Arrhythmia Classification and Diagnosis Based on ECG Signal: A Multi-Domain Collaborative Analysis and Decision Approach

Round 1

Reviewer 1 Report

     The authors developed a GNN model, a multi-domain collaborative analysis, to make the classification and diagnosis of arrhythmia more accurate. The ECG signals can be more clearly detected by convolution and sequential learning modules. The authors conduct extensive experiments on eight advanced models in the same field to demonstrate the effectiveness of the proposed method. However, I have several concerns.

1.     The grammar has some mistakes. I suggested revision native speakers.

2.     In Figure 2, there was not much difference between the raw and de-noised signals.

3.     In Figure 3, there was no detailed description for (a) to (d).

4.     In this study, the BIDMC Congestive Heart Failure Database was applied for data training and testing. I suggested applying the GNN model compared to other methods in other independent test sets to prove the robustness.

5.     What are the analyzed leads? Single or multi-leads? The composition and data quality was not clearly stated in the method section.

 

6.     Could the authors apply the GNN model in real-world ECG validation and provide AUC, sensitivity, specificity, negative predictive value, and positive predictive values rather than MAE, RMSE, or MAPE? It is difficult to assess the clinical value.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

I have one recommendation for the authors. The authors may provide the rationale behind the selection of the hyperparameters of the training. Will the change of hyperparameters produce better results?

Although this is a public dataset, a thorough description of the dataset will help readers understand the methodology. For example, what does it mean by "the training set, test set, and verification set are set up in a ratio of 7:2:1" ?

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

Manuscript entitled “Arrhythmia Classification and Diagnosis Based on ECG Signal: A Multi-Domain Collaborative Analysis and Decision Approach” gives an interesting and systematic research about proposal a novel and different approach in the classification and diagnosis of arrhythmia. It is based on a multi-domain collaborative analysis and decision approach, which make the classification and diagnosis of arrhythmia more accurate.

The topic is relevant in the field, because the rate of cardiovascular diseases is increasing round the world and, according to the authors, the prevalence rate of arrhythmia among elderly in China is as high as 2.4%.

The manuscript is well organized, authors used the scientific methods for the conducted analysis and they were adequately described. In the Introduction part the aim of the manuscript is described and in the Related work part, authors define the state and theme of the related researches. In the next part, they dive a detail classification of the application particulars of graph neural network in arrhythmia. The results of the study are clearly presented and the discussions support presented results. Results are proved at the end of the manuscript by presentation of the fitting results of the used graph neural network in comparison to other methods. Authors showed that proposed approach has improved the accuracy of prediction and classification of the sequence, and gave a foundation for online real-time monitoring of ECG to realize classification and diagnosis of arrhythmia.

The references cited in the manuscript are recent, mostly within the last 10 years. The figures, tables, images, and schematics are presented appropriately and clearly. Data presented in charts are properly presented and easy to interpret and understand.

The only suggestion for the authors is that authors change the index before the Introduction to 1 (instead of 0) and that in the chapter Related Work (line 65) they correct the first sentence. Is it in the sentence : “For more than ten years, scholars at home and abroad have proposed lots of classification algorithms for arrhythmia.” Authors should specify what "home" means.

The acceptance of the manuscript is suggested with the above mention corrections.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

The authors have adequately responded to the questions. However, I still have several suggestions.

1. Could the authors provide an English editing certificate?

2. The authors have changed Figure 2 for a better illustration of denoising signals. However, the baseline information has been altered after denoising, such as sinus rhythm with a clear P wave in the upper panel of Figure 2 but the absence of a P wave in the second beat in the lower panel. How will the authors explain this?

 

3. Thank you for providing the basic information of the four columns in Figure 3. It is better to provide the details of the lead information, such as blue one stands for lead V1, a green one for V5, and not just all in sinus rhythm but some examples for non-ectopic (N), ectopic (S), ventricular ectopic (V), fusion (F) and unclassified heartbeat (Q). 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

My previous comments were addressed

Author Response

Thank you for your review and suggestions.  We are honored that our reply can be approved by you.