Systolic Blood Pressure Estimation from PPG Signal Using ANN
Round 1
Reviewer 1 Report
Recommendation: Resubmission
Comments:
l The introduction should present the background, significant, other related works with more details concerning the AI algorithms, which are not shown in the paper.
l Figures: figure2 and 3 seems too blur and do not meet the fig requirement (300 PPI). Also please check all the rest figures which have the same resolution problems.
l Figure 5 and 6 can be combined into one figure, and also please check the necessities of using 11 figures in the paper instead of organizing them into a more logic and compact sequence.
l Figure with experiment of data Acquisition using PPG on volunteers is missing.
l Segmentation of text should be improved
l Parameters and more logics about the applied ANN should be descripted in details.
Author Response
Dear Review, thanks for your comments.
Please see the attachment
Author Response File: 
Author Response.pdf
Reviewer 2 Report
This research is not new and add a little to the recent bulk of knowledge.
The authors would greatly gain to complete a thorough review of the literature.
Author Response
Dear Review, thanks for your comments.
Please see the attachment
Author Response File: Author Response.pdf
Reviewer 3 Report
Comments to Authors (General)
· The keywords should be at least five and the authors should add some keywords in the paper.
· The authors shall have to express the long form of “SD” in Table 2 and 3.
· The authors shall have to modify the high resolution pictures in the whole paper.
Comments to Authors (Specific)
· The authors shall have to modify the “Abstract” with specific methodology for solving the identified problems in details.
· The “Introduction” section shall have to be revised again. The authors shall have to present the introduction of the research works with very relevant recent works with up-to-date citations in last three years’ period. The authors shall have to modify it.
· The authors shall have to mention the age level of dataset for 47 patients in ICU.
· The authors shall have to mention the mathematical model of the applied digital filters with specific frequency ranges.
· The authors shall have to present the system block diagram for specific model with Neural Network. And also the authors shall have to discuss on the specific ANN with specific functions such as train function and transfer function for each layers.
· The authors shall have to present the performance of the ANN model with specific discussions.
· The authors shall have to compare the obtained results with recent works of other researchers’ and the statistic table shall have to be presented before conclusion.
· The authors shall have to mention the robustness of the proposed model.
Author Response
Dear Review, thanks for your comments.
Please see the attachment
Author Response File: Author Response.pdf
Reviewer 4 Report
This paper implements and analyses a machine learning model to estimate the systolic blood pressure from the PPG signal. The topic of the article and the proposed method are very interesting. Overall, the paper is well written, well organized, and reports an interesting application of the machine learning model.
Please consider the following comments and recommendations:
-Limited literature review of deep learning models adopted for the analysis of PPG signals. Incomplete introduction or related work.
- Please clarify how the application of “zero-pad all the periods, so as to have one hundred samples each” can affect the model performance.
- In the paper, the authors state that “Moreover our reference is not a continuous signal, as it was in the case of the arterial pressure signal. Therefore, in order to obtain an SP value for each PPG period, two consecutive SP records have been averaged between two acquisition instants.” However, in Figure 11 a BP signal is represented with continuous BP values variation. Please clarify how BP values variation is present if an “averaged between two acquisition instants” was performed.
- The results analysis would benefit from a deeper comparison with results obtained by other methods and/or authors to the same problem.
Author Response
Dear Reviewer, thanks for your comments.
Please see the attachment
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
Too many seperated paragraphs, that makes reviewers difficult to catch the focus: If several minor paragraphs in the work share a common background or are related to the same subject, the structure must be significantly improved.
The revised part should be highlighted with a color.
Author Response
Dear Reviewer, we are going do upload the new version, following your advice.
Thank you,
Best Regards,
Benedetta
Reviewer 2 Report
Revision is relatively satisfying.
Author Response
We tried to improve References
Reviewer 4 Report
Review of the second version
The last version of this paper, which implements and analyses a machine learning model to estimate the systolic blood pressure from the PPG signal using ANN, was not improved enough in order to be accepted for publication in the present form.
- It continues with an incomplete introduction and state-of-the-art review. Very poor description of the nature and features analysis of PPG waveform without proper references. This is a very known topic and exists a lot of literature about it. For example, one of the recent reviews was published in March 2022, DOI: 10.3389/fphys.2021.808451
- The paper continues with a very limited literature review of deep learning models adopted for the analysis of PPG signals. The author's answer “We are trying to enrich this part but we already reported all the works we were able to find in literature” is quite strange because there are a lot of publications. For example, some of them are:
1) Deep learning models for cuffless blood pressure monitoring from PPG signals using attention mechanism Biomedical Signal Processing and Control 65(9455):10230 DOI:10.1016/j.bspc.2020.102301
2) The Application of Deep Learning Algorithms for PPG Signal Processing and Classification Computers 2021, 10, 158. https://doi.org/10.3390/computers10120158
3) Normalization of photoplethysmography using deep neural networks for individual and group comparison…. Scientific Reports | (2022) 12:3133 | doi.org/10.1038/s41598-022-07107-5
4) Deep recurrent neural network-based autoencoder for photoplethysmogram artifacts filtering” Computers & Electrical Engineering, Volume 92, June 2021, 107065 https://doi.org/10.1016/j.compeleceng.2021.107065
- It was also suggested that the analysis of the results would benefit from a deeper comparison with results obtained by other methods and/or authors to the same problem. However, conclusions and result discussion were not improved and continued almost the same;
- In References 7 and 8 are not clear, and complete information is needed;
- The new version of the article contain also the wrong format for the references in the following lines: 94, 99, 115, 117, 147, 168, 223, 228, 229.
Author Response
Dear reviewer, thank you very much for the suggestions; we referred to the papers you pointed out.
About the results, we already makes some improvement by comparing our model performance with that of similar project, in terms of aim, such as systolic blood pressure value estimation using artificial neural network.
As for the references we add some information about the references that were not clear at all. Instead, regarding the wrong format, we didn't really understand what you were referring to; is there something wrong on the footnote? If so, what?
Thank you and Best Regards
Round 3
Reviewer 1 Report
My comments are as follows:
1. Figure 5 depicts a generic ANN structure rather than the specific ANN algorithm that should be described for this work. please update the figure 5.
2. 12 figures is still too many; reorganize them into more sensible and compact groups.
Author Response
We removed some figures which were not extremely significant. Figure 1, Figure 5, Figure 9
