A Non-Contact Photoplethysmography Technique for the Estimation of Heart Rate via Smartphone

Round 1

Reviewer 1 Report

In the Figure 4 what is the y-axis's dimensions?

In the Figure 6 what means the cardiac frequency?

Author Response

what is the y-axis's dimensions?

Intensity mean of the pixel

 

what means the cardiac frequency?

The heart rate is the number of times the heart contracts in a period of one minute (BPM). The normal frequency, at rest, beats between 50 and 100 beats per minute.

Author Response File: Author Response.docx

Reviewer 2 Report

GENERAL

The authors present a technique for estimating heart rate using an image tracking technique.  Generally well written, although some errors in English are apparent, please read through carefully.  The techniques and scientific method of the paper are robust and sound, however there are issues with the statistics I would like answered. From the discussion and conclusion I fail to conclude with the authors' claims.

INTRODUCTION

Mentions in the introduction about measuring blood flow (Line 74).  Please re-evaluate this statement.  Photoplethysmography is primarily a measure of volume. Actual flow rates would not be possible in the technique described, please look at Laser Doppler Flowmetry for the closest optical method of estimating blood flow.

METHODS

These are robust and provide an excellent level of detail.

RESULTS

Authors claim an error rate between 3% and 5%.  There is no calculation or statistical graph to demonstrate this.  Figure 6, the Correlation graph, does not even show a correlation coefficient or line of identity, which are standard.  At least for this type of analysis a comparison between devices (app vs pulse oximeter) should be performed and analysed using a technique such as the Bland and Altman method.

DISCUSSION & CONCLUSION

The authors claim there is no app currently available that calculates pulse rate in real time.  I don't fully agree to this claim as i think there is some confusion (maybe on my part) about the techniques described.  Are the authors claiming they can calculate pulse rate continuously throughout the capture period or intermittently?  In either case it will be useful to see a demonstration of this in the paper by giving the readers a set of results that plot time against pulse rate.  This will add validity to their claim and should be shown.  This will also add extra value to the clinical usefulness of the data as quick sudden changes in pulse rate can be viewed as a measure in basic heart/pulse rate variability analysis.  If the case is that continuous monitoring is not yet achievable by this method then I am not sure the author's claim is true; there are many apps that use similar techniques described to perform a measurement in under 10 seconds (Cardiio for the iOS platform has been doing this since 2012).  These apps can be used semi-intermittently (with user interaction). Please comment for or against your app vs. other apps on the iOS/Android market place.

FINAL REMARKS

One has to ask what the practical application of such a technique is of use on a device that is mostly kept in the pocket of the user.  As a medical device specialist I would argue that this technology (contact-less pulse rate monitoring) is better suited to clinical or even ambulatory environments where permanent video capture systems could be deployed, and only if the algorithms developed can be used in a continuous capture mode.  Maybe this would be a useful tool for GPs and doctors to have.  I think the consensus in the phone app market space is that periodic recording of heart/pulse rate is good enough, and the use of smart watches with integrated PPG sensors to do continuous monitoring would always be more practical to the consumer.  I acknowledge the practical implications of using sensors that may cause irritation or uncomfortable feelings for the user, as implied in the introduction, but if this was really a major concern then less people would wear watches.  The use of ECG monitoring is a separate issue.  Only ECG monitoring can give a true picture of the electrical activity of the heart, and heart rate is only a small part of what ECG is capable of, in the cases where patients need ECG monitoring and they have reactions to the electrodes the small amount of discomfort caused by those sensors is vastly outweighed by the benefit of such monitoring. The authors should consider altering their discussion and conclusion to reflect this.

Author Response

We are grateful for the comments, We are made it the changes to the document with the suggestions provided.
Thank you for the observations.
Kind regards

Rodolfo Maestre

Author Response File: Author Response.docx

Reviewer 3 Report

The manuscript describes the use of smartphone cameras for the detection of vital signs (heart rate). Although not new, the idea can contribute to the remote monitoring of vital signs.

The authors have put an effort into describing the motivation of the work, however, the manuscript requires extensive improvements in the description of the methodology, presentation of the results and discussion. Please see below for more specific comments.

 

Introduction:

I believe the introduction can be improved significantly, especially in the description of the current state-of-the-art in non-contact monitoring via PPG) and the rationale behind the paper. The idea of vital sign monitoring using a (smartphone) camera or any other video-recording is not novel (please see the literature on non-contact PPG or application developed by Philips Vital Signs Camera App). However, using these principles in a new application would have contributed to the current knowledge on the subject.

Material and Methods:

The authors have attempted to describe the methods used, however, several details appear to be missing.

- Why the authors have used the forehead as Region of interest?

- Why using the green channel of the RGB? Does this provide better results? The literature provides some suggestions on the use of the green wavelength, however, the authors failed in describing the reasoning why.

- The methods used are based on PPG, but I could not recognise a typical PPG waveform from the signals acquired. I understand this may have a lower quality compared to a contact PPG sensor. Even after filtering, the signal quality appears to have a very low SNR.

- Which filtering technique or filter has been used?

- Can the authors explain in more details the reasoning for using the derivative of the PPG signal?

- There are no details on the FFT technique used.

 

Results:

- The system has been tested on subjects, but there is no detail on the population investigated.

- How did the author calculate the reported errors? There is no description of the analysis used?

- A scatter plot does not fit the purpose of assessing agreement. A Bland & Altman analysis (or any other error analysis) will be more appropriate. The authors state ‘correlation’ in Figure 6, but a formal correlation analysis has not been performed.

 

Discussion/Conclusion:

The discussion requires extensive improvement as it is very brief and fails in discussing the results obtained, the literature context, limitations etc.

 

General comments:

The captions of the figures should be self-explanatory The manuscript requires extensive editorial editing and grammar checking. The initials in the author's contributions section do not match with the authors names.

Author Response

We are grateful for the comments, We are made it the changes to the document with the suggestions provided.
Thank you for the observations.
Kind regards

Rodolfo Maestre

Author Response File: Author Response.docx

Round 2

Reviewer 2 Report

Whilst some improvements have been made regarding mine and the other reviewer’s suggestions these all seem very superficial. For example the Bland and Altman Plot was suggested by myself and another reviewer, and whilst this has been included there is no real discussion around the results, only a comment in the results section.  The error bars and mean difference bar are also not labelled, had they been then the added comments from line 188 to 194 would have made more sense as they are reliant on pointing to the "95%" limits of agreement.  This feels very rushed.  Correlation calculations were suggested by myself and another reviewer, the authors have failed to do this. I would not consider this for publication until all suggestions have been adequately responded to (either as a comment to the reviewers for justification or within the text of the paper).

I have just noticed that you do not include information regarding the test population (subjects) of your study.  Please include.  If this was done with limited volunteers under different scenarios, i.e. rest, standing, after exercise etc. then this can be extrapolated as a study population also, but I would expect to see no less than 20 subjects for a review of this type.

There is still very little consideration for current technology or acknowledgment of the suggestions I made regarding the use of this technology (in the section I labelled in my original review as “FINAL REMARKS”).  There is also little acknowledgement regarding the “Cardiio” app on the iOS platform which performs more or less the same function using the same method (from what I can see).  If it does not please comment specifically and in reference to.

Whilst you defend not showing traditional looking PPGs because of low SNR because of non-contact the fact that the figures 3 and 4 are not very discernible does not lend a lot of confidence.  Others doing non-contact PPG have been able to show PPG signals, for example please see the work done by Patel, Z et al. (The effects of optical sensor-tissue separation in endocavitary photoplethysmography, IOP publishing).  Please superimpose the plot of the PPG (or ECG) waveform from your reference device.  Periodicity of the signal you claim is there then may be more visible.  The same can then be done for the FFT plot showing that the two signals (your image-based PPG and the reference plot) are coherently linked.

Discussions and conclusions need improvement to acknowledge the Bland and Altman analysis, study population, and correlation of results.

All added content suffers poor English editing.

Author Response

Reviewer.- Whilst some improvements have been made regarding mine and the other reviewer’s suggestions these all seem very superficial. For example the Bland and Altman Plot was suggested by myself and another reviewer, and whilst this has been included there is no real discussion around the results, only a comment in the results section. 

Author.- According to the reviewer's suggestion, these comments are added and expressed on line 214 to 227

 

Reviewer.- The error bars and mean difference bar are also not labelled, had they been then the added comments from line 188 to 194 would have made more sense as they are reliant on pointing to the "95%" limits of agreement.

Author.- These comments are supported as suggested by the reviewer on line 201-206

 

Reviewer.- This feels very rushed.  Correlation calculations were suggested by myself and another reviewer, the authors have failed to do this.

Author.- The correlation calculation was performed according to what reviewers 1 and 2 suggested, and these are expressed in Table 2 and line 232 to 235

 

Reviewer.- I would not consider this for publication until all suggestions have been adequately responded to (either as a comment to the reviewers for justification or within the text of the paper).

Author.- I offer an apology for not taking into account all the suggestions or recommendations expressed by you, but in this document you will find all the changes made, point by point

 

Reviewer.- I have just noticed that you do not include information regarding the test population (subjects) of your study.  Please include.  If this was done with limited volunteers under different scenarios, i.e. rest, standing, after exercise etc. then this can be extrapolated as a study population also, but I would expect to see no less than 20 subjects for a review of this type.

Author.- This reviser’s recommendation, was described on line 204 to 206 and table 2

 

Reviewer.- There is still very little consideration for current technology or acknowledgment of the suggestions I made regarding the use of this technology (in the section I labelled in my original review as “FINAL REMARKS”).  There is also little acknowledgement regarding the “Cardiio” app on the iOS platform which performs more or less the same function using the same method (from what I can see).  If it does not please comment specifically and in reference to.

Author.- The consideration or recognition of technologies similar to the one we propose is expressed within the article on lines 74 to 79 and 238 to 239.

 

Reviewer.- Whilst you defend not showing traditional looking PPGs because of low SNR because of non-contact the fact that the figures 3 and 4 are not very discernible does not lend a lot of confidence.  Others doing non-contact PPG have been able to show PPG signals, for example please see the work done by Patel, Z et al. (The effects of optical sensor-tissue separation in endocavitary photoplethysmography, IOP publishing).  Please superimpose the plot of the PPG (or ECG) waveform from your reference device.  Periodicity of the signal you claim is there then may be more visible.  The same can then be done for the FFT plot showing that the two signals (your image-based PPG and the reference plot) are coherently linked.

Author.- On this point expressed and suggested by the reviewer, I offer an apology for this confusion and provide an explanation on this subject, the signal obtained from Figures 3 and 4, is the product of an average of 256 frames of images, whereas a result the graphs of figures 3 and 4 are obtained, this procedure is performed according to the process described in the flow chart of figure 1, and the explanation of this has been described in lines 137 to 144.

 

Reviewer.- Discussions and conclusions need improvement to acknowledge the Bland and Altman analysis, study population, and correlation of results.

Author.- According to the reviewer's suggestion, the analysis of Bland and Altman is included in the conclusions section on lines 278-285.

 

Reviewer.- All added content suffers poor English editing.

Author.- According to the reviewer's suggestion, we are aware that the language of English is not our mother tongue, but we promise that once the content of the article is validated, we will proceed to request the services of a grammar professional, which may be the service suggested by the editorial magazine.

Author Response File: Author Response.pdf

Reviewer 3 Report

The authors have not assessed all the comments raised in my first review. 

Also, I cannot find a point-by-point reply to my comments. 

All parts of the paper (Introduction, Methods, Results and Discussion) still requires extensive work and improvement in order to reach publishable quality.

Also, language and grammar improvements are required. I suggest adopting professional proofreading if required. 

I suggest rejection and re-submission.

 

Regards. 

Author Response

Reviewer.- The manuscript describes the use of smartphone cameras for the detection of vital signs (heart rate). Although not new, the idea can contribute to the remote monitoring of vital signs.

The authors have put an effort into describing the motivation of the work, however, the manuscript requires extensive improvements in the description of the methodology, presentation of the results and discussion. Please see below for more specific comments.

Introduction:

I believe the introduction can be improved significantly, especially in the description of the current state-of-the-art in non-contact monitoring via PPG) and the rationale behind the paper.

Author.- Suggestions for improving the document in the introduction section were made on lines 55 to 64 and 77 to 84.

Reviewer.- The idea of vital sign monitoring using a (smartphone) camera or any other video-recording is not novel (please see the literature on non-contact PPG or application developed by Philips Vital Signs Camera App). However, using these principles in a new application would have contributed to the current knowledge on the subject.
Author.- This consideration or recognition of technologies similar to that presented are expressed within the article on lines 74 to 77 and 238 to 239 and in turn describes the contribution or differentiating innovation of our application on lines 239 to 242.

Material and Methods:

The authors have attempted to describe the methods used, however, several details appear to be missing.

Reviewer.- Why the authors have used the forehead as Region of interest?
Author.- Regarding the descriptive question or suggestion of a concept to be explained in the article, mentioned by the reviewer, it is mentioned that the information was incorporated in the document on line 116 to 117.

Reviewer.- Why using the green channel of the RGB? Does this provide better results? The literature provides some suggestions on the use of the green wavelength, however, the authors failed in describing the reasoning why.
Author.- Concerning the question, it is mentioned that the information was incorporated in the document on line 129 to 133.

Reviewer.- The methods used are based on PPG, but I could not recognise a typical PPG waveform from the signals acquired. I understand this may have a lower quality compared to a contact PPG sensor. Even after filtering, the signal quality appears to have a very low SNR.
Author.- The question described by the reviewer was appended in the document on lines 139 to 141.

Reviewer.- Which filtering technique or filter has been used?
Author.- The question described by the reviewer was appended in the document on lines 145 to 148.

Reviewer.- Can the authors explain in more details the reasoning for using the derivative of the PPG signal?
Author.- The question described by the reviewer was appended in the document on lines 153 to 159 and 178 to 183.

Reviewer.- There are no details on the FFT technique used.
Author.- The question described by the reviewer was appended in the document on lines 163 to 175, 176 to 183 and in figure 5.

Results:

Reviewer.- The system has been tested on subjects, but there is no detail on the population investigated.
Author.- According to the reviewer's recommendation, information on the population investigated on lines 204 to 206 is attached.

Reviewer.- How did the author calculate the reported errors? There is no description of the analysis used?
Author.- Following the reviewer's recommendation, information on the calculations made for the reported errors is attached, using the Bland & Altman analysis and the descriptive statistics in Table 2.

Reviewer.- A scatter plot does not fit the purpose of assessing agreement. A Bland & Altman analysis (or any other error analysis) will be more appropriate. The authors state ‘correlation’ in Figure 6, but a formal correlation analysis has not been performed.
Author.- Following the reviewer's recommendation, information on the calculations made for the reported errors is attached, using the Bland & Altman analysis and the descriptive statistics in Table 2.

Discussion/Conclusion:

Reviewer.- The discussion requires extensive improvement as it is very brief and fails in discussing the results obtained, the literature context, limitations etc.
Author.- These suggestions were applied in the document, in the conclusions section on lines 273 to 284 and 300 to 310.

General comments:

Reviewer.- The captions of the figures should be self-explanatory The manuscript requires extensive editorial editing and grammar checking.
Author.- According to the reviewer's suggestion, we are aware that the language of English is not our mother tongue, but we promise that once the content of the article is validated, we will proceed to request the services of a grammar professional, which may be the service suggested by the editorial magazine.

Reviewer.- The initials in the author's contributions section do not match with the authors names.
Author.- I apologize for the error of the authors' initials presented in the document, and I thank you for the observation of that involuntary error, and I express to you that the change has already been made.

Reviewer.- The authors have not assessed all the comments raised in my first review.
Also, I cannot find a point-by-point reply to my comments.
All parts of the paper (Introduction, Methods, Results and Discussion) still requires extensive work and improvement in order to reach publishable quality.
Also, language and grammar improvements are required. I suggest adopting professional proofreading if required.

Author.- According to the reviewer's suggestion, we are aware that the language of English is not our mother tongue, but we promise that once the content of the article is validated, we will proceed to request the services of a grammar professional, which may be the service suggested by the editorial magazine.

Author Response File: Author Response.pdf

Round 3

Reviewer 2 Report

Now that you state your major innovation in this work is the mathematical and filtering technique used I feel that the paper will benefit from the added content and description of this process (equations, specific techniques and maybe a diagramatic representation of how the process is performed).

Considering the time that has obviously been put into this paper and the  author's efforts to address all reviewer comments I feel that the paper does have significant content and support from the results to constitute a publication.  Overall some major reconstruction is required to better incorporate all the added content as there text does feel "broken" in places.  This  may be, however, an effect of English not being the author's mother tongue, and I suggest editorial intervention to re-structure this manuscript.

Author Response

We added the content and description of the process, according to the reviewer's suggestion, these new information are added and expressed on line 118 to 133

We use the MDPI English Editing service according to the reviewer`s suggestion.

[Manuscript ID: applsci-622789] MDPI English editing - Editing completed

Author Response File: Author Response.docx