Efficient and Accurate Hemorrhages Detection in Retinal Fundus Images Using Smart Window Features

Round 1

Reviewer 1 Report

The authors propose an algorithm to segment and detects hemorrhages in retinal fundus images. The proposed method first performs preprocessing on retinal fundus images. Then a novel smart windowing-based adaptive threshold is utilized to segment hemorrhages and they found that the proposed method achieves a high F1 Score with 83.85% for the DIARETDB1 image dataset and 72.25% for DIARETDB0 image dataset. The proposed algorithm adapts adequately when compared with conventional algorithms, hence will act as a tool for segmentation.

The manuscript is well written and of a very interesting topic for de ophthalmologist community. The lack of continuous line at the right was very difficult to make the review of the paper.

1. Shortened the introduction into three main paragraphs of 500 words limit.
2. Do not give general and basic info of DR on intro that was known by ophthalmologist scientific community.
3. Reordered method section to improve comprehension, follow the flow chart diagram could be and excellent option
4. Re make Figure 10 and 11 to improve visual appearance.
5. In discussion, you may discuss the benefit of this technology for the clinical practice and why it is a good option to the daily clinical practice of the ophthalmologist.
6. References prior to 2005 must be updated to and actual one and the references out of indexed journals should be changed for a high indexed journals

Author Response

The point-by-point response to the reviewer's comments is in the attached file.

Author Response File: Author Response.docx

Reviewer 2 Report

Dear authors, 

thank you for presenting your work in this manuscript. You address a very interesting and important issue, as algorithms become more and more standard practice in clinical routine, especially detection of diseases. It is therefore important to modify and optimize these methods. Please consider to adapt the manuscript according to the following suggestions. 

First of all, it is difficult to review the paper when there are no page numbers and no line numbers. Did you use the mdpi/applied science template? Please double check if there was a formatting error. Therefore, the following points refer to paragraphs and not to lines. 

A source should be added in the WHO sentence about epidemiology in the introduction.

I find the last paragraph of the introduction (starting with "The paper is organized") not helpful. Since the manuscript follows a common format, I would remove the paragraph. 

Also in the introduction: I would not consider ocular ultrasound a popular methos for diabetic retinopathy. The clinical imaging paragraph should be improved and specified, consider an ohthalmological paper. 

You added a paragraph called "related work", which I think is very helpful for the reader. However, you introduce paragraph 1.2 but there is no 1.1. Otherwise this is an interesting paragraph which helps the reader understand the issue. 

The methods are thoroughly described and are easy to follow.

Compared to the rest of the manuscript, the Discussion is very short. I think this issue is of high clinical relevance so its clinical applications and limitations should be discussed more thoroughly. Please consider adding a paragraph concerning general limitations (e.g. are there any limitations to the DIARETDB datasets?). Same for the last two sentences of the discussion, the clinical relevance should be pointed out more detailed. 

Spelling/ grammar errors: english language should be improved, please consider these changes: 

Abstract: 

• "Diabetic retinopathy"
• "... to segment and detect hemorrhages..."

Introduction: 

• "diabetes mellitus"
• "...vision loss and alerts many people..."
• "...they are often blurry, poorly illuminated, and have a norrow field..."
• "...method for hemorrhage detection"

2.2.1. 

• "lighting conditions"

2.2.3. 

• "Previous experiments suggested that..."

3.3.

• "mutates histograms"
• "... does not have a problem with over-saturation"

Discussion

• "a higher area under the curve (AUC) value shows a higher capability of the classifiers"

Conclusion

• "... segments hemorrhages efficiently..."
• "Our results suggest"
• "has better segmentation characteristics, it could be helpful in other ophthalmological conditions as well" (

Thank you for your work and kind regards. 

Author Response

The point-by-point response to the reviewer's comments is in the attached file.

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

Comments solved