An ECG Signal De-Noising Approach Based on Wavelet Energy and Sub-Band Smoothing Filter
, Shanshan Wang 1,†, Feng Li 1, Jin Wang 1,*, Arun Kumar Sangaiah 3, Victor S. Sheng 4 and Xiangling Ding 5Round 1
Reviewer 1 Report
The authors claim to have developed a novel method for ECG de-noising based on wavelet energy and sub-band smoothing filters. While the design of the method is extensively described in details, the experimental setup is very poorly designed. Namely, even in the beginning, when selecting the wavelet basis, they have performed the analysis only on one record from the MIT-BIH database. Furthermore, while they compare the proposed technique with many other techniques for ECG de-noising, the comparison is again performed in several occasions only on small portion of the MIT-BIH database (only some records, which can be assumed that are pre-selected) or even only on one record in most cases. This represents a serious drawback of the experimental setup, and consequently the results and their interpretation. Therefore, I recommend major revision of the experimental evaluation.Author Response
Please see the attachment
Author Response File: 
Author Response.pdf
Reviewer 2 Report
Zhang et al. address the important and challenging task of ECG denoising by proposing an approach based on wavelet decomposition and sub-band smoothing. The proposed approach achieves a reduction in computational complexity by applying threshold processing only to a previously identified subsets of wavelet coefficients instead of all coefficients. The authors assess the performance of the proposed procedure on both real world and synthetic ECG signals contaminated by various noise sources and present promising results.
Specific Comments:
1. While being well-structured, the submitted manuscript in this reviewer's opinion is often poorly phrased. Authors should be encouraged to revise the manuscript in order to improve readability.
2. Section 3, which describes the proposed approach, needs to be revised; e.g. the authors make use of terms which have not previously been appropriately defined (for example, a clear definition of "tolerance value" and "change point" is missing). Also, the choice of the smoothing filter in Subsection 3.5 is not justified and appears arbitrary.
3. Similarly, in this reviewer's opinion, the authors should provide a clear motivation for the particular choice of signal excerpts out of the 48 signal traces in the MIT-BIH arrhythmia database (the authors allege to have selected 119 records but provide no further details). The same arguments holds for the results presented in Section 4. Crucially, specific excerpts of records have been selected for the provided figures while the tables appear to be based on yet another small subset of the MIT-BIH arrhythmia database. Accordingly, in this reviewer's opinion, due to the seemingly arbitrary record selection procedure the presented results do not sufficiently corroborate the conclusions as stated by the authors.
Minor Comments:
4. This reviewer respectfully submits that the qualitative denoising assessment by means of visual inspection is not appropriate as it has little to no evidentiary value.
5. Please replace "ectopic pulsations" with "ectopic complexes" (and provide a single sentence explanation of what they are). Also, please note that one assesses an ECG which then may be intergrated in the physician's decision making procress of diagnosis; there is, however, no such thing as a "diagnosis of ECG signals".
6. On lines 17-18, please strike/rephrase "electromyographically interference from organisms".
7. Singular Spectrum Analysis (SSA) has also been successfully used in ECG denoising. If possible, please add this to your related works/prior art discussion in Section 1.
Author Response
Please see the attachment
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
The reviewer still remarks that language and style spell check are mandatory before publication.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 2 Report
This reviewer would like to thank the authors for considering the comments raised by this and the other reviewer in round 1 and resubmitting a revised manuscript.
While the submitted manuscript has certainly been improved, this reviewer respectfully submits that author's statement - pertaining to why specific records and not others or all records were considered - remains ambiguous and hence unscientific.
This reviewer would strongly urge the authors to reconduct parts of their experiments, this time making use of the entire MIT-BIH arrhythmia database.
In the alternative, the authors should at the very least include - if not all - than additionally at least records 106, 113, 117, 119, 122, 200, 230 as is done in the 4 papers the brought to this reviewers attention pertaining the issue of qualitative denoising assessment.
Respectfully submitted
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
