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

Lian Liu, Han Zhou, Jiongquan Chen, Sijing Liu, Wenlong Shi, Dong Ni, Deng-Ping Fan, Xin Yang

Abstract

Deep neural networks have been widely applied in dichotomous medical image segmentation (DMIS) of many anatomical structures in several modalities, achieving promising performance. However, existing networks tend to struggle with task-specific, heavy and complex designs to improve accuracy. They made little instructions to which feature channels would be more beneficial for segmentation, and that may be why the performance and universality of these segmentation models are hindered. In this study, we propose an instructive feature enhancement approach, namely IFE, to adaptively select feature channels with rich texture cues and strong discriminability to enhance raw features based on local curvature or global information entropy criteria. Being plug-and-play and applicable for diverse DMIS tasks, IFE encourages the model to focus on texture-rich features which are especially important for the ambiguous and challenging boundary identification, simultaneously achieving simplicity, universality, and certain interpretability. To evaluate the proposed IFE, we constructed the first large-scale DMIS dataset Cosmos55k, which contains 55,023 images from 7 modalities and 26 anatomical structures. Extensive experiments show that IFE can improve the performance of classic segmentation networks across different anatomies and modalities with only slight modifications. Code is available at https://github.com/yezi-66/IFE.

Link to paper

DOI: https://doi.org/10.1007/978-3-031-43901-8_42

SharedIt: https://rdcu.be/dnwDQ

Link to the code repository

https://github.com/yezi-66/IFE

Link to the dataset(s)

N/A

Reviews

Review #2

  • Please describe the contribution of the paper

    1. A novel 2D dichotomous medical image segmentation (DMIS) dataset is proposed, termed as Cosmos55k, including 55023 images in terms of 7 modalities, such as CT, MRI, X-ray, etc.

    2. Instructive feature enhancement (IFE) is designed to select the deep learning-based features, which can improve the performances of 3 popular models: UNet, DeepLabV3++, and SINetV2.

  • 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. A novel insight is proposed to introduce the dichotomous image segmentation to medical image segmentation. Meanwhile, a new dataset is constructed based on the other public dataset.
    2. The proposed instructive feature enhancement (IFE) is interesting, which might be insightful to other community.
    3. The experimental results are comprehensive in terms of 9 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. The design motivation of IFE in UNet, DeepLabV3++, and SINetV2 might not be clear.
    2. Curvature-based Feature Selection seems like [1], [2]. It is suggested to discuss the differences between IFE and them.
    3. Cosmos55k is constructed based on 30 publicly available datasets, where the details of these public datasets should be cited.

    [1] Zuo, Zheming, et al. “Curvature-based feature selection with application in classifying electronic health records.” Technological Forecasting and Social Change 173 (2021): 121127. [2] Xing, Gang, et al. “Multi-scale pathological fluid segmentation in OCT with a novel curvature loss in convolutional neural network.” IEEE Transactions on Medical Imaging 41.6 (2022): 1547-1559.

  • 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
    1. Although the source code is not provided, the details of the proposed method are enough to reproduce.
    2. The proposed dataset, termed as Cosmos55k, is not public in this version. It is suggested to provide a clear and anonymous link to show some samples.
  • 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

    Minor issues:

    1. Fig.2 in Section 2.2 might not be introduced correctly.
    2. The legends of quantitative analysis plots in Supplement material might be not consist with the legends in the manuscript.
    3. The utilization of PNG format for CT, MR, and X-ray might not be suitable, since some metainfos should be contained, such as spacing.
  • 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?

    The presentation of this manuscript is clear and well-organized. The contribution of the dichotomous medical image segmentation (DMIS) dataset is solid. However, the design motivation instructive feature enhancement (IFE) for other deep learning models is not clear.

  • 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

    To select the instructive features for medical image segmentation, this study proposed an instructive feature enhancement approach (IFE) to adaptively select feature channels with rich texture cues and strong discriminability to enhance raw features. The mean curvature was used to select feature channels with rich texture cues, while entropy criteria was used to select feature channels with abundant discriminable information. To evaluate the proposed IFE, authors constructed a large-scale dataset Cosmos55k with various modalities and anatomical structures. Experiments show that IFE effectively can improve the performance of classic segmentation networks across different anatomies and modalities.

  • 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 study proposed an instructive feature enhancement approach (IFE) to adaptively select feature channels with rich texture cues and strong discriminability to enhance raw features via mean curvature and entropy criteria.
    2. This study constructed a large-scale dataset Cosmos55k with various modalities and anatomical structures.
    3. Experiments show that IFE can improve the performance of classic segmentation networks across different anatomies and modalities with only slight modifications.
  • 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. How to avoid the negative influence of complex background in computing mean curvature and entropy criteria? Whether the channels with abundant information of background but few information of foreground will be selected improperly?
    2. The reason of using curvature and entropy criteria should be further explained. And whether curvature and entropy criteria are qualify to describe the various texture and semantic information?
    3. Whether the potentially discriminable information in the discarded channels be missed?
    4. The English of the manuscript should be polished before further consideration for accept.
  • 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 code of this work was not provided. The reproducibility is slightly worse.

  • 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

    To select the instructive features for medical image segmentation, this study proposed an instructive feature enhancement approach (IFE) to adaptively select feature channels with rich texture cues and strong discriminability to enhance raw features. The mean curvature was used to select feature channels with rich texture cues, while entropy criteria was used to select feature channels with abundant discriminable information. To evaluate the proposed IFE, authors constructed a large-scale dataset Cosmos55k with various modalities and anatomical structures. Experiments show that IFE effectively can improve the performance of classic segmentation networks across different anatomies and modalities. This study provided a meaningful feature enhancement approach, but there are several weaknesses:

    1. How to avoid the negative influence of complex background in computing mean curvature and entropy criteria? Whether the channels with abundant information of background but few information of foreground will be selected improperly?
    2. The reason of using curvature and entropy criteria should be further explained. And whether curvature and entropy criteria are qualify to describe the various texture and semantic information?
    3. Whether the potentially discriminable information in the discarded channels be missed?
    4. The English of the manuscript should be polished before further consideration for accept.
  • 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?

    This study provided a meaningful feature enhancement approach IFE, using traditional feature descriptors. Experiments show that IFE effectively can improve the performance of classic segmentation networks across different anatomies and modalities. However, there are still several weaknesses. If the details can be added after major revision, I suggest receiving it after major revision.

  • 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

    This work grouped several datasets into a large dataset and adapted a new feature enhancement method into deep image segmentation model like unet, Deeplab V3+, for the binary medical image segmentation task.

  • 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.

    This paper introduced a new feature enhancement method for binary medical image segmenation. Two selected methods are evaluated. This method is eveluated on four different networks. Results show substantial improvement.

  • 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.

    The proposed curvature-based method utilized a designed filter. However, such a filter may be influenced by noise. Thus the performance may be various across different datasets. More detailed analysis for the experiment results is necessary. If the pages are not enough, some figures can be put in a supplementary material.

    The related work section is missing due to the limitation of the length. Some other related topics may contribute to this work. The explanation of Fig. 5 is not clear. Is it for comparison?

  • 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

    The model should be able to reproduce if the dataset is available.

  • 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

    Even though plenty of metrics have been utilized, the accuracy of the boundary should be further evaluated.

  • 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 method is simple and effective, the experiment results are reasonable.

  • 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



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.

    This paper proposes to select the features based on the local curvature or global information entropy which is evaluated on a collection of 55,023 images from 7 modalities and 26 anatomical structures. The proposed IFE module is pluggable to existing popular networks. The experiment results show its advantage. All 3 reviewers acknowledges the novelty of IFE and the evaluation on large dataset. It will be good if the author could address the reviewers’ question on weakness in the final version.



Author Feedback

N/A



back to top