Classical FE Analysis to Classify Parkinson’s Disease Patients

Round 1

Reviewer 1 Report

The purpose of the manuscript is to propose several experiments where PD patients and HC patients were classified based on facial images, using LBP and HOG. The authors claimed to achieve an accuracy of 80.4%. The manuscript lacks a proper literature survey, experimental design, or explanation of why their research is novel, which part is novel, also authors need to compare their results with other existing recently published work in a similar field. Significant revision is required before this manuscript can be accepted. The major issues as listed below:

Major points:

1.      Authors have used LBP and HOG but both of them are extensively researched while classification between PD patients and HC patients. Authors are requested to do a better literature survey and identify why their experiment using LBP and HOG is different from existing research. This is very critical to make this manuscript useful in any way.

2.      Support vector machine has been used here for classification purposes, but already in many research papers, SVM has been used for the same purpose. So authors need to clarify why they have again chosen SVM, what is novel here? (FYI, according to a review paper, “support vector machine (SVM) was the most dominant algorithm used”, https://downloads.hindawi.com/journals/acisc/2021/9917246.pdf)

3.      The author claimed to achieve an accuracy of 80.4% which is not significant enough compared to contemporary research. Authors need to compare their result in a separate table with similar and contemporary research and discuss elaborately why their result is worth publishing.

4.      The authors' literature survey is insufficient, first, it should contain more recent works from 2021 and 2022. Second, just stating existing research is not a proper literature survey, authors need to provide existing shortcomings with those research as well.

5.      The author should mention where LBP and HOG have been used for the same objective as presented in this manuscript.

6.      In section 2 authors keep describing the techniques which are already discussed in many papers and review articles, authors need to clarify why they have chosen those techniques, and how their implementation is different from existing research. Just describing the techniques which were discovered by someone else can not be accepted as a fruitful contribution to a journal paper.

7.      Authors mentioned that “The focus of this work is not only to be able to classify PD patients vs HC subjects, but to perform a more deep analysis of the results and to understand what the classifier is seeking to separate the classes”, which is that “deep analysis”?

 

8.      Authors need to dedicate a significant section related to the “limitations and constraints” of their research or experiment. 

Author Response

Dear reviewer, please find attached a rebuttal letter with the response to all comments, corrections, and criticisms performed by the reviewers and the editor.

Thank you for your comments.

 

Author Response File: Author Response.pdf

Reviewer 2 Report

1.       Kindly write the abstract in a concise and succinct manner; there is no requirement for background basic study in the abstract. The first three to four lines can shift to the introduction.

2.       Write specific author contributions

3.       Why SVM?

4.       Authors should try deep learning methods

5.       Script is completely confusing

6.       Final tuned parameters should mention in Table

7.       MCCNN is mentioned in the methodology, but it’s not explained or used for experimentation

8.       Results are not so effective

9.       Why machine learning approach preferred?

10.   Please remove unnecessary capitalizations in the manuscript

11.   Grammatically needs to recheck again

12.   Needs to recheck the structure of the paper as Introduction, Literature and Motivation, Contributions of the proposed work, Database, Proposed methodology, Experimental results, Discussion, Conclusion + Future scope.

13.   Needs to include an ablation study (If possible)

14.   Statistical tests are also missed

 

15.   Conclusions also need to re-write again

Author Response

Dear reviewer, in the attached fil you will find the response to all comments, suggestions, and corrections requested by the reviewers.

Thank you very much for your comments.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

The authors have answered the queries satisfactorily. "Presenting a baseline" can be an important issue but at the same time improving the accuracy and reliability of the technique are important too. That is where deep learning is doing better with a large chunk of data. I would like to suggest authors venture into the area where they can plan to use a hybrid deep learning model where they can combine the classical and deep learning techniques together so part of that can be explainable as well as the accuracy will be higher. 

 

Author Response

The authors of the manuscript would like to thank the reviewer for the suggestion. We certainly will incorporate mixtures of DL and ML methods in future studies.

 

Reviewer 2 Report

The authors included all the review comments provided by me. Please cite recent lit before publication.

1. Safi, Khaled, et al. "EMD-Based Method for Supervised Classification of Parkinson’s Disease Patients Using Balance Control Data." Bioengineering 9.7 (2022): 283.

2. Gómez-Gómez, L. F., Morales, A., Orozco-Arroyave, J. R., & Fierrez, J. Exploring Facial Expressions and Action Unit Domains For Parkinson Detection. Available at SSRN 4069648.

3. Prakash, Allam Jaya, and Samit Ari. "A system for automatic cardiac arrhythmia recognition using electrocardiogram signal." Bioelectronics and Medical Devices. Woodhead Publishing, 2019. 891-911.

Author Response

The authors would like to thank the reviewer for the comments. Suggested references have been added to the updated version of the manuscript.