Non-Invasive Blood Pressure Estimation Using Multi-Domain Pulse Wave Features and Random Forest Regression

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

In this manuscript, the authors have proposed a blood pressure estimation method, which can be used for simple and reliable daily blood pressure monitoring. Some meaningful work has been done, but the following issues should be addressed before publication.

(1) The Introduction is not sharp enough, and this section lacks critical comments on main  literatures.
(2) Many models of machine learning have been widely used to process complex signals, but what is the innovation and contribution of the used regression model？
(3) In Figures 13, 15 and 16 , the fonts of coordinates are not clear, so the font size should be larger enough.
(4) Can you explain the relationship between the blood pressure and sample data? In Figure 15(a), when the sample data changes from 4 to 10, the good estimated results are obtained, however, when the sample data is at 15 and 18, the results don't seem ideal. Why？
(5) In Part 5, the content of this paragraph only involves "Conclusions", not "Discussion". So it is suggested to change the title to "5. Conclusions".
(6) In addition, the English expression of this manuscript need to be further improved.

Comments on the Quality of English Language

In order to enhance readability, the English expression of this manuscript need to be further improved.

Author Response

Thank you very much for your review of our manuscript entitled "Non-Invasive Blood Pressure Estimation Using Multi-Domain Pulse Wave Features and Random Forest Regression" (No. 3528072).  The authors sincerely thank the editor and reviewers for their valuable suggestions.  We have carefully read the comments of the editor and reviewers, and provided corresponding relies point-by-point.  All the changes have been highlighted in a red color in the revised version (Marked Revision).  In addition, a final clean copy of this manuscript is also available after the changes are accepted.You can read the word document  to see our detailed response. Thanks again for your advice!

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

Recommendation: Minor

Comments: Cardiovascular disease remains a leading cause of mortality globally. Hypertension, in particular, can be a silent but severe threat to health. Without early detection and proper management, it can lead to serious complications such as heart attacks, strokes, and other cardiovascular and cerebrovascular issues. The application of machine learning in processing signals and mapping pulse signal features to systolic and diastolic blood pressures is indeed a significant and promising research direction. By leveraging machine learning algorithms, such as the random forest regression model mentioned earlier, it's possible to extract valuable insights from complex data and improve the accuracy and efficiency of blood pressure measurements. This can ultimately lead to better healthcare outcomes and more personalized treatment plans. In this article (electronics-3528072), the authors designed a flexible piezoelectric sensor and its corresponding circuit for pulse signal measurement and preprocessing. Diastolic and systolic blood pressures were determined successfully by using a random forest regression model. Here are some suggestions for this review article.

1. It is advisable that the written expression in the paper should steer clear of colloquial language.
2. In Figure 2, the voltage response of the pressure-sensitive sensor at 0.2N and 1N alone is insufficient. The sensor's response under different pressure levels should be investigated, and the linearity of this relationship should be analyzed.
3. If Figure 5c is derived from Figure 5b, this should be explicitly indicated within Figure 5b for clarity.

Comments on the Quality of English Language

No

Author Response

Thank you very much for your review of our manuscript entitled "Non-Invasive Blood Pressure Estimation Using Multi-Domain Pulse Wave Features and Random Forest Regression" (No. 3528072).  The authors sincerely thank the editor and reviewers for their valuable suggestions.  We have carefully read the comments of the editor and reviewers, and provided corresponding relies point-by-point.  All the changes have been highlighted in a red color in the revised version (Marked Revision).  In addition, a final clean copy of this manuscript is also available after the changes are accepted.You can read the word document  to see our detailed response. Thanks again for your advice!

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

The authors' study is quite interesting, and the paper is worthy of publication. However, I am not entirely convinced Electronics is the most suitable journal for this work. A more bio-oriented journal, such as Medical & Biological Engineering & Computing or Computer Methods and Programs in Biomedicine, might be a better fit. The paper presents an interesting approach to non-invasive blood pressure estimation; however, before publication, a thorough revision is required because several aspects require clarification and revision.

• Did the 45 volunteers sign an informed consent form? This information should be explicitly stated.
• How were the 27 volunteers selected for training and the 18 for testing? The selection process is crucial to assess the robustness of the proposed method.
• Why were age and sex not included in Table 1 as part of the volunteers' characteristics? These are relevant factors that should be reported.
• In Figure 5c, is the single-period pulse wave a single waveform, or was it obtained as an average? This should be clarified.
• The procedure described in lines 206-215 is unclear and requires a more detailed explanation.
• Figure 9 should be better explained and discussed to enhance its interpretability.
• I suggest either removing lines 268-272 or providing a clearer explanation of the differences with the adopted method, as the current description lacks clarity.
• Where are the time-domain features t1,t3,t4,T,h1,h3,h4 defined? Their definitions should be explicitly stated.
• The definition of the Spectral Energy Ratio (SERn) in Equation (7) is unclear and should be explained more clearly.
• The result reported in lines 341-342, "It can be found that in the wavelet domain, the pulse signal is mainly distributed in the frequency band of 0.78-1.56Hz"—is this consistent with the feature SER5? This should be discussed.
• In my opinion, all five machine learning methods used should be briefly discussed, along with a comparison of their performances.
• What exactly is meant by "hyperparameter search method" in line 452? A brief explanation should be provided.
• What is the "Gini index" in line 482? A definition or reference should be included.
• Are the results shown in Figure 16 consistent with a Principal Component Analysis (PCA) of the same data? This should be verified and commented on.

Addressing these issues will significantly improve the manuscript's quality and comprehensibility. Furthermore, I strongly recommend that the authors avoid using acronyms in the abstract and modify certain expressions for clarity and correctness. Specifically:

• Instead of "This paper uses" (lines 123-124), the phrase should be reformulated as "In this paper, [….] is used...".
• The expressions in line 141 and line 507 should be revised, as nothing is "designed in this paper"; rather, it is discussed.
• What does "for this reason" mean in line 334? The reasoning behind this statement should be explicitly stated.

 

Comments on the Quality of English Language

A careful review of language and technical presentation will improve the overall readability of the manuscript.

Author Response

Thank you very much for your review of our manuscript entitled "Non-Invasive Blood Pressure Estimation Using Multi-Domain Pulse Wave Features and Random Forest Regression" (No. 3528072).  The authors sincerely thank the editor and reviewers for their valuable suggestions.  We have carefully read the comments of the editor and reviewers, and provided corresponding relies point-by-point.  All the changes have been highlighted in a red color in the revised version (Marked Revision).  In addition, a final clean copy of this manuscript is also available after the changes are accepted.You can read the word document  to see our detailed response. Thanks again for your advice!

Author Response File: Author Response.pdf

Reviewer 4 Report

Comments and Suggestions for Authors

This article is devoted to the urgent problem of developing a non-invasive method for measuring blood pressure based on pulse wave signals. For this purpose, a flexible piezoelectric sensor and its electronic circuit were created, which recorded pulse signals. From these signals, 30 features were extracted in the time, frequency and wavelet domains, on the basis of which a random forest regression model was built. This model allowed estimating systolic and diastolic pressure with an average absolute error of 1.70 mm Hg and 1.41 mm Hg, which meets the accuracy standards of the American Association of Medical Devices.
Comments on the article:
1. It is necessary to formulate the research problems more clearly, showing what exactly the novelty of the proposed approach consists of.
2. The literature review should not only contain a list of works. It is also necessary to add a critical analysis of their shortcomings and a comparison with the proposed method.
3. It is advisable to add an explanation and comparison of why a flexible piezoelectric sensor was chosen, and not alternative solutions.
4. It is necessary to add information on how the system was calibrated before the experiments and what control methods were used to check the accuracy.
5. It is necessary to add an analysis of the characteristics of the sample of study participants (age, gender and physiological variability, etc.).
6. 5 algorithms were tested, but there is no explanation why, for example, deep neural networks or ensemble methods were not considered.
7. It is necessary to expand the evaluation metric by adding R², which will help to better characterize the prediction error and the model as a whole.
I recommend that this article be accepted with minor edits.

Comments on the Quality of English Language

This article is devoted to the urgent problem of developing a non-invasive method for measuring blood pressure based on pulse wave signals. For this purpose, a flexible piezoelectric sensor and its electronic circuit were created, which recorded pulse signals. From these signals, 30 features were extracted in the time, frequency and wavelet domains, on the basis of which a random forest regression model was built. This model allowed estimating systolic and diastolic pressure with an average absolute error of 1.70 mm Hg and 1.41 mm Hg, which meets the accuracy standards of the American Association of Medical Devices.
Comments on the article:
1. It is necessary to formulate the research problems more clearly, showing what exactly the novelty of the proposed approach consists of.
2. The literature review should not only contain a list of works. It is also necessary to add a critical analysis of their shortcomings and a comparison with the proposed method.
3. It is advisable to add an explanation and comparison of why a flexible piezoelectric sensor was chosen, and not alternative solutions.
4. It is necessary to add information on how the system was calibrated before the experiments and what control methods were used to check the accuracy.
5. It is necessary to add an analysis of the characteristics of the sample of study participants (age, gender and physiological variability, etc.).
6. 5 algorithms were tested, but there is no explanation why, for example, deep neural networks or ensemble methods were not considered.
7. It is necessary to expand the evaluation metric by adding R², which will help to better characterize the prediction error and the model as a whole.
I recommend that this article be accepted with minor edits.

Author Response

Thank you very much for your review of our manuscript entitled "Non-Invasive Blood Pressure Estimation Using Multi-Domain Pulse Wave Features and Random Forest Regression" (No. 3528072).  The authors sincerely thank the editor and reviewers for their valuable suggestions.  We have carefully read the comments of the editor and reviewers, and provided corresponding relies point-by-point.  All the changes have been highlighted in a red color in the revised version (Marked Revision).  In addition, a final clean copy of this manuscript is also available after the changes are accepted.You can read the word document  to see our detailed response. Thanks again for your advice!

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

All my main concerns have been carefully responded, So I have no further comments.