Review Reports

Round 1

Reviewer 1 Report

The authors present a paper about an automatic method for classifying chest x-ray as COVID-19/NON COVID-19. The classification is based on deep learning algorithm. A large dataset was also proposed. The topic is very interesting and the proposed idea is good. The draft is really well written. Anyway, in my opinion, some points should be clarified/commented on.

Major comments:

• Results are reported also in the introduction and materials and methods. I would suggest removing it from these sections.

• Flipping has been run for data augmentation. In my opinion, an upside-down patient is not credible.

• page 5, line 180. Missing/corrupted reference to a figure.

• Intensity normalization was used?

• The training/testing was executed on CPU or GPU? What about the timing?

• How the testing patients were selected. What about a k-fold strategy?

• Which stopping criterion was used?

• Please run some statistical tests to assess if the classifications are different.

• In order to augment the dataset, a possible choice could be generating digital radiography from CTs. Several covid19 positive CTs can be found online for free (as
  Zaffino P, Marzullo A, Moccia S, Calimeri F, De Momi E, Bertucci B, Arcuri PP, Spadea MF.

  An Open-Source COVID-19 CT Dataset with Automatic Lung Tissue Classification for Radiomics.
  Bioengineering. 2021; 8(2):26. https://doi.org/10.3390/bioengineering8020026 )

  Please discuss this option.

Minor comments:

• Typo in the abstract (n=umber)

• Please rephrase the sentence on page 11, lines314-315.

In summary: the topic of the manuscript is very interesting, the proposed strategy and the results are good, but some minor revisions are required.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

The authors report the application of pre-trained neural networks to predict COVID19 from chest x-rays as well as improved more balanced meta-dataset "COVID-ChestXray-15k" which combines various previously published but more limited studies.

Overall, the improvement in performance seems promising. Unfortunately, neither code nor data are provided, which makes further evaluation impossible.

minor comments:

line 2-3: change to "COVID-19 infected patients show abnormalities in chest X-ray images"

line 36: imprecise and cryptic, please improve this statement "Deep Learning is a 37 complex architecture of machine learning methods."

line 52: change to COVID-19, which is the WHO name of the disease, the D standing for disease "predict the Covid-19 disease", make sure to follow this throughout the manuscript.

line 59: missing period after "which leads to a total of 15,000 images"

line 60: language "To summarize, the contributions of this paper can 60 be summed as follow:" IMPORTANT: Please let a native speaker read this manuscript before resubmitting. The language is not on par with publishing standards. Ideally, do this before submitting a paper...

This statement is confusing numbering doesn't line up at all with numbering of the sections, please just remove: "The rest of the paper’s structure is as follows; the related work is highlighted in 69 Section 1. Section 2 outlines the methodology of the proposed method. In Section 3, 70 we show and compare the results of this work from different aspects. In section 4 we 71 present the discussion. Finally, we conclude the research in section 5 and propose future 72 work to enhance the proposed work." 

Figure 1: please remove, there is an abundance of information on COVID-19, you can incorporate this by reference. Besides, the formatting does not follow standards of scientific publishing. (i.e. font?!?)

In conclusion, I can't review the rest of the paper without having access to the author's code and data. Unless this is provided in a presentable form which makes it possible to follow what the authors did within a reasonable timeframe (i.e. jupyter notebook or at least a link to a github repository), I can not recommend it for publication.

Before resubmission please ask a native speaker with ML domain expertise to review the paper, there are countless language and grammar problems. 

Author Response

Please see the attachment.

Author Response File: Author Response.docx