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.
Reviewers are satisfied with the revisions, and recommend accepting this manuscript.
[# PeerJ Staff Note - this decision was reviewed and approved by Xiangjie Kong, a PeerJ Section Editor covering this Section #]
The author has responded to my concerns, and I have no further questions.
no comment
no comment
no comment
The reviewers have substantial concerns about this manuscript. The authors should provide point-to-point responses to address all the concerns and provide a revised manuscript with the revised parts being marked in different color.
**PeerJ Staff Note:** Please ensure that all review and editorial comments are addressed in a response letter and that any edits or clarifications mentioned in the letter are also inserted into the revised manuscript where appropriate.
**Language Note:** The review process has identified that the English language must be improved. PeerJ can provide language editing services - please contact us at [email protected] for pricing (be sure to provide your manuscript number and title). Alternatively, you should make your own arrangements to improve the language quality and provide details in your response letter. – PeerJ Staff
Lijuan Cui, et al investigated that towards reliable healthcare imaging: Conditional contrastive generative adversarial network for handling class imbalancing in MR images. However, there are some flaws that need to be addressed to improve the overall quality of this manuscript.
The following are the comments for the authors:
Major comments:
1. In fig1. The innovative 2C loss concept is mentioned, but a detailed explanation of this concept is not provided. Could the author provide more background on 2C loss so that readers can better understand this part?
2. In 3.1 CCGANs. The ultimate objective function is mentioned, and it is recommended to provide more explanation on why this objective function was chosen and how it can help with the semantic segmentation task.
3. In 3.5 Supervised Contrastive learning-based network (SCoLN). The use of contrastive learning has been mentioned, and it is suggested to provide a more detailed explanation of how contrast learning is used in this network, and what are the purposes and advantages of contrast learning.
4. The manuscript requires considerable grammatical corrections and there a several unclear phrases or words. (Line 477, “highlight” should be “highlights.”)
In this study, the authors developed a novel deep learning architecture, termed CCGAN, to address the issue of class imbalance within MRI datasets widely used for segmentation tasks. This model incorporates a class-specific attention mechanism, a region re-balancing module, and a discriminator based on supervised contrastive learning. The class-specific attention aims to selectively and effectively learn more relevant information from the input data. The authors integrated the RRM and class-specific attention modules into various deep learning models. Analysis of the results indicates a significant improvement in segmentation performance on highly imbalanced MRI datasets compared to baseline deep learning models.
Language can definitely be further polished. There are grammatical and style issues that need to be fixed.
Some mathematical terms need to be better defined.
The gap in the existing techniques need to be better clarified.
Methods need to be revised and communicated more rigorously and accurately.
Statistics can be better performed and described for the results in the table.
More details can be found in the attached file.
In the manuscript, the authors claimed that they developed a novel deep learning architecture namely CCGAN to tackle the problem of class Imbalancing in MRI dataset widely used for segmentation problem. The proposed model incorporates a class-specific attention mechanism, a region rebalancing module, and a discriminator based on Supervised Contrastive Learning. Though the results seem promising against other benchmark models. The structure of the CCGAN is still unclear: in the Figure 1,2,3, the authors did not give detailed structure info on 1). how contrastive learning worked, though mentioning in line 215: "...augmented with contrastive learning as shown in Figure 1."; 2). how attention mechanism worked, Figure 2 just shows regular CNN-like structures; 3). what region balance features are, in Figure 3, authors just added a few abstract boxes to the graph, which provided no actual meaning at all.
Additionally, the authors should re-structure their related work section, which is very lengthy right now. The authors should focus more on the contrastive learning work, attention mechanism etc. leaving fundamental knowledge out.
The authors claimed they devised a novel deep learning architecture namely CCGAN to tackle the problem of class Imbalancing in MRI dataset. The authors applied their architectures on various MRI datasets and did some ablation studies on certain modules to demonstrate their utilities: like so-called attention mechanism, class region balancing etc. However, the detailed structure of so-called contrastive learning, attention mechanism and region rebalancing it very ambiguous and none of them are clearly demonstrated in the figures. Particularly, the authors needs further explanation on the so-called attention mechanism, explaining on how attention is drawn and distributed across classes. Furthermore, the description of region rebalancing module is also very ambiguous, in the figure 3, only a few colorful boxes are added, without detailed explanation of region rebalancing modules. Such ambiguity in the method descriptions cast doubt on the validity of experimental design.
As mentioned earlier, the major concern of the manuscript is the ambiguity in the method descriptions, that cast doubt on the validity of experimental design and final findings. In all 3 figures, figures just provide very basic sketch of existing models, with no details on their specific designs on contrasive learning, attention and region rebalancing. Additionally, in the final tables of summarizing results, authors should mark the best performing groups with bold for all datasets
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.