Sparse Decomposition of Heart Rate Using a Bernoulli-Gaussian Model: Application to Sleep Apnoea Detection

Round 1

Reviewer 1 Report

In the submitted manuscript, the authors present an application of sparse decomposition of the heart rate in sleep apnoea detection. They described the background and internal structure adequately and they provided a very comprehensive description of methods. The proposed method gives encouraging results in sleep apnoea detection. However, there are a few points the authors should address:

-The introduction needs rewording and dividing into a few paragraphs (marked as indents) to improve the readability although it provides the background and objectives of the study.

-Materials and Methods:

-General: Use indents to separate different aspects of the text that do not need to be moved to designated subsections.

-Subsection 2.1:

“Sparse representations are widely used in signal and image processing,
computer vision, machine learning, array processing, data mining, system identification. . . [11 – 24 ].” could be paraphrased as “Sparse representations found many applications in signal and image processing, including computer vision, machine learning, array processing, data mining, and system identification [11-24]”.

-Material:

The authors use two datasets: internal dataset introduced in “Data inventory” section and PhysioNet Apnoea ECG Database introduced in subsection 3.2.

I suppose that the second dataset was applied to verify the method on a publicly available dataset. Am I right?

The description of the internal dataset should include the following important points:

-The number of subjects and their basic demographic data (age, gender, weight, height, cardiovascular conditions (yes/no)),

-Data acquisition setup (data acquisition protocol, devices, registered signals, including the number of recordings)

-Basic information on signals (sampling frequency, signal length).

Unfortunately, only the information on acquired signals (heart rate, accelerometer data, sleep scoring and hypnogram, episodes of apnoea, episodes of limb movements), the number of recordings, a usual length of recording, and a polysomnographic device, are available.

The description of datasets should be moved to a separate subsection within "Materials and Methods" section.

Author Response

Dear Reviewer,

Many thanks for your time and helpful comments. We did our best to take each suggestion into account to improve the paper overall quality.

You will find attached a document containing direct answers to your questions and suggestions, along with a global changelog between original submission and updated paper.

Warm regards,

Author Response File: Author Response.pdf

Reviewer 2 Report

The topic of the paper is very interesting; indeed, polysomnography is complex and expensive, so finding alternatives methodologies for apnoea detection is certainly of interest for both research and clinics.

Some points of the methods need to be better described, but, considering the general quality of the paper, I am confident that the authors will be able to improve the manuscript.

Page 7, lines 116-117. From what said in the Introduction section, I suppose that the authors themselves recorded the ECG signals. For sake of clarity, it would be better to specify it here.

Who defined the RDI index (eq. 5)? Who defined the apnoea severity ranges? (p. 7, lines 135-136)

Page 7, line 131: it is not clear for me if it is proper to consider a ratio between # of events and hours? Why is the "duration" of events not considered?

Fig. 4 is also not clear to me. First of all, the values ​​are all positive, while the chosen window goes from – 50 s to -5 s (page 8, line 151). Assuming there is an absolute value, why 5 coincides the minimum value of histogram, while 50 is not its maximum value?

Page 9, line 159: do the authors mean that the threshold is adaptive?

Fig. 10: On which basis, do the authors say that the signal is of bad quality? Did they add an SNR estimation to the software or is it just a visual analysis?

Fig. 15: the authors state that the model works perfectly; is this also true in the 4-6 h interval?

Conclusions: in the Introduction section (page 1, lines 37-38), the authors state that the literature on apnoea detection using HRV is extensive. However, a comparison with this literature is completely lacking. This is a fundamental part of the research work to understand if the proposed method is really promising. The correlation values ​​obtained with respect to the reference signals, in fact, are not very high (0.88 on the authors' signals and even lower on those of Physionet).

Author Response

Dear Reviewer,

Many thanks for your time and helpful comments. We did our best to take each suggestion into account to improve the paper overall quality.

You will find attached a document containing direct answers to your questions and suggestions, along with a global changelog between original submission and updated paper.

Warm regards,

Author Response File: Author Response.pdf

Reviewer 3 Report

This manuscript proposed a signal processing approach used for apnoea detection by means of heart rate, which is significance in practice.

I have some questions for this paper.

1. The ordinate marking should be supplemented in most of the figures of this manuscript,such as the variable name and unit should be standard.

2. In Section 3.2, the existing research using the Physionet Apnoea ECG Database should be compared with this the proposed method. Otherwise, the superiority of the proposed method can not be determined.

3. It is suggested to divide the Results Section into two parts, Results and Discussion, to ensure the sufficient discussions. 

4. And the future works could be supplemented.

Author Response

Dear Reviewer,

Many thanks for your time and helpful comments. We did our best to take each suggestion into account to improve the paper overall quality.

You will find attached a document containing direct answers to your questions and suggestions, along with a global changelog between original submission and updated paper.

Warm regards,

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

The authors have addressed my comments. However, there are a few minor points that should be addressed:

1. "This is the only pre-processing we perform before applying our algorithm" could be removed or moved to

2. Lines 288-292 may significantly benefit from proofreading and copy-editing

3. Caption of Figure 16: I recommend using "magnified" instead of "zoomed-in"

4. I would recommend proofreading and copy-editing the discussion.

5. Line 384: I would recommend citing the PhysioNet Apnoea Challenge results.

6. "begindocument/before" at the beginning of the manuscript is not necessary.

Author Response

Please find below our responses to each of your suggestions and commentaries. Thank you again for your time and input.

Warm regards,

1. "This is the only pre-processing we perform before applying our algorithm" could be removed or moved to

We moved the pre-processing remark at a more appropriate location

2. Lines 288-292 may significantly benefit from proofreading and copy-editing

We modified lines 288-292 to give clearer details on what we did.

3. Caption of Figure 16: I recommend using "magnified" instead of "zoomed-in"

We modified "zoomed-in" to "magnified" as suggested.

4. I would recommend proofreading and copy-editing the discussion.

We tried to improve the discussion as recommended.

5. Line 384: I would recommend citing the PhysioNet Apnoea Challenge results.

We are not sure about line 384 (in our document it is the bibliography); however, we specified Physionet's challenge best results in detection (new paragraph in Section 2.4.3). We point out in the new paragraph that the challenge's winner detected apnoea through the recurrence of apnoea rather than individual events as we do.

6. "begindocument/before" at the beginning of the manuscript is not necessary.

We apologise but we did not find any reference to "begindocument/before", neither in the compiled PDF nor in the LaTeX file.

Reviewer 2 Report

The authors have satisfied all my requests.

Author Response

Thanks for your time and helpful contributions in improving our paper,

Warm regards,