Robotic System for Blood Serum Aliquoting Based on a Neural Network Model of Machine Vision

Round 1

Reviewer 1 Report

Aiming at the problem of providing reliable and accurate information on the concentration or activity of the analytes in venous blood samples, this paper deals with the robotized aliquoting of blood serum. The authors propose application of a vision system for determination of pipette immersing depth. Two recognition algorithms are developed, one of them is based on the use of HSV color palette, the other one is based on convolutional neural network.

Specific remarks to the paper are following:

- References to scientific literature seem to be insufficient. For example, the authors claim that "it is proposed to use machine vision algorithms to determine the fluid contour" (page 2, lines 66-657) but do not refer to any papers. On the other hand, the authors unnecessarily describe in detail image processing algorithms presented in a very old paper published in previous century (page 2, lines 69-80 as well as page 10 lines 309-323).

- Due to above mentioned insufficient reference to the works of other scientists, the authors do not clearly indicate what the novelty of their approach consists in.

- References to literature on automatic image recognition and its application in robotics do not involve medical issues. More references to the newest research in domain of medical robotics should be added to the paper (perhaps you can find some interesting papers in the Special Issue "Medical robotics" of MDPI Sensors journal). 

- The text should be carefully read by the authors in order to correct the mistakes like accidentally removed parts of sentences e.g. "system for aliquoting was considered. biosamples." (page 3, line 123), repeated sentences (compare page 3, lines 112-113 and lines 117-118, as well as 122-123, or not understandable phrases like "to form training and other samples" (page 10, line 303).

Author Response

Dear reviewer, thank you for your thoughtful review. We have carefully considered your comments and edited the article in accordance with your suggestions. We believe that your suggestions are effective and improve the readability of the article. Thanks again for your comments. Below is the answer to your question.

Q1: - References to scientific literature seem to be insufficient. For example, the authors claim that "it is proposed to use machine vision algorithms to determine the fluid contour" (page 2, lines 66-657) but do not refer to any papers. On the other hand, the authors unnecessarily describe in detail image processing algorithms presented in a very old paper published in previous century (page 2, lines 69-80 as well as page 10 lines 309-323).

Response: References have been added to literature indicating the use of machine vision algorithms to determine the contour of a fluid. The overly detailed description of the algorithms and the use of old literature sources have been reworked.

Q2: - Due to above mentioned insufficient reference to the works of other scientists, the authors do not clearly indicate what the novelty of their approach consists in.

Response: Clarifications were added about the novelty of the approaches used in the work, which consists in the development of two algorithms for determining the levels of interfaces of blood fractions in a test tube, one of which is based on the HSV color model, and the other is based on the U-Net neural network.

Q3: - References to literature on automatic image recognition and its application in robotics do not involve medical issues. More references to the newest research in domain of medical robotics should be added to the paper (perhaps you can find some interesting papers in the Special Issue "Medical robotics" of MDPI Sensors journal).

Response: Research on the application of robotics and automatic image recognition in medicine was reviewed. References have been made to some work in this area.

Q4: - The text should be carefully read by the authors in order to correct the mistakes like accidentally removed parts of sentences e.g. "system for aliquoting was considered. biosamples." (page 3, line 123), repeated sentences (compare page 3, lines 112-113 and lines 117-118, as well as 122-123, or not understandable phrases like "to form training and other samples" (page 10, line 303).

Response: These inaccuracies and repetitions in the text have been corrected.

Reviewer 2 Report

The work presented in the paper is a comparative analysis of two algorithms applied to a application of blood sample is quite reflective. 

In my opinion, 

1. If the author presents the percentage of accuracy of HSV colour model as mentioned for CNN will be additive to the findings.

2. The paper was quite justifying in the context of Machine vision; a 3D model is found in fig2. As the title say a robotic system: the system is conceptually explained in section 2, As implementation is done with respect to robotic system, if those results provided in the paper would have been more supportive in this context.

Author Response

Dear reviewer, thank you for your thoughtful review. We have carefully considered your comments and edited the article in accordance with your suggestions. We believe that your suggestions are effective and improve the readability of the article. Thanks again for your comments. Below is the answer to your question.

Q1: - If the author presents the percentage of accuracy of HSV color model as mentioned for CNN will be additive to the findings.

Response: The method for assessing the accuracy of the two developed algorithms was added in section 5 "Comparative analysis of algorithms". The technique is the same for both algorithms, which makes it possible to compare the accuracy of their work.

Q2: - The paper was quite justifying in the context of Machine vision; a 3D model is found in fig2. As the title say a robotic system: the system is conceptually explained in section 2, As implementation is done with respect to robotic system, if those results provided in the paper would have been more supportive in this context.

Response: The section " Experimental results and analysis" was added to the work.

Reviewer 3 Report

The article proposes a serum detection robot system based on visual algorithms and delta robots for the actual serum detection problem. The algorithm part is comprehensively introduced and the analysis is detailed. However, the reviewers still have a few comments, which I hope the author can consider

 

1.         Among the existing introductions of intelligent monitoring systems based on robot systems, there are only literatures 8 and 9. It is obvious that the author's literature collection is not comprehensive enough.

2.         The author proposes a serum detection robot system, but most of the whole article only describes the algorithm, while for the delta robot part, there are only simulated images (figure, 2), and no real physical prototype is provided. Since there is a description of the experiment in the following article, the real image information of the whole system such as the delta robot should be provided instead of some photos of the serum. The above is just hope that the author can prove that he has indeed built a serum detection robot system instead of simply proposing an algorithm.

3.         In the article, the author provided a large number of pictures of serum, whether it is necessary to provide corresponding medical certificates, the source of blood samples and other information according to the regulations of the journal, proving that the provided liquid is blood rather than other liquids.

4.         Comparison need to be carried out with existing techniques if possible compare it with recent trends

5.         Try to include and refer recent papers in references and mention their contribution in your Literature Survey

6.         The quality of Figure 12 is too poor and not beautiful, I hope the author can correct it.

Author Response

Dear reviewer, thank you for your thoughtful review. We have carefully considered your comments and edited the article in accordance with your suggestions. We believe that your suggestions are effective and improve the readability of the article. Thanks again for your comments. Below is the answer to your question.

Q1: - Among the existing introductions of intelligent monitoring systems based on robot systems, there are only literatures 8 and 9. It is obvious that the author's literature collection is not comprehensive enough.

Response: Additionally, work on the use of robotic systems and machinec vision in medicine, as well as the use of machine vision directly for recognizing the contour of a liquid, were considered. References have been made to some of the studied work in this area.

Q2: - The author proposes a serum detection robot system, but most of the whole article only describes the algorithm, while for the delta robot part, there are only simulated images (figure, 2), and no real physical prototype is provided. Since there is a description of the experiment in the following article, the real image information of the whole system such as the delta robot should be provided instead of some photos of the serum. The above is just hope that the author can prove that he has indeed built a serum detection robot system instead of simply proposing an algorithm.

Response: The section " Experimental results and analysis" was added to the work

Q3: - In the article, the author provided a large number of pictures of serum, whether it is necessary to provide corresponding medical certificates, the source of blood samples and other information according to the regulations of the journal, proving that the provided liquid is blood rather than other liquids.

Response: The biomedical ethics committee's conclusion was later sent to the article, confirming the use of real blood samples. In some experiments, a liquid simulating blood was used, which was indicated in the article.

Q4: - Comparison need to be carried out with existing techniques if possible compare it with recent trends

Response: Additionally, the advantages of the selected methods in comparison with other segmentation methods were noted. Existing methods are sufficiently considered and their shortcomings in solving specific problems are indicated (end of section 1 "Introduction"). Developments and new approaches to solve these shortcomings were presented, an analysis of the effectiveness of the proposed solutions was made.

Q5: - Try to include and refer recent papers in references and mention their contribution in your Literature Survey

Response: Literature references have been revised and newer references have been added.

Q6: - The quality of Figure 12 is too poor and not beautiful, I hope the author can correct it.

Response: Figure 12 has been corrected and replaced, the image quality has been improved.

Round 2

Reviewer 1 Report

The manuscript has been substantially improved. All reviewer's comments were addressed. However, correctness of all sentences must be carefully checked (e.g. page 3 line 131).

Author Response

Thank you very much for the comments. Numerous corrections to the sentences have been made to the article. New changes are highlighted in yellow.