Back to top List of papers List of papers - by topics Author List
| Paper Info | Reviews | Meta-review | Author Feedback | Post-Rebuttal Meta-reviews |
Authors
Tan-Cong Nguyen, Tien-Phat Nguyen, Gia-Han Diep, Anh-Huy Tran-Dinh, Tam V. Nguyen, Minh-Triet Tran
Abstract
Polyps detection plays an important role in colonoscopy, cancer diagnosis and early treatment. Many efforts have been made to improve the encoder-decoder framework using global feature with an attention mechanism to enhance local feature, helping to effectively segment diversity polyps. However, using only global information derived from the last encoder block leads to the loss of regional information from intermediate layers. Furthermore, defining the boundaries of some polyps is challenging because there is visual interference between the benign region and the polyps at the border. To address these problems, we propose two novel modules: the Cascading Context module (CCM) and the Attention Balance module (BAM), aiming to build an effective polyp segmentation model. Specifically, CCM combines the extracted regional information of the current layer and the lower layer, then pours it into the upper layer - fusing regional and global information analogous to a waterfall pattern. The BAM uses the prediction output of the adjacent lower layer as a guide map to implement the attention mechanism for the three regions separately: the background, polyp, and boundary curve. BAM enhances local context information when deriving features from the encoder block. Our proposed approach is evaluated on three benchmark datasets with six evaluation metrics for segmentation quality and gives competitive results compared to other advanced methods, for both accuracy and efficiency. We shall release our code upon acceptance.
Link to paper
DOI: https://doi.org/10.1007/978-3-030-87193-2_60
SharedIt: https://rdcu.be/cyhMD
Link to the code repository
https://github.com/ntcongvn/CCBANet
Link to the dataset(s)
https://datasets.simula.no/kvasir-seg/
http://www.cvc.uab.es/CVC-Colon/index.php/databases/
http://www.cvc.uab.es/CVC-Colon/index.php/databases/cvc-endoscenestill/
Reviews
Review #1
- Please describe the contribution of the paper
This paper presents two novel modules: the cascading context module (CCM) and the attention balance module (BAM). The performance of the proposed method is tested by visual inspection and quantitative evaluation.
- 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.
An effective polyp segmentation model is presented in this paper. The CCM module is designed to capture the global context as well as the regional context and the BAM module is presented to balance the attention mechanism for background, foreground and boundary.
- 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.
Several abbreviations, e.g. CMMs and BCM seems to be used without definitions.
- 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
This paper has a good reproducibility.
- Please provide detailed and constructive comments for the authors. Please also refer to our Reviewer’s guide on what makes a good review: https://miccai2021.org/en/REVIEWER-GUIDELINES.html
What is the purpose of enhancing the background of polyps? We are looking forward to pursue the performance of the proposed method that the background attention of polyps is removed. Several abbreviations, e.g. CMMs and BCM seems to be used without definitions.
- Please state your overall opinion of the paper
accept (8)
- Please justify your recommendation. What were the major factors that led you to your overall score for this paper?
This paper presents two novel modules: the cascading context module (CCM) and the attention balance module (BAM). The experimental results prove that both modules can improve the performance of Polyp segmentation.
- What is the ranking of this paper in your review stack?
1
- Number of papers in your stack
6
- Reviewer confidence
Very confident
Review #2
- Please describe the contribution of the paper
This paper proposes to use the Cascading Context Module to combine the global context and the regional context information and develops the Balancing Attention Module which implements separately and balances the attention mechanism for all three regionals: background, foreground and boundary curve. Experimental results show the effectiveness of the proposed method.
- 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.
- The idea of combine the global context and the regional context information is well-motivated.
- The writing is easy to follow.
- The proposed method achieves the state-of-the-art performance.
- 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.
-
There is no ablation study to show the effectiveness of background, foreground and boundary curve in BAM.
-
- 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
This method is easy to be implemented.
- Please provide detailed and constructive comments for the authors. Please also refer to our Reviewer’s guide on what makes a good review: https://miccai2021.org/en/REVIEWER-GUIDELINES.html
N.A.
- Please state your overall opinion of the paper
Probably accept (7)
- Please justify your recommendation. What were the major factors that led you to your overall score for this paper?
Overall, the idea of the proposed method is interesting and well-motived. The proposed method achieves a new state-of-the-art performance. I am positive to this paper.
- What is the ranking of this paper in your review stack?
1
- Number of papers in your stack
3
- Reviewer confidence
Confident but not absolutely certain
Review #3
- Please describe the contribution of the paper
The Cascading Context module (CCM) and the Attention Balance module (BAM) are proposed aiming to build an effective polyp segmentation model. Specifically, CCM combines the extracted regional information of the current layer and the lower layer, then pours it into the upper layer - fusing regional and global information analogous to waterfall patterns. The BAM uses the prediction output of the adjacent lower layer as a guide map to implement the attention mechanism for the three regions separately: the background, polyp, and boundary curve. BAM enhances local context information when deriving features from the encoder block.
- 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.
The Cascading Context Module can effectively combine the global context and the regional context information, which provides the ability to enable multi-scale receptive fields to help enhance the consistency of pixel-wise segmentation. In addition, the Balancing Attention Module performs as an attention machine leading to a better feature representation.
- 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 paper designs two interactional modules among different layers in UNet-style frameworks for segmentation. Although the method achieves a good performance. Its novelty is not enough since it just simply fuses the global context and local information. Although it constructs a better multiscale feature representation. There were many works that proposed a similar idea. The major contribution of this paper needs to be further enhanced.
- 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 reproducibility of the paper is ok.
- Please provide detailed and constructive comments for the authors. Please also refer to our Reviewer’s guide on what makes a good review: https://miccai2021.org/en/REVIEWER-GUIDELINES.html
The contribution needs to be further improved. Its novelty is not enough since it just simply fuses the global context and local information for constructing a better multiscale feature representation. This idea has been applied in many fields.
- Please state your overall opinion of the paper
borderline reject (5)
- Please justify your recommendation. What were the major factors that led you to your overall score for this paper?
I argue that the contributions of this paper have been introduced in other papers. It is not the first work to propose the idea. It is just an application for medical images.
- What is the ranking of this paper in your review stack?
4
- Number of papers in your stack
5
- Reviewer confidence
Very confident
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 proposed the Cascading Context module (CCM) and the Attention Balance module (BAM) for an effective polyp segmentation model. Specifically, CCM combines the extracted regional information of the current layer and the lower layer and BAM enhances local context information when deriving features from the encoder block.
However, the major concern is the novelty of this paper, it seems just to combine existing ideas for polyp segmentation. The motivation and the differences compared with existing ideas should be added.
- What is the ranking of this paper in your stack? Use a number between 1 (best paper in your stack) and n (worst paper in your stack of n papers).
5
Author Feedback
We thank all Reviewers and Area Chair for the positive and constructive comments. Below are our answers (A) to the comments (Q).
Reviewer #1 Q: Several abbreviations seem to be used without definitions. A: Thank you very much. All abbreviations will be fully reviewed and defined in the final version.
Q: What is the purpose of enhancing the background of polyps? A: Thank you. Enhancing the background of polyps drives the segmentation network to sequentially discover new and complement object regions by erasing the previously detected salient regions in an adversarial manner. As in a regular segmentation framework, the predicted map is a rough segmentation result. Thus there are some parts of the polyp that have not been fully segmented. By erasing the segmented polyp region with high confidence and then joining the attention mechanism in its upper layer, the discriminative regions have been removed and no longer contribute to the segmentation prediction; the network is naturally driven to discover new regions. Another reason for background attention is that we want to solve the segmentation problem based on multiple tasks. The task of well-distinguishing the background through the attention mechanism is also helpful in some way to the polyp segmentation task.
Q: This paper presents two novel modules: the cascading context module (CCM) and the attention balance module (BAM). The experimental results prove that both modules can improve the performance of Polyp segmentation. A: Thank you for the positive comments.
Reviewer #2 Q: The idea of combining the global context and the regional context information is well-motivated. A: Thank you very much.
Q: There is no ablation study to show the effectiveness of background, foreground, and boundary curves in BAM. A: We will include the suggested study in the final version. Thank you very much.
Reviewer #3 Q: Regarding the novelty A: Thank you. As mentioned and also recognized by other reviewers, our two proposed modules, CCM and BAM, are well-motivated and elegantly complementary. The experimental results clearly demonstrate the effectiveness of the proposed work. We would like to emphasize the novelty of the two proposed modules as follows.
Cascading Context Module (CCM) refers to two new ideas. First, CCM proposes the concept of regional context information at the intermediate layers, which has not been mentioned in previous studies. Global context information can be useful for segmenting large polyps, but regional context information is more useful for small polyps. Here, the global context contains the features of the entire image; however, the small polyps occupy only a small region in the image. Second, CCM proposed a new way to effectively combine the global context and the regional context information. The current approaches use only global context information derived from the last encoder block may lead to ignoring some useful information of the intermediate encoder layers. In contrast, the CCM module can capture the global context and the regional context derived from E-Blocks of all deeper layers. In this way, the context information of regions in the image is weighted more precisely.
Attention Balance Module (BAM) performs as a general attention mechanism. The current approaches focus only on a regional semantic concept for the attention mechanism by using only either background or boundary. In contrast, BAM uses the background and boundary and extends the attention mechanism on the third region of interest that is the foreground. These three regions cover all the semantic space of the entire image. Therefore directing attention to individual regions will help strengthen the discriminative regions leading to a more accurate segmentation. BAM proposed a new approach that has not been mentioned in previous studies to implement separately and balances the attention mechanism for all three regions to learn a better feature representation.
Post-rebuttal Meta-Reviews
Meta-review # 1 (Primary)
- Please provide your assessment of the paper taking all information into account, including rebuttal. Highlight the key strengths and weaknesses of the paper, clarify how you reconciled contrasting review comments and scores, indicate if concerns were successfully addressed in the rebuttal, and provide a clear justification of your decision. If you disagree with some of the (meta)reviewer statements, you can indicate so in your meta-review. Please make sure that the authors, program chairs, and the public can understand the reason for your decision.
This paper proposed the Cascading Context module (CCM) and the Attention Balance module (BAM) for an effective polyp segmentation model. Specifically, CCM combines the extracted regional information of the current layer and the lower layer and BAM enhances local context information when deriving features from the encoder block. The major concern is the novelty of this paper, it seems just to combine existing ideas for polyp segmentation. The motivation and the differences compared with existing ideas should be added.
In the rebuttal part, the author gives detailed motivation for the proposed modules, which have some kinds of novelty. The authors are suggested to add these discussions related to the novelty issues in the final version.
- After you have reviewed the rebuttal, please provide your final rating based on all reviews and the authors’ rebuttal.
Accept
- What is the rank of this paper among all your rebuttal papers? Use a number between 1 (best paper in your stack) and n (worst paper in your stack of n papers).
4
Meta-review #2
- Please provide your assessment of the paper taking all information into account, including rebuttal. Highlight the key strengths and weaknesses of the paper, clarify how you reconciled contrasting review comments and scores, indicate if concerns were successfully addressed in the rebuttal, and provide a clear justification of your decision. If you disagree with some of the (meta)reviewer statements, you can indicate so in your meta-review. Please make sure that the authors, program chairs, and the public can understand the reason for your decision.
The major concern of reviewers is the limited novelty. The rebuttal responded this question. Thus, I prefer to accept.
- After you have reviewed the rebuttal, please provide your final rating based on all reviews and the authors’ rebuttal.
Accept
- What is the rank of this paper among all your rebuttal papers? Use a number between 1 (best paper in your stack) and n (worst paper in your stack of n papers).
10
Meta-review #3
- Please provide your assessment of the paper taking all information into account, including rebuttal. Highlight the key strengths and weaknesses of the paper, clarify how you reconciled contrasting review comments and scores, indicate if concerns were successfully addressed in the rebuttal, and provide a clear justification of your decision. If you disagree with some of the (meta)reviewer statements, you can indicate so in your meta-review. Please make sure that the authors, program chairs, and the public can understand the reason for your decision.
Though the paper is an integration of existing modules, it has shown to be effective for the polyp segmentation n this paper through extensive experiments and comparison on 3 different benchmark available datasets. The authors agreed to include an ablation study in the camera ready to further support their findings. I recommend an accept for this paper. The authors should update the camera ready to incorporate the justifications provided in the rebuttal.
- After you have reviewed the rebuttal, please provide your final rating based on all reviews and the authors’ rebuttal.
Accept
- What is the rank of this paper among all your rebuttal papers? Use a number between 1 (best paper in your stack) and n (worst paper in your stack of n papers).
2