Dementia Classification Using Deep Reinforcement Learning for Early Diagnosis

Round 1

Reviewer 1 Report

This work proposed a deep Reinforcement learning method for the dementia classification task. But the experiments have been implemented on only one dataset and compared with a few no-fashion existing approaches, which are not strong enough to support their work. Too many redundant figures are shown as expected to support the proposed method, but too much analysis with preliminary experiments is meaningless. The fate of this manuscript will be decided in the next round upon addressing the recommended suggestions adequately.

********Cons and general comments*********

- The logic of related work is messy. The section of **Related work** should not simply paraphrase other works but should divide them into categories and point out the improvement of this work compared with them. Table 1 is also redundant. Just profoundly focusing on the most related works. Also, I think it could be better if related works were categorized by their limitation.

- No description of CNN Layers is shown in Fig 1. What's the backbone of it? Add things precisely and technically.

- Wrong figure. The index of the citation paper appeared 38-40 in Fig.7, but these papers are not shown in the reference.

Above all, I think this work needs to be improved in two aspects.

- Rewrite the section of related work.

- Implement more experiments and add further results, analysis, and evaluation metrics to support the superiority of the proposed work instead of trivial analysis. (To do so, you can request additional time from the journal/Editor)

- Quality of Fig-1 is poor. Likewise, please pay attention to the other figures and their unnecessarily bigger sizes.

- Add a proper caption to Fig-1. Likewise, there should be sub-numbers for Fig -2 and Fig-3, such as (a) and (b), (i) and (ii), respectively.

- Many typos and wrong sentences were found throughout the manuscript, such as

In Figure1depict step 1 input Dementia dataset.

The above is just one example, and there could be more; please revise accordingly. In fact, the paper is poorly written and a bit hard to understand due to poor writing style. Therefore, checking the whole manuscript from scratch and fixing the issues is recommended. The best option would be to proofread your manuscript from any proofread service.

-The conclusion is too short and could be improved.

 

- Provide the exact reference for each cited dataset upon appearing for the first time in the text. 

Comments for author File: Comments.pdf

Author Response

Dear Sir 

Greetings

I wish to submit revised version of original research paper for publication in Applied Sciences Journal, titled “Dementia classification using Deep Reinforcement learning for early Diagnosis”. I am uploading the response of the First reviewer queries attached herewith for  kind consideration.

Thank you for your consideration. I look forward to hearing from you.

Sincerely,

Arshad Hashmi

 

Author Response File: Author Response.docx

Reviewer 2 Report

In this paper, Deep Reinforcement Architecture (DRA) is utilized to classify dementia.  In particular, the dataset's most important problem, the class imbalance, is resolved via the XgBoost Deep Reinforcement method, Finally, the proposed technique increases performance measures.In this work, the authors achieved the following main contributions.

1.Improving class imbalance through deep reinforcement learning and iterative policy by balancing instances of each class.

2.Convolution neural network with deep reinforcement learning is used to improve segmentation.

3.Use structure-based learning to improve classification.

However, there are some problems with this paper.

In Abstract, The advanced methods and outstanding contributions of this study in dementia diagnosis are unclear.The authors need to rethink the plausibility of the presentation. 

The RELATED WORK portion should include author synthesis and evaluation of past research, rather than solely enumerating a list of past studies. The authors should clearly describe gaps in the current state of the science, and how the current paper will contribute something new to the scientific body of knowledge. The literature review does not give readers a clear picture of the current research progress in Dementia early diagnosis.

Comments for author File: Comments.pdf

Author Response

Dear Sir 

Greetings

I wish to submit revised version of original research paper for publication in Applied Sciences Journal, titled “Dementia classification using Deep Reinforcement learning for early Diagnosis”. I am uploading the response of the Second reviewer queries attached herewith for  kind consideration.

Thank you for your consideration. I look forward to hearing from you.

Sincerely,

Arshad Hashmi

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

The suggested comments are adequately addressed. Therefore, we do accept the revised version of the submitted manuscript.