A pH Monitoring Algorithm for Orifice Plate Culture Medium

Round 1

Reviewer 1 Report

The authors in the article propose a solution to one of the applied problems using computer vision methods. The topic of the article is relevant. The structure of the article is classical. All necessary sections are present. The level of English is acceptable. The article is easy to read. The quality of the figures is mediocre. For example, to make out what is furrowed in Fig. 6 is difficult. Formulas (all two) are written carelessly. The References section is also carelessly designed. The article cites 22 sources, not all of which are relevant.

The following remarks can be made on the material of the article:

1. Formally, the article contains all the necessary sections, but in fact, the methods section does not contain a description of new scientific results. It contains an extensive algorithm for processing and analyzing images instead of a mathematical apparatus that substantiates the rationality of the necessary transformations.

2. At one time, I solved the problem of converting an image frame from RGB to HSV using OpenCV. I blurred the frame and applied the color lookup function. The inRange() function does a search on a "from and to" basis, i.e. from what to what color to select pixels for it. Then I processed each pixel of the frame. If the pixel is white, then fill it with gray. Well, then the rectangle () function selected objects, simply drawing a rectangle around the selected area. In my opinion, this is no worse than what the authors proposed. I ask authors to substantiate my assertion.

3. In chapter 3.1.4 of the Experiments section, the authors presented a theorem without a clear formulation and proof. If this is a well-known theorem, then a reference should be made. If unknown, then it must be clearly formulated and proved.

4. Looking at the graphs shown in Fig. 7, it becomes obvious that these are non-linear functions (power and potentially trigonometric). In this case, the use of linear regression by the authors to describe the experimental data is puzzling.

5. Curve in Fig. 8 is based on 6 data points. This is absolutely not enough to draw any conclusions about the object under study. Mathematical statistics clearly determines how many experiments need to be carried out in order to prove the adequacy of the research method (depending on the number of factors, measurement accuracy, etc.) I ask the authors to justify the adequacy of the results obtained using appropriate statistical criteria.

Author Response

According to the comments of the reviewer, the author revised the description of the manuscript, including the normalization of formulas and charts. And carefully check the relevance of references. The groove in Figure 6 is caused by the reflection of the orifice cover.

The main corrections in the paper and the responses to the reviewer’s comments are as flowing:

Comment 1: Formally, the article contains all the necessary sections, but in fact, the methods section does not contain a description of new scientific results. It contains an extensive algorithm for processing and analyzing images instead of a mathematical apparatus that substantiates the rationality of the necessary transformations.

Response 1: In terms of the innovation of the article, compared with the traditional contact measurement method, this paper proposes a rapid non-contact whole process monitoring method. Moreover, the components related to pH in the HSV color model are analyzed, and the relationship between them is modeled by regression. Finally, the experimental verification proves that the method proposed in this paper can be applied to practical measurement. This is not mentioned in previous methods.

Comment 2: At one time, I solved the problem of converting an image frame from RGB to HSV using OpenCV. I blurred the frame and applied the color lookup function. The inRange() function does a search on a "from and to" basis, i.e. from what to what color to select pixels for it. Then I processed each pixel of the frame. If the pixel is white, then fill it with gray. Well, then the rectangle () function selected objects, simply drawing a rectangle around the selected area. In my opinion, this is no worse than what the authors proposed. I ask authors to substantiate my assertion.

Response 2:The first step of monitoring in this paper is to extract the effective region of the captured image to obtain the solution part in the image. In fact, there are many ways to extract effective regions from images. Drawing a rectangle around the selected area proposed by the reviewer is also a method. However, this requires manual operation. In order to realize the automatic extraction of the solution region, this paper uses edge detection and binarization to realize automatic extraction of the solution part. Thus, the whole pH monitoring process is more efficient without human intervention.

Comment 3: In chapter 3.1.4 of the Experiments section, the authors presented a theorem without a clear formulation and proof. If this is a well-known theorem, then a reference should be made. If unknown, then it must be clearly formulated and proved.

Response 3: We have modified the description of this part. The second section method adds the mathematical modeling of the third section. In chapter 3.1.4, formula 1 is not a theorem. It is a definition established to study the significance of the three components of HSV. These three expressions are also common statistical concepts, namely first-order moment, second-order moment, and third-order moment. Corresponding references are added to the formula.

Comment 4: Looking at the graphs shown in Fig. 7, it becomes obvious that these are non-linear functions (power and potentially trigonometric). In this case, the use of linear regression by the authors to describe the experimental data is puzzling.

Response 4: In Formula 3, the relationship between pH and Hmean has been modified as an exponential function.

Comment 5: Curve in Fig. 8 is based on 6 data points. This is absolutely not enough to draw any conclusions about the object under study. Mathematical statistics clearly determines how many experiments need to be carried out in order to prove the adequacy of the research method (depending on the number of factors, measurement accuracy, etc.) I ask the authors to justify the adequacy of the results obtained using appropriate statistical criteria.

Response 5: There are 30 data points in Figure 8. In each solution image, we have selected five different positions, and there are a total of six solutions with different pH values. The total number of actual measurement data obtained in this way is 30, which can ensure the adequacy of fitting data. See Table 3 for specific data. Moreover, in 3.3, the established model is analyzed and verified.

Here are the parts I modified and the reasons.

1. In the abstract, we reduced some descriptions about the use of cells and increased the innovation of this paper. The innovation of this paper is to avoid using immersion sensors for pH measurement in order to prevent cell pollution when culturing cells with an orifice plate. To solve this problem, we proposed to use the HSV algorithm to determine pH. We modified the description of the linear relationship between pH and Hmeans.
2. In lines 108-114, we added a description of HSV.
3. The change made in line 146 was to modify the subsequent description of the verification method of the built model.
4. We modified the position of the picture in Figure 3.
5. “Color moment” in line 204 was removed.
6. The scale and marking of Figure 7 had been changed.
7. In line 247, the selection rules of the points established by the experimental model were changed.
8. Table 3 was changed to use a pH meter to measure the pH of cell fluid with 30 holes.
9. Equation 3 changed to a newly fitted curve.
10. Figure 8 was changed to a new modeling curve.
11. In lines 259-268, the verification method of the previous fitting curve is changed. Modified to verify the fitted curve using SSE and MSE.
12. In Table 4, the pH position errors of hole 1 and hole 3 caused by filling errors in the previous version had been modified.
13. In section 3.3.2, the verification of the feasibility of the model was added. The accuracy of the fitting curve was analyzed by calculating the root mean square error of the pH predicted by the fitting curve and the pH measured by the pH meter.
14. In formula 4, we recalculated the root mean square error.
15. In line 323, we updated RMSE.
16. Finally, some references are revised and added.

Author Response File: Author Response.pdf

Reviewer 2 Report

This manuscript reports a pH monitoring algorithm for orficie plate culture medium.  This approach is not novelty. There several approaches at literature, which are based on color changes using HSV cameras or smartphones cameras. However, those approaches are not for pH determination.

·         At Figure 6 authors present cropped liquid images . There are visible shadows. How authors solve the problem of shadow effect on extracted data from picture.

·         Theorem 1 – Color moment color moment eigenvalue , should be changed to color moment eigenvalue

·         Figures 6 cropped images should be aligned.

·         Authors mentioned: However, traditional methods present changes in the color of the culture fluid as a yellow hue at pH 6.8-7.2. So it is difficult for the naked eye to judge the specific value of  its pH, which can only be determined empirically

Why then authors didn’t extend the range of prediction in both ways ?

·         All conclusions and discussion are made only on 6 measurements?  Did authors made replicates or external samples set to test prediction ability of obtained model ?

·         Page 11 line 318 InGivenhe ?

Author Response

Dear  Reviewers:

It has been revised according to the suggestions, and the descriptions in other places in the manuscript have been checked and proofread.

In terms of the innovation of the article, compared with the traditional contact measurement method, this paper proposes a rapid non-contact whole process monitoring method. Moreover, the components related to pH in HSV color model are analyzed, and the relationship between them is modeled by regression. Finally, the experimental verification proves that the method proposed in this paper can be applied to practical measurement. This is not mentioned in previous methods.

Comment 1:At Figure 6 authors present cropped liquid images. There are visible shadows. Response 1: How authors solve the problem of the shadow effect on extracted data from pictures.

In Figure 6, the brightness is indeed inconsistent, which is caused by the reflection of the orifice cover. In this paper, the HSV color model is used to extract the H component of a color image, so as to eliminate the influence of shadow on pH measurement to a certain extent. At the same time, in the later experiment, using Hmean to describe the average value of the color of the whole solution is equivalent to shadow filtering, which can also play a role.

Comment 2: Theorem 1 – Color moment color moment eigenvalue, should be changed to color moment eigenvalue.

Response 2: Theorem 1 – Color moment color moment eigenvalue, had been changed to color moment eigenvalue.

Comment 3: Figures 6 cropped images should be aligned.

Response 3: The position of Figure 6 has been changed.

Comment 4: Authors mentioned: However, traditional methods present changes in the color of the culture fluid as a yellow hue at pH 6.8-7.2. So it is difficult for the naked eye to judge the specific value of its pH, which can only be determined empirically? Why then authors didn’t extend the range of prediction in both ways?

Response 4: Because when most cells are in an acidic environment, it will affect their metabolism and other behaviors. The downward trend of pH of cell culture medium is stepwise, and as seed cells and other therapeutic cells that need strict culture conditions, if their pH is already acidic, their culture medium may have exceeded its replacement time.

Comment 5: All conclusions and discussion are made only on 6 measurements?  Did authors made replicates or external samples set to test prediction ability of obtained model ?

Response 5:There are 30 data points in Figure 8. In each solution image, we have selected five different positions, and there are a total of six solutions with different pH values. The total number of actual measurement data obtained in this way is 30, which can ensure the adequacy of fitting data. See Table 3 for specific data. Moreover, in 3.3, the established model is analyzed and verified. According to the review opinions, this part has been improved in the text.

Comment 6: Page 11 line 318 InGivenhe ?

Response 6: It has been corrected.

Here are the parts I modified and the reasons.

1. In the abstract, we reduced some descriptions about the use of cells and increased the innovation of this paper. The innovation of this paper is to avoid using immersion sensors for pH measurement in order to prevent cell pollution when culturing cells with an orifice plate. To solve this problem, we proposed to use the HSV algorithm to determine pH. We modified the description of the linear relationship between pH and Hmeans.
2. In lines 108-114, we added a description of HSV.
3. The change made in line 146 was to modify the subsequent description of the verification method of the built model.
4. We modified the position of the picture in Figure 3.
5. “Color moment” in line 204 was removed.
6. The scale and marking of Figure 7 had been changed.
7. In line 247, the selection rules of the points established by the experimental model were changed.
8. Table 3 was changed to use a pH meter to measure the pH of cell fluid with 30 holes.
9. Equation 3 changed to a newly fitted curve.
10. Figure 8 was changed to a new modeling curve.
11. In lines 259-268, the verification method of the previous fitting curve is changed. Modified to verify the fitted curve using SSE and MSE.
12. In Table 4, the pH position errors of hole 1 and hole 3 caused by filling errors in the previous version had been modified.
13. In section 3.3.2, the verification of the feasibility of the model was added. The accuracy of the fitting curve was analyzed by calculating the root mean square error of the pH predicted by the fitting curve and the pH measured by the pH meter.
14. In formula 4, we recalculated the root mean square error.
15. In line 323, we updated RMSE.
16. Finally, some references are revised and added.

In all, I found the comments quite helpful, and I revised my paper point-by-point. Thank you and the review again for your help!

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

I formulated the following comments to the basic version of the article:

1: Formally, the article contains all the necessary sections, but in fact, the methods section does not contain a description of new scientific results. It contains an extensive algorithm for processing and analyzing images instead of a mathematical apparatus that substantiates the rationality of the necessary transformations.

2: At one time, I solved the problem of converting an image frame from RGB to HSV using OpenCV. I blurred the frame and applied the color lookup function. The inRange() function does a search on a "from and to" basis, i.e. from what to what color to select pixels for it. Then I processed each pixel of the frame. If the pixel is white, then fill it with gray. Well, then the rectangle () function selected objects, simply drawing a rectangle around the selected area. In my opinion, this is no worse than what the authors proposed. I ask authors to substantiate my assertion.

3: In chapter 3.1.4 of the Experiments section, the authors presented a theorem without a clear formulation and proof. If this is a well-known theorem, then a reference should be made. If unknown, then it must be clearly formulated and proved.

4: Looking at the graphs shown in Fig. 7, it becomes obvious that these are non-linear functions (power and potentially trigonometric). In this case, the use of linear regression by the authors to describe the experimental data is puzzling.

5: Curve in Fig. 8 is based on 6 data points. This is absolutely not enough to draw any conclusions about the object under study. Mathematical statistics clearly determines how many experiments need to be carried out in order to prove the adequacy of the research method (depending on the number of factors, measurement accuracy, etc.) I ask the authors to justify the adequacy of the results obtained using appropriate statistical criteria.

The authors responded to all comments. I didn't like all of their answers, but the edited version of the article isn't so bad that it shouldn't be published. In general, I support the publication of this article.

Reviewer 2 Report

Authors made a correction according to suggestions  however in conclusion part they didn't change InGivenhe