All reviews of published articles are made public. This includes manuscript files, peer review comments, author rebuttals and revised materials. Note: This was optional for articles submitted before 13 February 2023.
Peer reviewers are encouraged (but not required) to provide their names to the authors when submitting their peer review. If they agree to provide their name, then their personal profile page will reflect a public acknowledgment that they performed a review (even if the article is rejected). If the article is accepted, then reviewers who provided their name will be associated with the article itself.
The manuscript can be accepted now.
[# PeerJ Staff Note - this decision was reviewed and approved by Paula Soares, a PeerJ Section Editor covering this Section #]
The authors have replied and addressed all the comments I raised. Thank the authors for their work. I recommend publishing this article.
Has been improved.
Has been improved.
The authors have addressed all my comments carefully by supplementing experiments and explanation of their model. Also, the presentation is improved significantly in the revised manuscript. In my opinion, the paper is suitable for publication in PeerJ.
no comment
no comment
Author has made the modifications as per the suggested revisions, but couldn't find the keywords in the text. Please add the keywords.
Author has made the modifications as per the suggested revisions, but couldn't find the keywords in the text. Please add the keywords.
Author has made the modifications as per the suggested revisions, but couldn't find the keywords in the text. Please add the keywords.
Please carefully consider the comments raised by reviewers.
[# PeerJ Staff Note: Please ensure that all review and editorial comments are addressed in a response letter and any edits or clarifications mentioned in the letter are also inserted into the revised manuscript where appropriate. #]
The structure of this article is well organised. Esophageal cancer (EC) is the seventh most common cancer. Radiotherapy is one of the most important treatment strategies for EC. Thorax CT images are usually used for EC radiotherapy planning. The critical step in the planning pipeline involves the extraction of the esophageal gross tumor volume (GTV) from CT data by manual segmentation. However, manual or even semi-automatic algorithms are usually operator-dependent, time-consuming and subject to high inter and intra-observer variability in clinical practice. Besides, the spectral images obtained using SDCT can provide additional information to tranditional CT. Therefore, the authors evaluated the performance of deep learning model for fully automated segmentation of EC on SDCT scans acquired for Radiotherapy treatment planning, and compare performance between CI and VMI40kev data. They found that the VMI 40kev as input data resulted in higher learning efficacy in the training datasets and slightly higher accuracy in the test datasets in comparison with CI.
1. What are the criteria for dividing the training, validation and test sets?
2. "VMI 40kev as input data resulted in higher learning efficacy in the training datasets", how to define the higher learning efficacy.
3. "Fully automated segmentation of EC with 3D Res-UNet has high reliability", how reliability is demonstrated
4. "However, very few automatic segmentation models for EC have been proposed based on CT images", a brief summary of the results of these few relevant studies is suggested.
5. Can the inclusion and exclusion criteria for the sample be represented in a flow chart?
6. "All density scopes were -50 HU (min) to 400 HU (max); the density was normalized and artificially set from 0 to 1.", can you tell me exactly what normalization method was used and if you can give me the formula?
7. Can the image pre-processing code be provided, or at least the software, toolkit and version in use?
8. It is recommended that loss convergence plots are provided for the training and validation sets.
9. "...in the train and validation datasets and test datasets were compared by the t-test.", does the t-test here refer to a two-sample t-test? Is it a two-tailed.
10. Is there a test for consistency between the two observers when EC segmentation?
11. What does each number in Table 1 mean when statistically describing the size of the tumour?
The image pre-processing code should be provided, or at least the software, toolkit and version in use?
In this study, the authors use 3D Res-UNet to assess the performance of fully automated segmentation of esophageal cancer. The experimental results show the 3D Res-UNet works well on CI and VMI_40kev data. It is also proved that the VMI_40kev is helpful for modelling the segmentation performance. In addition, the authors demonstrated that the 3D Res-UNet is clinically feasible and has practical value in planning patient detection, diagnosis and radiotherapy.
The paper is well organized into sections and the presentation is clear to me. I think the paper is suitable for publication in PeerJ if the authors can address the comments properly.
(1) Please describe the 3D Res-UNet model in detail, including the convolution strides, paddings, and batch sizes, etc.
(2) Although this paper compares the results from other papers, it does not use its own data to compare the results of the classical or state-of-the-art models with 3D Res-UNet, which fails to comprehensively assess the superiority of the 3D Res-UNet in the segmentation of esophageal cancer.
(3) Why is the epoch set to 600? In my opinion, such an epoch can lead to a big computational cost. Why not change the learning rate during training phase? Can the learning rate of 0.0001 reach the optimal value? I suggest the authors consider the dynamic learning rate.
(4) How is the deconvolution layer in the decoder defined. Will the directly deconvolving the connected tensor cause the deficiency of data fusion, thus resulting in missing feature vectors？
(1) Why do the authors give the median of each evaluation index instead of the average? As shown in Figure 3, the split performance for the test set is not very stable. The authors should explain the reason for this?
(2) In the paper, the authors mentioned that low-kev VMI can significantly improve the attenuation and signal-to-noise ratio for the primary tumor. Also, the CNR of the primary tumor vs circumjacent anatomic is significantly higher in the low-kev VMI, peaking at VMI_40kev. It can be seen that VMI_40kev should have a better segmentation than CI. So what is the significance of this study?
no comment
Clear and unambiguous, professional English used throughout.
Research question well defined, relevant & meaningful.
Conclusions are well stated, linked to original research question & limited to supporting results.
This study mainly utilized the 3D Res-UNet technique to compare the performance of conventional images and 40-keV virtual monoenergetic images (VMI) in automatic segmentation of esophageal cancer (EC). It was found that using 40-keV VMI can improve the accuracy of EC segmentation. Meanwhile, the 3D Res-UNet method showed better performance in EC detection compared to traditional imaging methods. These findings may contribute to the improvement of diagnostic and therapeutic approaches for EC. However, there are still some minor issues in this study that need further refinement.
Introduction:
1.Please list the methods used in previous studies to segment esophageal lesions and their respective advantages and disadvantages. After analyzing these previous studies, explain why this method was chosen for the present study, its advantages, and the problems it can solve.
2.Please add the relevant introduction for VMI40kev.
3.Please confirm again whether 'informed consent was waived' is correct.
4."Automatic Segmentation Model and training" section could provide more detailed data processing steps.
5.Please thoroughly check for minor grammatical errors throughout the entire manuscript.
All text and materials provided via this peer-review history page are made available under a Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.