Rapid and Accurate Diagnosis of COVID-19 Cases from Chest X-ray Images through an Optimized Features Extraction Approach

Round 1

Reviewer 1 Report

This paper proposed an interesting idea that used feature extraction and machine learning for rapid COVID-19 diagnosis. However, this paper did not reveal enough technical details and did not handle some details in a careful way.

1. For example, there a numerous grammar and format errors, which makes the paper hard to understand. Below are just some examples of language errors.

(a). Line 38: “It can able to transmit …..” should be “It can transmit ……”

(b). Line 66-67: “Thus, automatic classification is required in the X-ray image to detect the infection quick manner.” should be “Thus, automatic classification is required in the X-ray image to detect the infection in a quick manner.”

(c). Line 101-102: “In this model, 100 X-ray images are used and evaluate the performance based on threshold level (T). 90% sensitivity and 87.8% specificity (at T=9=0.25) or 96% sensitivity and 70.6% specificity (at T=0.15).” – these two sentences have grammar errors.

(d). Line 109-111: “A deep convolutional neural network model called COVID-Net [25], Wang et al., 2020. COVID-Net improves the transparency and trust to detect the positive cases from open-source X-ray images.” – these sentences have format and grammar errors.

(e). Line 113-120: “In each model, to analyse the intensity of normalization to classify the status of the COVID-19 cases. COVIDX-Net model required 50 X-ray images to validate the experiment. It has been done 80 for training the model and 20 for testing and achieves the F1-score of 0.91 for COVID-19 which is …” – There are grammar and spelling errors in these sentences. I’m not sure what the authors want to say. Do the authors want to say COVIDX-Net used 80% data for training and 20% data for testing?

(f). Line 124-125: “For optimization, a stacking algorithm is utilized and integration of gradient-based discriminative localization is to discriminate the COVID-19 infected regions from different pneumonia of X-ray images.” – There are grammar errors.

(g). Figure 1 has spelling mistakes. The box “Gamma correction techique” should be “Gamma correction technique”.

 

2. Besides language problems, this paper does not contain enough important technical details either.

For example, the authors did not describe the datasets used in this study. They only mentioned that they used 60 X-ray images of size 224x224 from Kaggle. However, the authors didn’t mention any background information about this dataset. For example, X-ray acquisition conditions affect image quality and therefore affect diagnosis, so under what conditions (e.g., dose, kVp, etc.) were these X-ray images collected? Usually, different X-ray machine vendors provide different X-ray image processing algorithms which affect the image quality. So, from what machine (e.g., vendor and model#) were the images collected? What pre-processing algorithms were applied to these images? Diagnostic X-ray images are larger than 224x224, therefore these images must have been downscaled, so what downscale algorithm was applied?

Patient information is also lacking from these datasets. For example, on what day after the onset of symptoms were the images collected? What is the age distribution of the patients? How severe were the patients?

I understand the dataset used in this study is an open dataset and such information may not be available. However, the authors should reveal as much information as possible and make clear statements about what information is not available.

 

3. The authors made a comparison of their results with others (e.g. DarkCovidNet model etc.), however, the datasets used in other studies are not the same. Therefore, such a comparison is not a fair comparison. The authors should rerun other’s algorithms on the same dataset instead of doing a direct comparison. Or the authors should explain why others’ results can be compared directly to their own results. Without a fair comparison, it is hard to draw the conclusion that the method proposed by this paper is better than others.

 

 

Author Response

Dear Reviewer,

Thanks for your valuable time and efforts to improve our submitted manuscript.

Please find the responses to your review comments as an attachment.

Thanks

Dr Arunachalam V

Associate Professor, SENSE,

Vellore Institute of Technology, Vellore

 

Author Response File: Author Response.docx

Reviewer 2 Report

I) Description: This paper proposes a feature extraction algorithm that combines Zernike Moment Feature (ZMF) and Gray Level Co-occurrence Matrix Feature (GF), which is used to detect the COVID-19 infection accurately.
II) strength points:
- the paper is well written and well organized, which makes it easy to follow.
-the contributions are clearly described
-
III)weaknesses:
- the proposed model design is not well justified, a proper justification of every choice should be given, for example, you select 8 features for high level, and 24 for low level, why is that, why not 30 and 100 for instance? it is not enough to describe your proposed model, you should also justify why you made these chooces.

VI) general comments:

- in the abstract "Deep Neural Network (DNN) is preferred over standard AI models to speed up the classification with minimal datasets.", this is erroneous, deep learning models particularly require a large amount of data to be trained.

-in the contributions list, remove 'the machine (processer core i7 at 2.80Ghz, 8GB RAM) used to run the MATLAB environment.) and move it to 'experimental settings description section', also remove the citations ([20-23, 25-27].) in the third contributions, move it to the section where you describe the baselines that you are comparing with.
- you should add the following related works:
[1] Alghamdi, Hanan S., Ghada Amoudi, Salma Elhag, Kawther Saeedi, and Jomanah Nasser. "Deep learning approaches for detecting COVID-19 from chest X-ray images: A survey." Ieee Access 9 (2021): 20235-20254.
[2] Zhang, Wenyin, Yong Wu, Bo Yang, Shunbo Hu, Liang Wu, and Sahraoui Dhelim. "Overview of multi-modal brain tumor mr image segmentation." In Healthcare, vol. 9, no. 8, p. 1051. MDPI, 2021.
[3] Serena Low, Woan Ching, et al. "An overview of deep learning techniques on chest X-ray and CT scan identification of COVID-19." Computational and Mathematical Methods in Medicine 2021 (2021).

 

 

 

Author Response

Dear Reviewer,

Thanks for your valuable time and efforts to improve our submitted manuscript.

Please find the responses to your review comments as an attachment.

Thanks

Dr Arunachalam V

Associate Professor, SENSE,

Vellore Institute of Technology, Vellore

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

Thank you for providing the requested additional information. The revised version looks good.