Paper Info Reviews Meta-Review Author Feedback Post-rebuttal Meta-Reviews

Authors

Xukun Zhang, Zhiming Cui, Changan Chen, Jie Wei, Jingjiao Lou, Wenxin Hu, He Zhang, Tao Zhou, Feng Shi, Dinggang Shen

Abstract

Fetal brain segmentation from Magnetic Resonance (MR) images is a fundamental step in brain development study and early diagnosis. Although progress has been made, performance still needs to be improved especially for the images with motion artifacts (due to unpredictable fetal movement) and/or changes of magnetic field. In this paper, we propose a novel confidence-aware cascaded framework to accurately extract fetal brain from MR image. Different from the existing coarse-to-fine techniques, our two-stage strategy aims to segment brain region and simultaneously produce segmentation confidence for each slice in 3D MR image. Then, the image slices with high-confidence scores are leveraged to guide brain segmentation of low-confidence image slices, especially on the brain regions with blurred boundaries. Furthermore, a slice consistence loss is also proposed to enhance the relationship among the segmentations of adjacent slices. Experimental results on fetal brain MRI dataset show that our proposed model achieves superior performance, and outperforms several state-of-the-art methods.

Link to paper

DOI: https://doi.org/10.1007/978-3-030-87199-4_55

SharedIt: https://rdcu.be/cyl4P

Link to the code repository

N/A

Link to the dataset(s)

N/A

Reviews

Review #1

Please describe the contribution of the paper

In this paper, the authors propose a method to delineate the brain of fetuses in anisotropic stacks of in-utero MRI scans. The main contribution of the paper is the definition of a method to include the segmentation confidence in specific slices of the stack in order to increase segmentation accuracy in regions of the volumes affected by image distortions (prob due to motion, magnetic field effects).
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 paper is well written and the presentation is clear. The method is explained in sufficient detail and the evaluation is appropriate. The results compare favorably to a range of previous methods.
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.

While it is of course preferential to include the whole segmentation in a single framework, I cannot help but wonder how the results compare to simple post-processing such as through-slice smoothing (maybe median-filtering) of any of the other proposed methods. Also, the novelty of the method is somewhat questionable, as measures of inter-slice consistency have been used in segmentation tasks before (see for instance 10.1155/2018/4185279)
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

Reproducibility of the paper is rather low. Only proprietary data was used and no code is provided.
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

Apart from the general comments raised before, the authors should re-format their formulas: use only single letters as variables (eg “GT_s” in eq. (1)), “Entropy(t) in eq. (2). Also, the first sentence in 2.2 is not a complete sentence, and there is a typo in 3.1 (“sacns” instead of “scans”)
Please state your overall opinion of the paper

borderline accept (6)
Please justify your recommendation. What were the major factors that led you to your overall score for this paper?

While the method’s novelty is at least conceptually questionable, the comparison with other methods on the same task is very exhaustive and favorable to the presented method. I would therefore recommend acceptance of the paper, albeit weakly.
What is the ranking of this paper in your review stack?

2
Number of papers in your stack

4
Reviewer confidence

Very confident

Review #2

Please describe the contribution of the paper

This paper presents a confidence-aware multi-stage 2D segmentation framework to perform fetal brain segmentation in 3D MR images. In particular, a localization model was used to coarsely localize the brain regions and to calculate the confidence of the segmentation.
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 proposes a segmentation consistency loss to exploit the consistent relationship among adjacent image slices.
2. Experimental results show improved performance compared to several existing segmentation frameworks.
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 authors need to carefully revise the decription of Guidance Module, which is inconsistent with the Figure 2, i.e. the softmax operation is absent in the description.
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 authors need to carefully review the Guidance Module, which is inconsistent with the Figure 2, i.e. the softmax operation is absent in the description.
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

This paper presents a confidence-aware multi-stage 2D segmentation framework to perform fetal brain segmentation in 3D MR images. In particular, a localization model was used to coarsely localize the brain regions and to calculate the confidence of the segmentation. Fetal MRI scans usually suffer from motion artifacts, thus, the proposed framework with confidence loss makes sense. Experimental results show improved performance compared to several existing segmentation frameworks. I would like to suggest the authors to carefully revise the Guidance Module, which is inconsistent with the Figure 2, i.e. the softmax operation is absent in the description.
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?

The main contribution of this work is to laverage the information embeded within the neibourhooding slices to improve the poor segmention performace with motion artifacts, which makes a lot sense. Overall, it is an interesting paper with careful experimental studies.
What is the ranking of this paper in your review stack?

1
Number of papers in your stack

5
Reviewer confidence

Very confident

Review #3

Please describe the contribution of the paper

This paper introduces a cascaded network to segment brain from MRI. The pipeline included two networks, first stage coarsely segment brain and generate segmentation certainty per slice. The localized brain in the second network segmented finely and slices with low confidence value guided by neighboring slices
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 proposed method Using cascaded networks for detecting slices with low confidence and using neighboring slices for generating segmentation is novel. The ablation evaluation and comparing with state-of-art methods are nicely done
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.

Dataset: The dataset is not describes well especially what proportion of slices had artifacts, are they included in the training set? How GT has been done on slices with sever artifacts? Did doctors segment based on neighboring slices? and the evaluation could have been reported separately on slices with artifacts Fetal MRI suffer from inter slice movements, consistency loss optimizing prediction with neighboring slice using coordinates. This assumption is only true if images are well registered. It is mentioned bias field correction method was applied on all images in page6, figures 3, 4 illustrated bias field artifacts. Was that method did not work or figures are made prior to processing? Method: the confidence measurements are not pixel level but slice. Considering many slices only part of slice has artifact not the entire slice
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

I can not comment on that
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

typo in page 6 Shiah et al –> Salehi et al [9]

The following paper presented fetal MRI segmentation method in images with bias artifacts, please comment how you compare with using biased artifacts as an augmentation https://www.sciencedirect.com/science/article/abs/pii/S0730725X18306106
Please state your overall opinion of the paper

borderline accept (6)
Please justify your recommendation. What were the major factors that led you to your overall score for this paper?

The method is novel and the application is well chosen, there is room for further improvement to describe dataset and method more precisely.
What is the ranking of this paper in your review stack?

1
Number of papers in your stack

4
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.

The authors proposed a novel cascaded network for fetal brain segmentation by leveraging uncertainties and inter-slice consistency. The three reviewers have consensus that the proposed method is interesting and the results are promising compared with existing methods. They all recommended an acceptance for this work. However, some minor concerns from the reviewers should be addressed by the authors, such as the formulations of the Guidance module, and clarifications on the dataset and reasonability of using inter-slice consistency.
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).

2

Author Feedback

N/A

back to top

Confidence-aware Cascaded Network for Fetal Brain Segmentation on MR Images