Efficient Learning of Healthcare Data from IoT Devices by Edge Convolution Neural Networks^†

Round 1

Reviewer 1 Report

The authors present a hybrid architecture named EdgeCNN which merges the potential of Cloud and Edge computing for learning about healthcare data from IoT devices. They also present a deep learning model for ECG inference which can be deployed to run on edge smart devices. They design a data enhancement method for ECG based on deep convolutional generative adversarial network to expand ECG data volume. They carry out experiments on two datasets to evaluate  the deep learning model of ECG classification based on EdgeCNN. Their approach has advantages respect to the state of the art, for what concerns: inference time, storage, running memory and power consumption.

The work is well written and it builds on a conference paper that has been substantially extended. 

Minor Comments:

1. Not clear the meaning of "Agile" in the title.
2. row 50 something is wrong with the sentence "which is successfully deploy it on edge devices.
3. row 120 avoid using the contracted form
4. Algorithm 1 reads badly
5. Figure 3, try to better place the captions
6. row 282 a reference to a Figure is missing, in fact at the top of the page there is a Figure with no caption at all.
7. row 329 "platform ,which" --> "platform, which"
8. row 362 "privacy.Fifth" --> "privacy. Fifth"
9. Figure 11 if printed in black and white does not allow to distinguish the three lines. The authors may use the same approach used in Figure 10

Author Response

Dear reviewer,

Based on the revision comments, we made the following revisions to the paper:

1. Our structure has the advantages of low latency and high accuracy, so we used ‘agile’ before, but after thinking about it, we think it’s more appropriate to replace agile with efficient, so we modified the title.
2. We corrected this problem, and the revised sentence is "Then we successfully deploy it on edge devices."
3. We use cannot instead of can’t.
4. We modified this algorithm.
5. We reset the title of figure 3.
6. We captioned this picture which missed before.
7. We fixed this error.
8. We fixed this error.
9. We repainted Figure 11 using the method of Figure 10.

Thank you for your review and look forward to your reply.

Reviewer 2 Report

Thank you for submitting the paper to the journal. The scope of the paper is compatible with the aims and scope of the journal. The title is clear and concise; there is no need to change it.

I assess the paper as well-prepared. However, I would suggest and even require preparing the revision to improve the quality of the paper. To some extent, they are only technical.

Here I list them in chronological order:

1) Line 5. Please rephrase the sentence part: „The contributions include: „

2) Line 12. Please explain abbreviation „I/O”, before that was also not explained ECG and CNN

3) Generally, the abstract is properly prepared; I have no comments on the content.

4) Line 18 – The introduction should be numbered as 1. Then you should renumber the other sections of the manuscript.

5) Line 20/21 please revise „to monitor the physiological

21 and psychological status of the human body.”

6) Line 22 should be „face” and now is „faces”

7) Lines 36-39  -the source(s) needed, you mention MIT-BIH research, and there is no information about the publishing it

8) Section Related Work: This part is not fully prepared. Please enhance the literature review. Especially health informatics, data management, real-time monitoring are, as you mentioned, very up-to-date and popular topic, there is a lot of publications about similar research. Please check Medline or Web of Science to find them. They also should be discussed. Please avoid colloquial speech like „hot topic” (line 73) – please amend it in the whole paper. Please mention data security and other risks/challenges in this part of the paper, as well as AI and IoT development.

9) Section 2.2. is well prepared. This is one of the best in the paper.

10) Line 182 – Please do not use capital letters in headings (check it in the whole paper)

11) Line 188 – Please amend the big letter at the beginning

12) Line 260 amend „figure”, also big letters in line 274, 283, 288

13) The assumptions of the study, all the steps of the research procedure are well described. However, there is some kind of discussion in 4.1.3. Please refer better to „old architecture” in Table 3 – it is not clear – is this the „traditional cloud” or not? It should be mentioned in the text.

14) Conclusion – this part is highly underdeveloped. Firstly, you should present the impact on knowledge and organisations/companies which may benefit from your results. You present the filling some research gap, but this impact is significant, also on the development of knowledge in the field you study.

Secondly, please indicate the limitations of your study. You even mentioned some of them in the main body, sometimes not directly. Please address that matter shortly in one paragraph.

Thirdly, I would like to ask you to add some proposal of future research directions for the studied topic – should the similar research be prepared with the use of another method(s), research framework, hybrid approaches and so on? What are your views on the future research?

15) Please revise the English language. There are especially some minor issues like a lack of „the”, „a”, „an”, wrong form of the verb etc.

Taking into consideration the above comments, for now, I require major revision. However, the revision is not severe, and it will not influence the main body so much, e.g. results, methods and assumptions.

I look forward to seeing a revised paper.

Author Response

Dear reviewer,

Based on the revision comments, we made the following revisions to the paper:

1. We have changed the sentence to "These contributions include:" and the capitalization of the following content. We have some questions about this comment. Do we need to modify the content or just the format is wrong.
2. We have added the full names of I/O, ECG and CNN.
3. Thank you for your evaluation.
4. We renumber it.
5. We change this sentence to “Internet of things devices are widely used in human health monitoring.”.
6. We corrected this error.
7. We have added references to MIT-BIH data.
8. We read related papers on health informatics, data management, real-time monitoring, AI and IoT, and added discussions on these papers in related work, and corrected some wording issues.
9. Thank you for your review.

10, 11, 12. We corrected this error.

13. We added an explanation of the two architectures and changed old architecture to traditional cloud architecture.
14. We rewrite the conclusion part in the paper.
15. We re-read the article and corrected some grammatical errors.

Thank you for your review and look forward to your reply.