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.
Dear Authors,
Thank you for addressing the last reviewer's minor comments.
Best wishes,
-
-
-
-
Dear Authors,
Thank you for the revised paper. Although one reviewer accepts your paper, one reviewer suggests minor revision. We encourage you to address the Reviewer2' suggestion on the number of independent runs and resubmit your paper.
Best wishes,
no comment
no comment
no comment
no comment
No comment
You say that "The BPSO algorithm was evaluated over 5 independent runs." You need to run it at least 20 times.
No comment
No comment
Dear Authors,
Your manuscript is not suggested for publication in its current form. The reviewers think that your paper requires major revision. We encourage you to address the concerns and criticisms of reviewers and resubmit your paper once you have updated it accordingly.
Best wishes,
**PeerJ Staff Note:** Please ensure that all review, editorial, and staff comments are addressed in a response letter and that any edits or clarifications mentioned in the letter are also inserted into the revised manuscript where appropriate.
you can find detailed review of this manuscript as in the Additional comments section
you can find detailed review of this manuscript as in the Additional comments section
you can find detailed review of this manuscript as in the Additional comments section
This review critically evaluates the article titled “Enhanced Swarm Optimization for Feature Selection in EEG Classification: Investigating Visibility Graph and Persistent Homology-Based Features.” The study focuses on classifying non-medical and medical datasets EEG recorded under different auditory conditions, aiming to contribute to the field by extracting topological and network-based features and integrating them with an improved binary particle swarm optimization (BPSO) method for optimal feature selection. However, in its current form, the manuscript presents several drawbacks that require attention. This review assesses the novelty, technical soundness, contribution to the field, and overall relevance of the work. The suggested revisions outlined below are expected to enhance the clarity, rigor, and impact of the article.
1. The abstract does not sufficiently reflect all contributions (e.g., dataset details, novelty of enhanced BPSO, multimodal EEG/sEMG potential). It should highlight both methodological innovations and comparative performance improvements more explicitly..
2. The related work section is underdeveloped. More recent state-of-the-art studies should be cited, especially on PSO enhancements, visibility graphs, and persistent homology in EEG. A comparative positioning would strengthen the gap of the field.
3. The “Discussions" section should be added in a more highlighting, argumentative way. The evaluation results should be described in more detail including the discussion about algorithm complexity. For example, Expand with deeper arguments: why does alpha outperform beta? Why does WVG dominate? Discuss algorithmic complexity and interpretability.
4. The limitations of the algorithm and the evaluation environment should be discussed in the paper. What are the capabilities, benefits and limitations of study? The discussion section could benefit from an exploration of potential limitations, such as computational demands or limitations in real-time processing. Suggestions for future research, like optimizing the model for faster inference or exploring its application to other types of medical imaging, would add depth to the study.
5. Discuss dataset size (45 participants), generalization limits, computational costs of BPSO, and real-time feasibility. Suggest future work on faster inference, multimodal fusion, or hardware optimization
6. The complexity of the proposed model and the model parameter uncertainty are not enough mentioned. Provide parameter counts, hyperparameter sensitivity analysis, and explain trade-offs.
7. Incorporate visual explanations (e.g., Grad-CAM or saliency maps) to illustrate how each module contributes to the segmentation. Consider Grad-CAM, saliency maps, or node-importance visualizations on visibility graphs to clarify feature contributions.
8. The authors should consider discussing the computational efficiency of the model, including training time and inference time. Training/inference time, hardware used, and scalability should be explicitly reported.
9. The author should analysis the reason why the tested results are achieved. The reasons behind obtained performance (e.g., PH feature richness vs. persistence statistics redundancy) should be interpreted more explicitly.
10. The Conclusions should be reviewed again. The original aspect of the study and its difference from other studies should be clearly explained (A more detailed explanation should be given for the conclusion part.) This section needs strengthening. Clarify originality (combining PH + VG + enhanced BPSO for non-medical EEG) and differences from prior methods.
11. How did you set the parameters of proposed method for better performance? Justify choice of swarm size, iterations, and inertia weights. Did you tune hyperparameters empirically or via validation?
12. Only accuracy/F1 reported in some sections. Add per-class precision, recall, F1, confusion matrices, and PR-AUC.
13. Contributions should be summarized clearly in Introduction and Conclusion (e.g., methodological, dataset analysis, comparative validation).
14. Provide more details on subject demographics, recording conditions, preprocessing steps, and artifact removal criteria. For Bonn dataset, justify preprocessing choices (e.g., trimming to 5.9s).
15. Clarify the internal architecture of the enhanced BPSO: velocity update equations, swarm division, mutation strategies. Present hyperparameters systematically in a table.
Thre is no problem with the English of the paper.
The introduction section summarizes the literature, but some recent work (e.g., deep learning-based EEG classification papers from the last 1–2 years) could be further cited.
Structure of the article, figures and tabels are proper. Raw data has been shared.
Yes. The study was designed and reported comprehensively. The research question was to examine the effectiveness of PH, VG, and iBPSO methods in classifying non-medical EEG. The experiments aimed to test this hypothesis. The results are directly related to this hypothesis; in particular, the PH+VG features and the improved BPSO were shown to significantly increase accuracy and F1 scores.
all key concepts and methods are clearly defined, with equations and algorithmic details provided to ensure clarity and reproducibility.
Yes. This manuscript presents original primary research that fits well within the aims and scope of PeerJ Computer Science.
The manuscript clearly explains how the proposed approach contributes to filling this gap.
The study has been conducted to a high technical standard, with detailed descriptions of data preprocessing, feature extraction, and classification procedures. Multiple experiments, statistical validation, and comparisons with baseline methods support the rigor of the investigation. Ethical standards were also appropriately followed, with prior approval obtained for the non-medical EEG dataset and informed consent from participants.
Methods are described with sufficient information.
The study includes meaningful replication by testing the proposed methods on both non-medical EEG data and the widely used Bonn dataset. This dual evaluation provides clear rationale and demonstrates the benefit to the literature by validating the approach across different EEG contexts.
The underlying data, including both the non-medical EEG dataset and the Bonn benchmark dataset, have been clearly described and made available.
The conclusions are clearly stated and directly linked to the original research question regarding the effectiveness of combining PH, VG, and enhanced BPSO for EEG classification. The claims are appropriately limited to the supporting results, highlighting the observed performance improvements and the relative contribution of different feature types without overstating the findings.
1) Strengths:
Methodological novelty: combination of PH + VG and BPSO improvements.
Validation on both medical and non-medical EEG data.
Comprehensive experimental results supported by statistical tests.
2) Weaknesses:
Dataset descriptions should be provided in more detail.
More solid justification is needed for parameter selections.
A deeper discussion of computational cost should be included.
Literature coverage is insufficient.
3) Comments:
The Introduction section should explicitly summarize the study’s findings.
While the Introduction summarizes the literature, it could benefit from additional references to more recent works (e.g., deep learning–based EEG classification papers from the last 1–2 years).
Although the Discussion section is strong, the computational cost aspect is treated somewhat superficially; a more detailed consideration of the high cost of PH and VG methods in terms of practical usability would be valuable.
Parameter choices (e.g., swarm size, number of iterations) are reported in the experimental section, but providing clearer justifications would strengthen the methodology.
It should be specified how many independent runs were performed for BPSO.
1. Clearly state the research gap in the introduction and how this study addresses it explicitly, mentioning the work advances highlighting the novelties with respect to previous studies.
The knowledge gap being investigated should be identified, and statements should be made as to how the study contributes to filling that gap.
2.Provide the relevance and challenges of Non-medical EEG analysis focusing on how they impact or enrich the significance of chosen problem when compared with traditional medical EEG studies.
3. The manuscript would benefit from overall language polishing to improve readability and precision. (e.g lines 43-46)
1. A clear step-by-step explanation for each enhancement and the parameters set in the swarm optimization is needed
2. A more structured explanation of the feature extraction and selection stages is expected
3. Methods should be described with sufficient information to be reproducible by another investigator.
1.The clarity of results could be improved by refining figure legends.
2.The conclusions should be appropriately stated, should be connected to the original question investigated, and should be limited to those supported by the results. In particular, claims of a causative relationship should be supported by a well-controlled experimental intervention
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.