Predicting Pattern Standard Deviation in Glaucoma: A Machine Learning Approach Leveraging Clinical Data
, Daniel R. Chow 2, Andres C. Ramos 1, Guillermo Rocha 1 and Paul Harasymowycz 1,3,*Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsCongratulations to the authors on a well designed study.
How does RNFL compare to the automated neural network results alone? This still required technicians and an expensive OCT technology?
Can the La Roche Glaucoma Risk Calculator can be used to flag individuals who might benefit from more advanced assessment tools like the artificial neural network (ANN) for predicting Pattern Standard Deviation (PSD). The La Roche calculator estimates risk for glaucoma based on three clinical factors: age, intraocular pressure (IOP), and central corneal thickness (CCT). It is designed to stratify patients by glaucoma risk with high sensitivity and specificity, identifying those at higher risk (e.g., with a composite score ≥6) who may require more detailed evaluation or monitoring. This does not require expensive OCT and VF testing initially. Does the ANN have a higher sensitivity and specificity? Please add to discussion.
The authors end by stating “These findings open new avenues for obtaining practical, technician-independent, and objective VF estimations in glaucoma patients” but technicians are require to obtain IOP, CCT, RNFL oct images?
Do the authors have any specific recommendations on the most efficient and cost effective way to perform glaucoma screening in higher risk elderly unserved populations base on their experience.
Author Response
We greatly appreciate the reviewers’ thoughtful evaluation of our manuscript. Please find enclosed the revised version along with our detailed responses to each of the reviewers’ comments.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
Comments and Suggestions for AuthorsThe research on glaucoma in this paper is very interesting, but the following problems need to be improved:
1) The description of the method is very limited. The author needs to carefully describe how to use neural networks.
2) The quality of Figure 1 is too poor and contains very limited information, so it needs to be remade.
3) Why use paired two tailed t-test? Is the sample size appropriate? Does it conform to the t distribution?
4) The author should collect some real world data to test the model.
Author Response
We greatly appreciate the reviewers’ thoughtful evaluation of our manuscript. Please find enclosed the revised version along with our detailed responses to each of the reviewers’ comments.
Author Response File: Author Response.pdf
Reviewer 3 Report
Comments and Suggestions for AuthorsThe manuscript "Predicting Pattern Standard Deviation in Glaucoma: A Machine Learning Approach Leveraging Clinical Data" is interesting. Several comments are listed below.
1) Two publicly accessible repositories from a study conducted at Dankook University Hospital and Gyeongsang National University Hospital and the GRAPE data set were chosen for the study. However, the reason for this selection is not clear.
2) Relatively, the glaucoma case's number (n = 541) is higher than control case number (202), two times higher. Matching the number of the cases may not be needed?
3) Glaucoma cases may vary like mentioned in the manuscript: non-glaucomatous eyes, primary open-angle glaucoma, or primary angle-closure glaucoma. How this can be divided and analyzed?
4) Although it is written, age and country's factors should be well-discussed with several different cases in the discussion part.
Author Response
We greatly appreciate the reviewers’ thoughtful evaluation of our manuscript. Please find enclosed the revised version along with our detailed responses to each of the reviewers’ comments.
Author Response File: Author Response.pdf
Round 2
Reviewer 2 Report
Comments and Suggestions for AuthorsI think the paper can be accepted.
