Semi-Supervised Medical Image Classification Based on Attention and Intrinsic Features of Samples
Round 1
Reviewer 1 Report
Regarding the content, I do not have any changes to recommend, it makes a good literary review to support the relevance of the problem to be studied and a good structuring of the content, it uses the correct methodology for this type of study and it is a consistent and well-detailed methodology to give significance to the results they show, makes a good discussion of the results with respect to the studies carried out previously, and marks the conclusion obtained well.
Although I advise looking at these things:
Never two sections without a paragraph of text in between. You should put a couple of lines describing/naming the subsections you are going to deal with within that section. You must correct this between sections 2 and 2.1.
A table or figure should never appear without reference to it in the text before it.
Citations must be numbered in order of appearance, citation 30 cannot come after 17 and before 18. Correct this throughout the article and be careful if these changes may affect the numbering of the “References” section.
In the section “5. Conclusion”, it is necessary to develop a deeper analysis of the conclusions, implications and limitations of the study. In addition to the possible future lines of research opened with this research.
Don't forget to add after the “Author Contributions” section, the “Funding” section. Look at the template.
And the references in the 'References' section must follow the model set by the journal. You must correct the errors that exist. Look at this in the template.
Author Response
Hello, dear reviewer! I really appreciate your taking the time to review my manuscript. After seeing your advice to me, I carefully reviewed my paper and made the following modifications.
First, I briefly summarized the main contents of our section 2 in sections 2 and 2.1, then I corrected the position of the table to ensure that it appeared after the corresponding text describing it, and I rearranged the serial numbers of the cited documents in the manuscript to ensure that they appeared in order. And in the manuscript, I joined the discussion module and discussed the conclusions and limitations of the research in the conclusion part. Finally, there is no fund support for this research at present.
Reviewer 2 Report
Manuscript titled Semi-supervised Medical Image Classification based on Attention and Intrinsic Features of Samples propose a new semi-supervised model for medical image classification.
Below I attached my concerns:
Abstract is lacking material and methods as well as informative conclusions.
Authors mentioned about the medical images applied as a input data. There is no information about this element. Clearly description of medical images should be described.
No discussion part was presented. Results should be compared to the existing literature and critical conclusions should appear.
No limitation to the study was presented. It should be clearly stated at the end of the discussion prat.
Conclusions should be more informative. Currently it is too general and do not refer to the presented results.
Author Response
Hello, dear reviewer! I really appreciate your taking the time to review my manuscript. After seeing your advice to me, I carefully reviewed my paper and made the following modifications.
First of all, I explained the results of our experiment in detail. I described the medical image in detail when I introduced the dataset in Section 4, and then I added the discussion part to the manuscript. In this part and conclusion, we compared our method with the previous methods and made a detailed comparison, and discussed the limitations of our method. Finally, thank you very much for your suggestion and look forward to your reply.
Reviewer 3 Report
1) Numeric achievement is missing of your work both in Abstract and conclusion.
2) The novelty of this paper is not very clear. Would you please add the novelty of the work either in abstract or in conclusion which differs from literature?
3) Please review more relevant works and find the research gap from there.
4) A general high level block diagram or framework of complete work in this paper is required to add at the beginning of chapter 3 so that readers can follow up the entire technical work has been done in this paper. Although you have provided overview of our semi-supervised framework however, you can modify the figure including training and testing steps.
5) Network architecture of semi supervised method is missing. Please add.
6) Feature extraction process is not clear.
7) Please add strength, limitation and impact/significance of this work in real life scenarios under the discussion subsection.
8) Specific future research directions are missing. Please add those at the end of conclusions.
Author Response
First of all, thank you very much for your suggestions on my manuscript. I carefully reviewed my manuscript and made the following modifications. I added the digitized results obtained by our experiment to the abstract and conclusion of the manuscript. Regarding the novelty of the paper, we introduced a super-parameter that changes with the performance of the classifier based on focal loss, which can well adjust the weight of the sample supervision loss of correct and wrong classification. Secondly, we introduce a sample intrinsic relationship loss similar to unsupervised loss to make better use of the intrinsic characteristics of unlabeled samples, which is the difference between our manuscripts. These have already been described in the abstract. We have made a little modification to the general layout of semi-supervised framework in Section 3, in which the colored cube module represents the feature extraction process of the network. Then, we added a discussion module to the manuscript, which compared our method with the previous method with the conclusion part, described the advantages and disadvantages of our method, and looked forward to the future research work.
Finally, I sincerely thank you again for your suggestion on my manuscript, and look forward to your reply.
Round 2
Reviewer 2 Report
Thank you for your answers.
Author Response
Dear reviewer, thank you again for taking the time to review my manuscript. I feel very honored.
Reviewer 3 Report
I would like to thank the authors to address most of my comments. However, still you may need to:
1. Review more relevant recent works
2. Please add significance of this work in real life scenarios under the discussion subsection.
3. English corrections is required through the entire manuscript.
Author Response
Dear reviewer, thank you very much for taking the time to review my manuscript and make valuable suggestions. In response to your questions, I have made the following modifications to my manuscript:
I consulted the recent related papers, but the number was small and I didn't have any good ideas, so I didn't change them. I added the significance of our work in the real scene to the discussion summary and carefully checked my manuscript and corrected it in English.
Thank you again for your help and look forward to your reply.
Author Response File: 
Author Response.docx
