore designed as a lightweight task-oriented module to compensate this specific attenuation pattern without introducing heavy enhancement networks, preserving clinical efficiency. We sincerely thank R2 for this suggestion. The illumination-perturbation ablation would be helpful to assess prediction stability under different illumination settings. Q5.Depth-modality justification (R3): Our encoder design follows D2GPLand+, whose depth-modality ablation (Table 10) demonstrates that adding depth features clearly improves performance over RGB-only input, and using a stronger depth estimator further improves the results. This supports that depth provides useful geometric cues for landmark detection. Q6.Parameter clarification and convergence (R2, R3): We will clarify that the logit function is: logit(x)=log(x/1−x), where x is clamped to [10^−6, 1–10^−6] for numerical stability. We will also specify all hyperparameters, including λcurve=0.2, λillum=0.1, and the top-K=16.Ablations on key parameters and curve-update convergence will be included in future extended work. Q7.Dataset protocol and characteristics (R3): The validation-test gap on L3D-2K is mainly due to the distribution difference between the patient-level validation and test set. Similar gap are also observed in D2GPLand+ under the same dataset setting, as reflected in Table 2 (test set) and Table 4 (validation set). For P2ILF, we will release our code to facilitate future comparison. Despite the small test set, our improvements remain statistically significant. We will also discuss data diversity, annotation quality, and label uncertainty in journal version. Q8.Limitations and clinical validation (R1-R3): Due to the inherent difficulty of curvilinear structure segmentation in laparoscopic liver images, the absolute segmentation metrics leave room for improvement, especially in failure cases such as ambiguous junctions between silhouettes/ridges and ligaments, and false predictions under tissue or instrument occlusion. Our method is robust against reflections and there are few bleeding incidents in current datasets. We will further validate the detected landmarks in downstream pre-to-intraoperative 2D-3D liver registration to assess their clinical utility for AR guidance in the future.



Meta-Review

Meta-review #1

  • Your recommendation

    Provisional Accept

  • 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.

    All three reviewers rated this paper positively. The reviewers acknowledged the methodological soundness, improved results, and clear presentation. Authors are encouraged to incorporate the reviewers’ suggestions into their final manuscript to the fullest extent possible.



back to top