Suitability Classification of Retinal Fundus Images for Diabetic Retinopathy Using Deep Learning

Round 1

Reviewer 1 Report

The authors report an interesting study on the use of Deep Learning in classification of Retinal Fundus Images in diabetic patients. The work is well written and interesting, and it can be accepted for publication after minor English editing (e.g. see lines 50, 134, 245). 

Author Response

We appreciate your comments and thank you for taking the time to review our article. An edited English version is attached below. We hope that we have addressed them satisfactorily and that the manuscript is now suitable for publication.

Yours sincerely,

Susana Orteca-Cisneros

Author Response File: Author Response.pdf

Reviewer 2 Report

RFI quality control is an important step in machine learning approaches and several articles have been published in recent years. The authors proposed a new approach to purify the training data set for their CNN network to improve the efficiency of RFI classification. 

In the preprocessing, K-mean classtering was applied to categorize the images into low and high groups. Further CNN was trained to classify the RFI images to low and high groups as well. 

The results presented in the tables  5 and 10 are very close; it would be interesting to see the performances of clustering and CNN in terms of efficiency, senstitivity, specificity and accuracy, so that the advantages of the CNN approach would be clear. 

Author Response

We appreciate your comments and thank you for taking the time to review our article. A document is attached below where we explain the changes. We hope that we have addressed them satisfactorily and that the manuscript is now suitable for publication.

Yours sincerely,
On behalf of all the authors,
Susana Orteca-Cisneros

Author Response File: Author Response.pdf

Reviewer 3 Report

This paper proposed a deep neural network based retinal fundus images classification method. The proposed method obtained better performance according to the experiments. However, I still have some concerns.

1. For better understanding the architecture, I suggest the authors could give a figure of DRNet-Q framework.

2. I suggest the authors could update the references information, some arXiv version may have the final version.

Author Response

We appreciate your comments and thank you for taking the time to review our article. A document is attached below where we explain the changes. We hope that we have addressed them satisfactorily and that the manuscript is now suitable for publication.

Yours sincerely,
On behalf of all the authors,
Susana Orteca-Cisneros

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

I am satisified with the revision and have no further comments.

Reviewer 3 Report

The authors have successfully answered all my concerns.