Heart Sound Signals Classification with Image Conversion Employed

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Heart Sound Signals Classification with Image Conversion

Employed

The paper is about heart sound classification. The paper is well-written and well-organized. It is missing important references and some methods have to be explained better. The figures must be explained in detail. Conclusions are supported by the results.

About paper writing: the author repeated several time extended form and acronyms along the paper. Please revise this and remove all duplicated definition.

Detailed comments:

line 54: please remove Numbered lists can be added as follows:

line 76,126,130 abbreviation convolutional neural networks CNN is repeated

line 74 empirical wavelet transform (EWT), missing reference

line 102 Bayesian Long Short-Term Memory (LSTM) missing reference

line 114 PCG missing extended form

line 149 DWT missing extended form

line 176-179 repetition of acronyms…

line 211: in the accompanying figure: please give exact reference

line 220: Butterworth bandpass filter please give reference

figure 2: please add measurement unit to Y axis. Max and Min Y must be the same (left and right, before and after filtering).

Figure 3 is unclear. What the circle represents? The square? The solids? Also it make use of terms not explained in the paper, like inception,concat…

Figure 4 and figure 5 are the figure 3 zoom. So the same comment of figure 3 apply

Line 296: inception is poorly explained. Can you give detailed information about the method?

Table1: concat not explained

Table 1 shows transformation from 2D images (e.g. 256×256×1) to 3d images (128×128×4). Can you explain in detail how the algorithm works? How inception, concat etc are modifying the initial matrix of values?

Line 323 please define the ReLU function and/or add reference

Line 363: number of channel? Please explain or give reference

Line 387-392: are these indicators already used in the literature? Please add references

Table 8: please define as maxpooling and average pooling.

From table 8

Line 405: “average pooling is more sensitive to background information” is repeated in line 408-409 “Table 4 shows that average pooling is more sensitive to background information, which can integrate all information for decision-making and help classification”

Table 5: sigmoid not defined in the text please add explanation and reference. Relu: please capitalize as in the text

Table 6: what is ration? Ration not defined in the text please add explanation and reference

Table 10: 60% must be 60.00%

Line 446: GASF not defined - missing extended form

Line 472-473,487 please add measurement units

Table 11: LSBTM not defined - missing extended form

 

 

 

 

 

 

 

Comments on the Quality of English Language

sufficient

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The main question addressed by the research is how to improve the classification accuracy of heart sound signals for the detection of cardiovascular diseases.

 The topic of heart sound signal classification is relevant in the field of cardiovascular disease detection, as it provides a direct means of identifying such diseases. The research addresses the limitations of traditional statistical feature methods and temporal dimension features in achieving good classification accuracy. The proposed PANet framework, which incorporates a new partition attention module and Fusionghost module, aims to improve the feature extraction and classification tasks for heart sound signals. The research also compares the performance of the PANet framework with other advanced algorithms, demonstrating its effectiveness in achieving high classification accuracy rates.

 

 However:

 

 - The authors could provide more details on the implementation of the partition attention module and Fusionghost module in the PANet framework.

 

- Further explanation of how the bispectrum representation is obtained from the heart sound signal would enhance the clarity of the methodology.

- The authors could consider conducting additional experiments on different datasets to validate the performance of the PANet framework in diverse scenarios.

- The authors should include a comparison with more state-of-the-art algorithms in the field to provide a more comprehensive evaluation of the PANet framework's performance.

- It would be beneficial to include an analysis of the computational resources required by the PANet framework to assess its practicality and scalability.

- The authors could consider conducting a sensitivity analysis to evaluate the robustness of the PANet framework to variations in input parameters and noise levels.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

The article is well-written in general. The following points are required to be addressed:

11. The first two paragraphs in the introduction section should be merged. Here, the focus should be on the importance/motivation of automated analysis and classification of heart sound signals using computer-based methods. In other words, it should clearly state the target problem with appropriate references.

a2. Similar to the previous comment, paragraph 3 and paragraph 4 in the introduction section should also be merged. It should focus on a brief overview of existing methods along with their limitations. 

33. The introduction section should also summarize the significance (in the light of achieved results) of the proposed work at the end of the Introductions Section.

44. The related work section should include another subsection named as "Novelty". This subsection should clearly explain various novel points of the proposed approach as compared to CNN Methods, RNN Methods, and traditional Machine Learning methods.

55. A comparative table (in the related work section) in terms of various attributes can be added to compare different state-of-the-art approaches and the proposed approach.

66. Figure 1 in the methodology section can be enriched by putting more details about the input and outputs of each block. Moreover, the methodology in Figure 1 is straightforward and represents a typical AI-based classification phenomenon. The novelty of the proposed approach is not evident from this figure.

77. The signal preprocessing step should provide the required mathematical formulations for this step.

98. The first paragraph in section 4 provides an overview of the entire section. It is recommended to cite each subsection with its number in this paragraph.

99. Appropriate references must be provided in Section 4.1.

110. The motivation for the selected database in Section 4.1 must be given.

111. The title of Section 4.2 should be improved. The authors do not present the network structure here. The authors are mainly evaluating the PA module and FusionGhost module. The presentation of results can be made more compact and illustrative through some figures.

112. Table 11 (Comparison of with state-of-art) has been provided. Nevertheless, it has never been discussed in the text. How different numbers in Table 11 prove the superiority of the proposed work.

113. The results in Table 11 show that the achieved results with the proposed method are almost similar to the results of RNN(LSTM, BLSTM, GRU, BiGRU).  Although authors have provided Table 12 (Comparative Analysis of RNNs and PANet Performance Across Different Frequency Bands and Noise Levels), however, the significance of the proposed work over the RNN is required to be explained in the context of Table 12.

 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The authors fully reply to my reviews, comments and suggestions.  i believe that the paper is acceptable for publication

Comments on the Quality of English Language

acceptable

Reviewer 2 Report

Comments and Suggestions for Authors

The authors have addressed the proposed comments and the manuscript has improved considerably with respect to the original version

Reviewer 3 Report

Comments and Suggestions for Authors

Authors have addressed all the raised concerns. 

The article can be published in its current form.