All reviews of published articles are made public. This includes manuscript files, peer review comments, author rebuttals and revised materials. Note: This was optional for articles submitted before 13 February 2023.
Peer reviewers are encouraged (but not required) to provide their names to the authors when submitting their peer review. If they agree to provide their name, then their personal profile page will reflect a public acknowledgment that they performed a review (even if the article is rejected). If the article is accepted, then reviewers who provided their name will be associated with the article itself.
The authors have addressed all the concerns.
[# PeerJ Staff Note - this decision was reviewed and approved by Massimiliano Fasi, a PeerJ Section Editor covering this Section #]
There are some remaining minor concerns that need to be addressed.
1) Comment from previous round: Specific problem statement of HOC method in artifact removal/peak detection
Authors response: Literature is updated and summarized in lines 98-112
Reviewer feedback: Comment is partially addressed. In the section describing the use of higher-order cumulants (HOC) for artifact removal, most cited works appear to focus primarily on component analysis or decomposition methods rather than directly employing HOC or combining methods with HOC, for artifact removal or peak detection. Perhaps the author can clarify the relevance of these references to the stated method, and, if applicable, elaborate on how the approaches in those papers are connected to HOC-based artifact removal in the context of your study.
2) Comment from previous round: Please include more recent references for related works on HOC for eye blink detection/removal.
Authors response: updated literature synthesis has been incorporated into the Introduction section (lines 98-106).
Reviewer feedback: The authors have included some recent studies; however, as noted in Comment #1 above, it would be helpful if they could provide a clearer explanation of how these references are directly relevant to HOC-based artifact removal in the context of this work.
3) Comments from previous round: Please clarify how the proposed approach differs from prior work that use HOC as mentioned in the literature review.
Authors’ response: In the Introduction section (lines 107-112), we have further clarified the distinctions between our proposed method and existing HOC-based approaches mentioned in the literature review
Reviewer feedback: The authors revision in lines 104 – 112 has adequately addressed the problem statement in HOC and the proposed approach.
1) Comment from previous round: Elaborate on the process of determination thr1 and thr2
Authors response: we have supplemented some details in the “Further adjustments to detected blink intervals” section (lines 194-204) and “Determination of Parameters” section (347-361) to elaborate on the two thresholds and their operational mechanisms in blink artifact boundary identification
Reviewer feedback: The additional details are adequate and have clarified the process of determining thr1 and thr2. This revision improves the transparency of the methodology
2) Comment from previous round: The novelty of the approach is not clear.
Author’s response: The authors add the explanation on novelty (lines 107-112) to provide a clear explanation
Reviewer response: Adequately addressed.
1) Comment from previous round: Please clarify how CC and RRMSE are computed for real EEG data, given that the true artifact-free EEG (ground truth) is not available
Authors response: The author have now supplemented the description of evaluation parameters applied to real data at lines 335-345
Reviewer feedback: The explanation of the MAE ratio has clarified the problem in the context of real data
2) Comment from previous round: a zoomed-in version of Figure 7B in the high-frequency region to better illustrate preservation of signal components
Authors response: an amplified analysis of the β-band power spectrum in Figure 7(B) were conducted
Reviewer feedback: The comment has been adequately addressed. As an additional suggestion, it would be helpful to include a zoomed-in view of the time-domain signal. In Figure 7(a), the proposed method illustrates how the eye-blink region is identified; a zoomed-in version around higher-frequency components near the eye-blink artifact would further demonstrate that the method successfully preserves the non-artifact signals.
3) Comment from previous round: Suggest to add discussion on the decrease performance (reduced CC and increase in RMSE) when SNR increase (Figure 6)
Author response: Analysis and discussion on the performance drop was added in lines 455-466
Reviewer feedback: Adequately addressed
The reviewers have substantial concerns about this manuscript. The authors should provide point-to-point responses to address all the concerns and provide a revised manuscript with the revised parts being marked in different color.
**Language Note:** PeerJ staff have identified that the English language needs to be improved. When you prepare your next revision, please either (i) have a colleague who is proficient in English and familiar with the subject matter review your manuscript, or (ii) contact a professional editing service to review your manuscript. PeerJ can provide language editing services - you can contact us at [email protected] for pricing (be sure to provide your manuscript number and title). – PeerJ Staff
no comment
no comment
no comment
As a pilot study, this article is good enough to propose novel eye blink detection method based on higher-order cumulants. The performance examination procedure is sufficient to represent the superior of this proposal. Some minor comments:
[1] The figure quality can be improved. The figure size is too small to be read.
[2] The math formula location are not consistent, some are in the center, some are on the left side.
[3] The impact on filtered EEG rhythm power is not shown. It may be the topic in the coming article, I can understand. Perhaps it can be mentioned in the discussion part.
Strengths
The manuscript is written in clear, professional English, making it accessible to an international audience.
The introduction provides sufficient context and motivation for the study, detailing the importance of artifact removal in EEG signal processing.
The literature review is comprehensive, covering relevant techniques such as VME and MSDW.
Figures and tables are well-labeled, clear, and provide sufficient information to support the claims.
Raw data has been provided, and appropriate references are included.
Areas for Improvement
Clarity in Explanation:
Some sections, particularly the mathematical derivations, could be made clearer with additional explanations or intuitive descriptions.
The rationale behind selecting specific parameters for the HOC method (e.g., threshold values) needs more elaboration.
Literature Coverage:
The manuscript focuses primarily on existing techniques like VME and MSDW but does not adequately compare with newer deep-learning-based methods for artifact removal.
Grammar & Typographical Errors:
Several minor grammatical errors and awkward phrasings need revision. For example:
Line 22: "However, the eye blink artifacts can contaminate the EEG signal, potentially affecting its clinical application." → "However, eye blink artifacts can contaminate EEG signals, potentially impacting their clinical utility."
Line 81: "This method is capable of adaptively identifying blink artifacts at different interval lengths." → "This method can adaptively identify blink artifacts of varying durations."
Strengths
The research question is well-defined and aims to improve blink artifact detection using higher-order cumulants.
The methodology is described in sufficient detail, enabling reproducibility.
The study uses both semi-simulated and real EEG data, ensuring robustness.
Areas for Improvement
Parameter Justification:
The selection of parameters such as HOC order, sliding window size, and error tolerance thresholds should be better justified.
The paper briefly mentions that ROC curves were used for parameter optimization (Figure 4), but it does not discuss why the chosen values are optimal for different types of EEG signals.
Validation Against More Methods:
The study compares the sliding window HOC method against VME and MSDW. However, modern deep-learning approaches for artifact detection (e.g., CNN-based or LSTM-based models) should be included in the comparison.
Statistical Significance of Results:
Although results are presented with statistical measures, p-values should be explicitly mentioned to confirm whether differences between methods are significant.
Strengths
The proposed sliding window HOC method demonstrates superior detection performance in both direct accuracy metrics (Youden index) and indirect artifact reduction effectiveness (CC, RRMSE, MAE).
Results from real datasets show strong performance in maintaining EEG signal integrity.
The study provides clear evidence that the proposed method improves boundary identification for blink artifacts.
Areas for Improvement
Discussion of Limitations:
The manuscript briefly acknowledges limitations but should expand on practical constraints, such as:
How this method would generalize across different EEG setups (e.g., different electrodes, varying noise conditions).
Whether it can handle simultaneous contamination from multiple artifacts (e.g., muscle artifacts, line noise).
Real-Time Applicability:
Since EEG artifact removal is often applied in real-time applications (e.g., brain-computer interfaces), the computational efficiency of the proposed method should be analyzed.
Benchmarking on Different EEG Tasks:
The effectiveness of blink artifact removal is well demonstrated, but would this method perform equally well in task-based EEG paradigms (e.g., cognitive load studies, event-related potentials)?
Figures and Tables:
Figures are generally well-presented, but some could be better explained in the captions (e.g., ROC curves in Figure 4 should explicitly state the tested parameter ranges).
Table 2 and Table 3 contain valuable comparisons but could be more compact by summarizing key takeaways.
Supplemental Code and Data:
The availability of supplemental code is appreciated, but it would be useful to provide example scripts for reproducibility.
1) Interesting topic and the presented writing is adequate.
2) Perhaps, the author can add more specific problem statement on HOC method for artifact removal and elaborate on how the proposed method fill in the gap.
3) Please include more recent references for related works on HOC for eye blink detection/removal. Please revise the literature review to better explain the state of the art instead of just listing out relevant works. Please clarify how the proposed approach differs from prior work that use HOC as mentioned in the literature review.
1) The proposed method utilizes thr1 and thr2 as predetermined threshold values. Please elaborate on how these values are selected and justified. A more detailed explanation of the threshold determination process, including whether they are based on empirical analysis, statistical criteria, or optimization techniques, would enhance clarity.
2) While the proposed methods align with the problem statement, the novelty of the approach is not clearly evident. Although the methodology is well explained, further clarification on what differentiates this work from existing studies would strengthen its contribution. Highlighting key innovations, improvements, or unique aspects compared to prior research would be beneficial.
1) The author should clarify how Correlation Coefficient (CC) and Relative Root Mean Square Error (RRMSE) are computed for real EEG data, given that the true artifact-free EEG (ground truth) is not available. A detailed explanation of the reference signal or estimation method used for evaluation would improve clarity.
2) It is recommended to include a zoomed-in version of Figure 7B in the high-frequency region to better illustrate how proposed HOC preserves signal components compared to other methods. This will provide clearer visual evidence of its effectiveness.
3) Suggest to add discussion on the decrease performance (reduced CC and increase in RMSE) when SNR increase (Figure 6). Explain the factors contributing to this reduction, or limitations of the proposed method in handling high/low-SNR data
Clear and unambiguous, professional English used throughout. However a more precise explanation of higher order cumulant is required in the manuscript.
In introduction, sufficient explanation is expected to understand the importance of single channel EEG analysis with special interest on frontal lobe.
In introduction, references should be given along with the specific biomedical applications where HOC has been used in para 3, line no. 82-85.
A flow diagram may be helpful.
Professional article structure, figures, tables. Raw data shared.
Results supports the hypothesis and well defined.
The experimental design of the manuscript found to be original.
All underlying data have been provided; they are robust, statistically sound, & controlled.
Conclusions are well stated and linked to original research questions. Limitations are also discussed elaborately in the discussion section
All text and materials provided via this peer-review history page are made available under a Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.