Fully Automatic Analysis of Muscle B-Mode Ultrasound Images Based on the Deep Residual Shrinkage U-Net

Round 1

Reviewer 1 Report

It can be published in the present form.

Author Response

Thanks for your comments. We have made some improvements to the paper.

Reviewer 2 Report

The authors present a method for medical semantic segmentation. The method is based on a Unet architecture, which is embellished with residual shrinkage modules. The dice loss is used for the network's training. The method is described in detail and has some novelties and thus is worth publishing.

What is lack in the paper is some ablation study and some comparison with other methods. More specifically,

a) The methods that are used in comparison are derivatives from the basic unet, which acts as a partial ablation study, showing how some modules work. The reviewer would like to see more direct comparison with other methods in the field, such as the method in [13], [22], [23].

b) It would be interesting to check the effect of shrinkage, i.e. soft-thresholding with an ablation experiment.

c) What would be the effect of using hard thresholding, instead of soft-thresholding.

d) Which deconvolution approach do you use ? Is it transposed convolution ? Please define.

e) For clarity, you should define that the features produced on the left are concatenated to those by upscale at the right in a typical unet fashion. This is not mentioned in the text.

 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

This study aimed to proposes an automatic muscle ultrasound image analysis method based on image segmentation. I have the following suggestions.

1. What is the novelty of this study although several muscle ultrasound image analysis approaches have been proposed earlier?
2. Please add a paragraph about the contribution of this article in a bulleted form at the end part of the Introduction section.
3. Authors should describe the source and details of the dataset used in this study.
4. Figures are unclear and of low quality. Authors should add more details in the captions of the figures.
5. Authors should mention the feature extraction process of muscle parameters such as muscle thickness in detail.
6. Authors should present the training and validation accuracy graphs of the proposed model with changes in the size of the dataset
7. Authors should report more performance measures of their model, such as accuracy, sensitivity, specificity, precision, and negative predictive value.
8. The results and discussion section need to be extended and improved. Authors must make discussion on the advantages and drawbacks of their proposed system with other studies adding a discussion section.
9. From the writing point of view, the manuscript must be checked for typos and the grammatical issues should be improved.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 4 Report

This paper presents automatic analysis of muscle B-Mode ultrasound images based on deep residual shrinkage U-Net. It`s a nice research paper. The authors mentioned the details in an efficient way. The authors should add future work and give some details about it. Nice research work.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 3 Report

Thanks for addressing the comments.