Paper Info Reviews Meta-Review Author Feedback Post-rebuttal Meta-Reviews Back to top
List of Papers By topics Author List
Paper Info Reviews Meta-review Author Feedback Post-Rebuttal Meta-reviews

Authors

Haojie Han, Hongen Liao, Daoqiang Zhang, Wentao Kong, Fang Chen

Abstract

Dynamic contrast-enhanced ultrasound (CEUS) video with microbubble contrast agents reflects the microvessel distribution and dynamic microvessel perfusion, and may provide more discriminative information than conventional gray ultrasound (US). Thus, CEUS video has vital clinical value in differentiating between malignant and benign thyroid nodules. In particular, the CEUS video can show numerous neovascularisations around the nodule, which constantly infiltrate the surrounding tissues. Although the infiltrative of microvessel is ambiguous on CEUS video, it causes the tumor size and margin to be larger on CEUS video than on conventional gray US and may promote the diagnosis of thyroid nodules. In this paper, we propose a novel framework to diagnose thyroid nodules based on dynamic CEUS video by considering microvessel infiltration and via segmented confidence mapping assists diagnosis. Specifically, the Temporal Projection Attention (TPA) is proposed to complement and interact with the semantic information of microvessel perfusion from the time dimension of dynamic CEUS. In addition, we employ a group of confidence maps with a series of flexible Sigmoid Alpha Functions (SAF) to aware and describe the infiltrative area of microvessel for enhancing diagnosis. The experimental results on clinical CEUS video data indicate that our approach can attain an diagnostic accuracy of 88.79% for thyroid nodule and perform better than conventional methods. In addition, we also achieve an optimal dice of 85.54% compared to other classical segmentation methods. Therefore, consideration of dynamic microvessel perfusion and infiltrative expansion is helpful for CEUS-based diagnosis and segmentation of thyroid nodules. The datasets and codes will be available.

Link to paper

DOI: https://doi.org/10.1007/978-3-031-43987-2_17

SharedIt: https://rdcu.be/dnwJz

Link to the code repository

N/A

Link to the dataset(s)

N/A

Reviews

Review #4

  • Please describe the contribution of the paper

    This paper1 proposes a novel framework to diagnose thyroid nodules based on dynamic contrast-enhanced ultrasound (CEUS) video by considering microvessel infiltration. Specifically, the Temporal Projection Attention (TPA) is proposed to complement and interact with the semantic information of microvessel perfusion from the time dimension of dynamic CEUS. In addition, the paper employs a group of confidence maps with a series of flexible Sigmoid Alpha Functions (SAF) to aware and describe the infiltrative area of microvessel for enhancing diagnosis. The experimental results on clinical CEUS video data indicate that their approach can attain an accuracy of 88.79% for thyroid nodule and perform better than conventional methods1.

  • Please list the main strengths of the paper; you should write about 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 article is well-written and easy to follow
    2. the topic about using segmentation to improve the diagnosis is interesting
  • Please list the main weaknesses of the paper. Please provide details, for instance, if you think a method is not novel, explain why and provide a reference to prior work.

    authors should explain more about the works that use segmentation masks to improve the diagnosis[3][4], and compare with them also conduct the experiments on Thyroid Nodule Diagnosis. the author should also discuss more about the fluctuation on the diagnosis caused by the uncertainty in the segmentation[1,2]. Since the segmentation is sometime objective, and the pixe–level uncertainty would have significant effect on the diagnosis results.

    would the authors release the dataset? since the data is all collected from the same hospital, would it have the large domain similarity and cause it easy to overfit? plus, the number of samples is also not large enough to train a well-established deep learning model. authors should release the dataset or provide more detailed about the data analysis.

    [1] “Opinions Vary? Diagnosis First!.” MICCAI 2022 [2] “Calibrate the inter-observer segmentation uncertainty via diagnosis-first principle.” arXiv preprint arXiv:2208.03016 (2022). [3] “SeATrans: Learning Segmentation-Assisted Diagnosis Model via Transformer.”MICCAI 2022. [4] “Leveraging undiagnosed data for glaucoma classification with teacher-student learning.” MICCAI 2020

  • Please rate the clarity and organization of this paper

    Very Good

  • Please comment on the reproducibility of the paper. Note, that authors have filled out a reproducibility checklist upon submission. Please be aware that authors are not required to meet all criteria on the checklist - for instance, providing code and data is a plus, but not a requirement for acceptance

    yes

  • Please provide detailed and constructive comments for the authors. Please also refer to our Reviewer’s guide on what makes a good review: https://conferences.miccai.org/2023/en/REVIEWER-GUIDELINES.html

    address the concerns in point 6

  • Rate the paper on a scale of 1-8, 8 being the strongest (8-5: accept; 4-1: reject). Spreading the score helps create a distribution for decision-making

    5

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

    overall interesting research topic with novel method, but need the supplement of more discussions and verification.

  • Reviewer confidence

    Very confident

  • [Post rebuttal] After reading the author’s rebuttal, state your overall opinion of the paper if it has been changed

    N/A

  • [Post rebuttal] Please justify your decision

    N/A



Review #3

  • Please describe the contribution of the paper

    This paper introduces an explanatory framework for the diagnosis of thyroid nodules based on dynamic CEUS videos. Compared to the former method, the proposed network benefits from the Temporal Projection Attention as well as the use of confidence maps instead of binary masks in the perception of the infiltrative expansion area.

    Experiments show the strong competence of the proposed method on a dataset.

  • Please list the main strengths of the paper; you should write about 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 proposed method is technically sound and novel.
    2. The method outperforms all baselines on the dataset, in terms of all the metrics, which is impressive.
    3. The paper is well-organized and easy to understand.
  • Please list the main weaknesses of the paper. Please provide details, for instance, if you think a method is not novel, explain why and provide a reference to prior work.

    Generally, the paper is written in a clear way. However, there are some missing parts in the method and experiment section. Please check constructive suggestions for details.

  • Please rate the clarity and organization of this paper

    Very Good

  • Please comment on the reproducibility of the paper. Note, that authors have filled out a reproducibility checklist upon submission. Please be aware that authors are not required to meet all criteria on the checklist - for instance, providing code and data is a plus, but not a requirement for acceptance

    Satisfactory.

  • Please provide detailed and constructive comments for the authors. Please also refer to our Reviewer’s guide on what makes a good review: https://conferences.miccai.org/2023/en/REVIEWER-GUIDELINES.html

    Major points:

    1. In equation 10, are the hyperparameters selected manually? If yes, please provide the ablation study result.
    2. How did you pre-train the TALR backbone, please provide more details.

    Minor points:

    1. There is an extra blank after 600x in ‘Implementation Details’
    2. What is the batch size?
    3. It would be more clear to learn the model performance if the authors could report the time complexity of the proposed model
  • Rate the paper on a scale of 1-8, 8 being the strongest (8-5: accept; 4-1: reject). Spreading the score helps create a distribution for decision-making

    7

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

    This paper is technically sound, novel, and with a suitable topic for MICCAI. The main drawback of this paper is the missing description of details of some experiments instead of the method itself, which shows an impressive performance compared to the baselines. Consequently, I would suggest an acceptance.

  • Reviewer confidence

    Very confident

  • [Post rebuttal] After reading the author’s rebuttal, state your overall opinion of the paper if it has been changed

    N/A

  • [Post rebuttal] Please justify your decision

    N/A



Review #2

  • Please describe the contribution of the paper

    A dynamic CEUS video and microvessel infiltration-based framework has been proposed to diagnose thyroid nodules. The performance of the proposed technique has been validated with clinical datasets. .

  • Please list the main strengths of the paper; you should write about 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. Well-written manuscript
    2. Novel technical formulation
    3. Incorporation of several quantitative metrics
  • Please list the main weaknesses of the paper. Please provide details, for instance, if you think a method is not novel, explain why and provide a reference to prior work.
    1. No details on the parameter selection strategy
    2. Insufficient discussion
    3. Result improvements for a few cases are minimal
  • Please rate the clarity and organization of this paper

    Good

  • Please comment on the reproducibility of the paper. Note, that authors have filled out a reproducibility checklist upon submission. Please be aware that authors are not required to meet all criteria on the checklist - for instance, providing code and data is a plus, but not a requirement for acceptance

    The details on the parameter selection process will make the work reproducible.

  • Please provide detailed and constructive comments for the authors. Please also refer to our Reviewer’s guide on what makes a good review: https://conferences.miccai.org/2023/en/REVIEWER-GUIDELINES.html
    1. Abstract: The problem statement should be clearer. Please mention the specific margin of improvement achieved by the proposed technique over the existing one(s).
    2. Methods: Please elaborate on the parameter (such as \alpha, \lambda) selection process. Did you adopt any automatic criterion? If not, how did you ensure the optimality of the parameter values? Did you conduct any robustness analysis?
    3. Experiments: Did you obtain written consent from every patient?
    4. Implementation details: Why 100 epochs? What was the criterion to select the number of epochs?
    5. Table 1: The percentage improvement over V-Net + TMP + IPO is minimal. Did you conduct any formal statistical test to assess if the performance improvement is significant?
    6. Please list the limitations of the presented study.
    7. Conclusion: The conclusion section is too long and not precise. Please try to make it succinct.
    8. Spacing: A space is missing before every citation and figure number. Please fix this issue.
  • Rate the paper on a scale of 1-8, 8 being the strongest (8-5: accept; 4-1: reject). Spreading the score helps create a distribution for decision-making

    6

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

    The technical formulation is novel. The results are promising in general. However, there are scopes for improvement in a few cases.

  • Reviewer confidence

    Very confident

  • [Post rebuttal] After reading the author’s rebuttal, state your overall opinion of the paper if it has been changed

    N/A

  • [Post rebuttal] Please justify your decision

    N/A



Review #1

  • Please describe the contribution of the paper

    (1) This paper proposed a Temporal Projection Attention (TPA) to complement and interact with the semantic information of microvessel perfusion from the time dimension.

    (2) This paper adopted a group of confidence maps instead of binary masks to perceive the infiltrative expansion area from grayscale US to CEUS of microvessels for improving diagnosis.

  • Please list the main strengths of the paper; you should write about 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) This paper proposed a multi-task framework for CEUS videos, which is used for lesion area recognition and differential diagnosis.

    (2) In order to better identify the lesion area during the microvessel fusion period, the author proposed Temporal Projection Attention (TPA) to enhance the time information extraction.

    (3) Instead of using common segmentation strategies, the author uses Sigmoid Alpha Function (SAF) and Iterative Probabilistic Optimization (IPO) to generate lesion area maps.

  • Please list the main weaknesses of the paper. Please provide details, for instance, if you think a method is not novel, explain why and provide a reference to prior work.

    (1) The concept of modularity is not reflected in Figure 2. It is difficult to quickly find the proposed CFA/TLAR/TPA modules in the framework. This has caused confusion for readers.

    (2) The necessity of CFA is unknown. The pixel classification usually requires multi-scale features, but VNet itself has multi-scale characteristics. If only image classification requires multi-scale features, why design a cross-task shared feature extraction module. The 3D inception block may have many variations, but author did not describe it. Are those two learnable weights scalar or tensor? The results learned were not displayed either.

    (3) In TPA, the channel information re-weighting operation remains mysterious. The author is unwilling to spend two sentences introducing it.

    (4) SAF module is interesting, but not proposed by the author.

    (5) In IPO, the necessity of asymmetric convolution is unknown.

    (6) The core of this paper is to establish a multi-task framework to enhance the lesion area recognition and differential diagnosis by attention mechanism and iterative probabilistic optimization. But TPA module and SAF module were proposed by reference [12] and [13], respectively. The IPO module may be not author’s contribution either.

    (7) Too many hyper-parameter.

    (8) In the experiments, the called SOTA methods are outdated. UNet3D and V-Net were proposed in 2015 and 2016. Although Transunet was proposed in 2021, many studies have pointed out that it is difficult to achieve the performance claimed by the author. Please refer the paper [1].

  • Please rate the clarity and organization of this paper

    Satisfactory

  • Please comment on the reproducibility of the paper. Note, that authors have filled out a reproducibility checklist upon submission. Please be aware that authors are not required to meet all criteria on the checklist - for instance, providing code and data is a plus, but not a requirement for acceptance

    The reproducibility is low, CEUS data is not publicly available, and it is difficult to collect and annotate it. The code is not publicly available, and some method details are unclear.

  • Please provide detailed and constructive comments for the authors. Please also refer to our Reviewer’s guide on what makes a good review: https://conferences.miccai.org/2023/en/REVIEWER-GUIDELINES.html

    please see the detailed comments in the main weakness. (1) The writing and drawing of the method section need improvement, and the current version is not friendly for readers.

    (2) Please supplement the missing details in the method section.

    (3) Increase original contribution.

    (4) This paper may be more suitable for journals rather than conferences, as your method contains many details but is not introduced.

  • Rate the paper on a scale of 1-8, 8 being the strongest (8-5: accept; 4-1: reject). Spreading the score helps create a distribution for decision-making

    4

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

    limited method novelty and writing problems.

  • Reviewer confidence

    Confident but not absolutely certain

  • [Post rebuttal] After reading the author’s rebuttal, state your overall opinion of the paper if it has been changed

    N/A

  • [Post rebuttal] Please justify your decision

    N/A



Primary Meta-Review

  • Please provide your assessment of this work, taking into account all reviews. Summarize the key strengths and weaknesses of the paper and justify your recommendation. 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.

    The proposed work features multiple interesting techniques including Temporal Projection Attention (TPA), but more methodological details are needed for better understanding.



Author Feedback

N/A



back to top