Electrocardiogram Signal Classification Based on Mix Time-Series Imaging

Round 1

Reviewer 1 Report

Thank you for providing me with the opportunity to read “ECG signal classification method based on Mix time-series imaging”. The paper proposes an electrocardiogram classification method that encodes one-dimensional ECG signals into the three-channel images, named ECG classification based on Mix Time-series Imaging (EC-MTSI). Specifically, this hybrid transformation method combines Gramian angular field (GAF), recurrent plot (RP), and tiling, preserving the original ECG time series’ time dependence and correlation. I have the following comments:

·         Line 4. Add the other part for the sentence ending on “these signals”. What happens when the causes mentioned are ignored? Then present what your paper does

·         Line 4, replace electrocardiogram with ECG since you already established the acronym

·         Please add the results (in 1-2 lines) to the abstract.

·         How much is abundant? Please provide the exact number in line 10. Again, avoid using vague words and clearly present the results.

·          Line 19, rethink the use of “essential” here.

·         Add more references to lines 19-35.

·         The problem is not well presented in the study. Please establish the context first through proper references and literature and proceed to provide the solution.

·         Please provide the objectives of the study in a numbered format for the ease of readers to locate and apprehend.

·         The novelty of the study should be elaborated on in a separate paragraph in the introduction section. What are the innovations brought in by this study? This must be added to the introduction section.

·         Line 56 again, how much is abundant? Please provide the number here and at other places in the paper to avoid such vagueness.

·         The ECG line in figure 1 is not properly visible; please use a higher resolution.

·         Please move the figure next to where they are first referred to enhance the readability of the paper. For example, figures 2 , 3 and 4 are very far from where these are mentioned.

·         What is the value addition of Figure 4? Please explain in detail or remove the figure if there is not much value-added.

·         Please justify the reason for selecting and using data from 1975-1979. Why not use recent data that may show more accurate data? How does the technology development from ECG in between 1979 and 2022 affect your data? What were checks in place to make sure there is no such effect on the data obtained from machines in 1979 versus machines in 2022?

·         Please explain each sub-section of Figure 5 (from a to e) in detail. All the sections must be properly discussed and key takeaways presented and elaborated on.

·         The discussion on Tables 2 and 3 must be improved beyond the mere representation of numbers in the text. What are the key takeaways here? Please add real discussion here.

·         The paper needs a standalone, detailed, and comprehensive discussion section. The authors must compare the current study with previously published ones in the discussion section in a tabular format. The results of the study should be compared with other similar studies, and the key improvements should be discussed.

·         Add the exact improvements to the conclusion section. The readers would benefit from knowing the exact improvement percentages and how your paper has superior or comparable values.

·         The limitations are not present in the conclusion section. These must be added.

·         Please add the implications (practical and research) to the conclusion section.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

In this study, Cai et al. applied a combination of Gramian angular field (GAF), recurrent plot (RP), tiling, and artificial neural networks to classify ECG signals. The authors have concluded that the ResNet50V2 + DenseNet169 network presents the highest classification accuracy.

This manuscript suffers from some issues that prevent me from suggesting publishing its current version. My comments are listed below:

1-      I recommend not to use abbreviations in the title.

2-      The “method” word in the title seems redundant.

3-      The abstract should present the key numerical finding of the study. Your abstract presents no numerical results.

4-      The following reference states that cancer is the leading cause of death globally, which conflicts with your statement in Lines 14-18.

Bray, F., Laversanne, M., Weiderpass, E. & Soerjomataram, I. The ever-increasing importance of cancer as a leading cause of premature death worldwide. Cancer 127, 3029–3030 (2021).

5-      Many classification algorithms have been applied in different scientific and industrial applications, including the KNN algorithm in medical (https://doi.org/10.1109/ACCESS.2019.2955754), fuzzy-logic in fault diagnosis (https://doi.org/10.1016/S0098-1354(02)00214-4), Wavelet transform and recurrent neural networks scheme in petroleum engineering (https://doi.org/10.1115/1.4047595), PLS-DA in chemistry (https://doi.org/10.1039/C3AY40582F), and so on.

It is better to mention these algorithms in your paper to help readers know about other classification algorithms.

6-      It is not traditional to assign a separate section to the literature review. Combine sections 2, 2.1, and 2.2 with the introduction.

7-      The last paragraph of section 2.2 should clearly state the contribution and novelty of your study (based on my previous comment, this paragraph should appear at the end of the introduction of the revised manuscript).

8-      Please remove the red underlines in Figure 1.

9-      The authors should provide some valid references for those equations obtained from the literature. This article (https://doi.org/10.1115/1.4050781) may be used as a reference for Eqs. (6) to (9).

1-  Since your databank is not available in your mentioned reference (i.e., [27]), it is better to report it in the supplementary material.

1-  I highly advise the authors to report the graph of the confusion matrix of the best classifier (i.e., ResNet50V2+DenseNet169) for the training, testing, and all databank.

1-  Rewrite the conclusion by focusing on your own achieved results.

 

 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Thank you for addressing my comments. Some minor observations:

1. Group the references in line 302 as 40-48. Also such a huge cluster may not be needed. Consider reducing these.

2. Table 5 must not appear in Conclusion. Please move to most relevant place (discussion)

3. The comma at the end of equation 1 can be removed

All the best

 

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

Dear authors

Thank you for thoroughly addressing my concerns.

I think the revised manuscript is now acceptable in Electronics.

Congratulation on accepting your article.

Regards

Author Response

Dear reviewer,

Thank you for your revision!

Regards