Enhancing Diabetic Retinopathy Detection Using Pixel Color Amplification and EfficientNetV2: A Novel Approach for Early Disease Identification

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

In this paper, the authors introduce an enhancing diabetic retinopathy detection method based on pixel color amplification and EfficientNet V2. This work is interesting but the paper is not well-written. The detailed comments are as follows:

1.  The technical contributions of the paper are not clearly summarized. At least for now, it seems that the proposed method is just a combination of the existing algorithms.
2. The logic of the paper is quite confusing, so it is suggested to adjust the structure of the paper.
3. The figures in the paper are of poor quality. For example, Figure 1, as a workflow diagram of the proposed method, does not highlight the innovation. For another example, there is no uniform style for the confusion matrices.
4. The experimental section should focus on analyzing the causes of the results.
5. The authors should discuss the drawbacks of the proposed method and plan for future improvements. The following papers may be helpful. 10.1109/ACCESS.2024.3377690, 10.1109/TIE.2023.3301546, 10.1109/TNNLS.2014.2330900.

Comments on the Quality of English Language

Extensive editing of the English language is required.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

Recommendation: Recommend publishing in MDPI Electronics after minor revision.

Dear editor and authors,

In this work, Lin and colleagues present a new approach for early diabetic retinopathy identification by employing pixel color amplification techniques to make retinopathy image more accurate and efficient. Overall, it is a comprehensive study and recommended to be published once below changes have been made.

1.                  Recommend removing the file name of each picture in Fig.3.

2.                  For those pictures without a full circular structure, do they impact training efficiency and accuracy?

3.                  Did you employ cross-validation, holdout validation, or another validation strategy?

 

4.                  Please discuss more on the architectural feature of the EfficientNetV2 that makes it more accurate and efficient than other CNN networks.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

This manuscript introduced an AI approach to early detection of diabetic retinopathy. A very detailed description of preprocessing steps was explained. The author also compared various CNN models, and proposed that EfficientNetV2 has better performance for accurate classification of DR.

A few minor issues are:  

1. In line 89, the term DR has been defined previously. 

2. In figure 3, the author should add the data after cropping for comparison. And figure 4 should be placed before figure 3, explain the theory first, then show the results. 

3. Are the images in figure 1b figure 5 the same? This is a bit confusing. If not, the author should also add figure 1b as an additional example in figure 5. 

4. Many places need proper references, for example: a) introduction section, line 34-42; b) In section 2.2.2, line 176 to 183, these contents need proper references. Please check the rest of the manuscript.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The authors have revised this paper properly according to my comments. The revised manuscript can meet the requirements of Electronics.