Computer-Aided Diagnosis for Early Signs of Skin Diseases Using Multi Types Feature Fusion Based on a Hybrid Deep Learning Model

Round 1

Reviewer 1 Report

 

 In this paper, the authors presented a classification method of different skin lesion categories using dermoscopic images. They applied a number of CNN models for feature extraction and then ML methods were used for the classification process.

The paper needs extensive revisions before considering for publication, such as:

-        The novelty is completely not clear. As noticed, the applied method all existing methods, I did not see any new method developed. Or, maybe the description of the contribution is not clear. So, clarify this issue.

-        The complexity and computation time should be studied and discussed.

-        What about the dividing of the training and testing samples? Did you test the model with completely unseen samples?

-        More details about the model are need. For example, pseudocodes, or source codes.

-        In your writing, care about the words that you used, for example, “early detection” kindly, note that this is not an application, it is just a classification approach, in case of making it as an application, you can claim it is an early detection system. Also, it is not an automatic system? How did you consider it like that? Fix those exaggerating words.

-        Re-plot figure 2, it looks not original.

-        The results in figure 5 cannot be seen clearly. Re-plot them.

-        Limitations and challenges must be discussed.

-        A proof reading is also needed.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

- The article has no novel contribution because it replicates the same architectures as other networks shown in the state of the art, with some difference, but in my opinion they have no great contribution.

- They present experimentation with a set of public images and do not offer experimentation or testing with one of their own.

- The images are of poor quality

- They do not explain in detail how the training and classification results were obtained and they are not expressed graphically. 

- In line 37, it is outdated (2016) what they report from the academy of dermatology.

- In 81 they do not mention which methodology they followed to determine the architecture of the neural network or was it trial and error?

- In 90 it is mentioned qualitatively, that the proposed work showed promising results to diagnose 90 different skin diseases, but it is not expressed quantitatively.

 

- It is not explained in detail why they removed the last fully connected layer.

- In 176 It is not explained in detail why they used these parameters for training. Did they follow a methodology?

 

- In 245, stage 3 it is mentioned that the training time of the model is significantly reduced, in my opinion it would be clearer to mention quantitatively what is meant by significantly.

 

- The main stages of the proposed approach should be explained in detail and not just mentioned.

 

- You do not mention if you are working with a hospital and are generating your own training set?

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments for author File: Comments.pdf

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

The paper was improved and it can be accepted for publication. 

Reviewer 3 Report

I accept the manuscript in its current form.