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, we are pleased to verify that you meet the reviewer's valuable feedback to improve your research.
Thank you for considering PeerJ Computer Science and submitting your work.
Kind regards
PCoelho
[# PeerJ Staff Note - this decision was reviewed and approved by Xiangjie Kong, a PeerJ Section Editor covering this Section #]
The manuscript is clearly written in professional English and follows a coherent academic structure. The introduction provides sufficient background on the MVQA problem, relevant datasets, and the motivation for applying mixup and label smoothing. The literature review is comprehensive and well-connected to the research objectives. Figures, tables, and algorithms are well-labeled and contribute effectively to the narrative. Raw data and implementation details are properly shared through Zenodo, ensuring transparency and reproducibility
Improving medical visual question answering through data augmentation and explainability, is well-defined, relevant, and addresses an identified gap in dataset imbalance and interpretability.
The methodology is rigorous and well-documented, including dataset normalization, augmentation algorithms, model architectures (VGGNet-LSTM and ResNet-LSTM), and quantitative metrics (accuracy and BLEU). The inclusion of statistical significance testing and confidence intervals strengthens the robustness of the analysis. Ethical considerations for medical data and clinical validation steps are explicitly stated.
Performance improvements across all MVQA datasets are clearly demonstrated and contextualized against state-of-the-art baselines. The use of LRP XAI for visualization adds interpretability and reinforces the validity of the findings. Conclusions are logically connected to the research question and are appropriately scoped to the reported results.
Dear authors,
Thanks a lot for your efforts to improve the manuscript.
Nevertheless, some concerns are still remaining that need to be addressed.
Like before, you are advised to critically respond to the remaining comments point by point when preparing a new version of the manuscript and while preparing for the rebuttal letter.
Kind regards,
PCoelho
The manuscript is well written and presented in a clear and logical manner, with professional use of English throughout. It incorporates recent literature (2022- 2025), which provides sufficient field context and situates the study within current research. The inclusion of figures, tables, and appendices such as abbreviations and symbols improves readability, while the availability of raw data and implementation details on GitHub strengthens reproducibility. A few minor improvements would further enhance the manuscript. figure captions, particularly for Figures 4 to 6, could be expanded to better explain the insights conveyed and guide the reader more effectively.
The research question is clearly defined, focusing on the challenges of dataset scarcity and imbalance in MVQA, and it offers a meaningful solution using Mixup and Label Smoothing, combined with LRP for explainability. The investigation is carried out rigorously, with strong attention to ethical considerations, datasets are sourced from de-identified repositories, and the augmented samples have been reviewed by a certified pathologist to ensure clinical validity. The methods, algorithms, and equations are described in detail, and the availability of code and data makes the work reproducible. The description of hyperparameters, computational setup, and reproducibility steps further strengthens the credibility of the study. Algorithm 1 and Functions 1- 2 add more descriptive explanations alongside the mathematical details.
The comparisons between original and augmented datasets are thorough, and the inclusion of statistical tests such as confidence intervals and p-values strengthens the credibility of the results. The use of LRP for visualization adds clarity by showing how the model makes its predictions, effectively linking the outcomes back to the research question.
Dear authors,
You are advised to critically respond to all comments point by point when preparing an updated version of the manuscript and while preparing for the rebuttal letter. Please address all comments/suggestions provided by reviewers, considering that these should be added to the new version of the manuscript.
Kind regards,
PCoelho
**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
Clear and unambiguous, professional English used throughout:
The manuscript is written in professional and technical English; however, there are some grammatical inconsistencies and awkward phrasing that may impact readability.
Certain sentences, particularly in the methodology and results sections, could be refined for better clarity and fluency. A thorough proofreading or professional editing would be beneficial.
Literature references, sufficient field background/context provided:
The paper provides a well-rounded literature review, citing relevant studies on dataset augmentation, MVQA models, and Explainable AI (XAI) techniques.
Some recent advancements in XAI, particularly beyond Layer-wise Relevance Propagation (LRP), could be referenced for completeness.
Professional article structure, figures, tables, and raw data shared:
The manuscript follows a professional and logical structure, with a clear introduction, methodology, results, and discussion.
Figures and tables are well-structured and provide meaningful insights. However, some tables (e.g., Table 1) could be reformatted for improved readability.
The authors provide a GitHub link for dataset access, but it should be explicitly stated whether all raw data and model implementation details are publicly available for full reproducibility.
Self-contained with relevant results to hypotheses:
The manuscript is self-contained, addressing its research question effectively.
The proposed dataset augmentation methods and MVQA model improvements are directly linked to the hypotheses stated in the introduction.
Formal results should include clear definitions of all terms and theorems, and detailed proofs:
The manuscript presents a solid mathematical foundation, defining key functions and equations used in the augmentation and validation processes.
Some derivations and notations (e.g., Equations 1–25) could be better explained for clarity, especially for readers unfamiliar with dataset augmentation techniques.
Original primary research within the Aims and Scope of the journal:
The study presents original research focused on dataset augmentation and explainability in Medical Visual Question Answering (MVQA) systems.
The topic aligns well with the scope of the journal, as it contributes to advancements in AI-based medical image analysis and interoperability.
Research question well defined, relevant & meaningful. It is stated how research fills an identified knowledge gap:
The research question is clearly defined: how to enhance MVQA model performance using dataset augmentation (Mixup and Label Smoothing) and Explainable AI (Layer-wise Relevance Propagation - LRP XAI).
The study addresses the challenge of imbalanced and limited medical QA datasets, which is a critical issue in the development of robust MVQA models.
The work improves upon previous methods by applying systematic dataset augmentation techniques and validating results using explainability methods, filling a clear gap in medical AI model robustness and interoperability.
Rigorous investigation performed to a high technical & ethical standard:
The paper employs mathematical formulations, statistical validation (accuracy, BLEU score), and qualitative explainability analysis to ensure a rigorous investigation.
Ethical considerations for medical data handling are briefly mentioned, but could be expanded, especially regarding privacy and bias in dataset augmentation.
Methods described with sufficient detail & information to replicate:
The methodology is detailed and well-structured, covering dataset augmentation, model training, and explainability analysis.
Mathematical formulations (Equations 1–25) are provided for dataset transformations and performance evaluation. However, some equations could be explained more clearly for better readability.
Implementation details (e.g., hyperparameters, training setup, dataset preprocessing) are included, but a more explicit description of computational resources and reproducibility steps would strengthen the study’s replicability.
Impact and novelty not assessed. Meaningful replication encouraged where rationale & benefit to literature is clearly stated:
The manuscript does not explicitly assess the impact or novelty of the study, but provides a clear rationale for dataset augmentation and explainability in MVQA.
The authors emphasize the importance of their approach by comparing augmented datasets with original datasets and showing measurable performance improvements.
The study’s methodology and open-source dataset availability (via GitHub) suggest that meaningful replication is possible, but a more explicit discussion on how other researchers can apply these techniques to different datasets would strengthen this aspect.
All underlying data have been provided; they are robust, statistically sound, & controlled:
The manuscript claims that the enhanced MVQA datasets and results are available; however, it is important to verify if all raw data, preprocessing scripts, and model checkpoints are accessible for full transparency.
Statistical analysis is sound, using accuracy and BLEU score to compare performance between different models and datasets.
The experiments are controlled, with clear comparisons between original and augmented datasets. However, reporting additional statistical significance tests (e.g., confidence intervals, p-values) would further support the robustness of the findings.
Conclusions are well stated, linked to the original research question & limited to supporting results:
The conclusions effectively summarize the improvements observed in MVQA performance due to dataset augmentation and explainability methods.
The findings are directly tied to the original research question and do not overextend beyond the presented results.
The study acknowledges some limitations, particularly in dataset challenges and model biases, but a more detailed discussion on potential weaknesses and future improvements would be beneficial.
Areas for Improvement:
Refinement of language and writing clarity.
More explicit discussion of statistical significance.
Verification of dataset and code availability for full reproducibility.
- The quality of figures (e.g., Figure 1) and caption descriptions needs improvement, as they lack sufficient detail.
- Algorithm 1 should be converted into a text algorithm, not an image. The same for others such as Function 1.
- The equations should be described well.
- Include a table of abbreviations in the revised manuscript to aid reader comprehension.
- Include a table of symbols in the revised manuscript for clarity, especially if mathematical notations are used.
- Several sentences lack citations. Address this by adding appropriate references where necessary.
- The manuscript should be reviewed to correct any typographical errors.
- Incorporate recent citations from 2022 to 2025 to ensure the manuscript reflects the latest advancements in the field.
- What is the novelty of the proposed approach? Highlight how the approach advances the state-of-the-art.
- Format Reference 34 according to the journal guidelines.
- Clearly identify the research question and gap in the manuscript to ensure the study's purpose is well-articulated.
- Include a clear justification for the research to strengthen the rationale behind the study.
- Discuss the suggested approach more scientifically, providing additional technical and methodological details.
- What distinguishes the current study from related research? This should be explicitly stated to emphasize the study's uniqueness.
- Why are the sizes of heatmaps in Figure 4 larger in height compared to the images?
-
-
-
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.