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.
Reviewers are happy with the revisions and I concur to recommend accepting the current version.
[# PeerJ Staff Note - this decision was reviewed and approved by Jyotismita Chaki, a PeerJ Section Editor covering this Section #]
Compared to the original manuscript, the revised version has made significant improvements in English. The author has made significant improvements to the introduction section which clearly demonstrates the objective and specific contribution of this paper. In the revised manuscript, the author has supplemented the the number of the original cases/images and the CT and MRI data used in the proposed method.
In the revised manuscript, the author provided reasonable explanations and improvements to the experimental design.
Conclusions are well stated, linked to original research question & limited to supporting results.
In summary, the author has fully responded to the review comments and made significant modifications to the original manuscript
The revised manuscript used clear and professional English. And it satisfies the publish criteria and answers my question.
The revised manuscript has a good experimental design.
The revised manuscript has a good validity of the findings.
I agree to accept this manuscript.
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 tracked in a different color.
**PeerJ Staff Note:** Please ensure that all review and editorial comments are addressed in a response letter and any edits or clarifications mentioned in the letter are also inserted into the revised manuscript where appropriate.
**Language Note:** The review process has identified that the English language must be improved. PeerJ can provide language editing services - please contact us at copyediting@peerj.com for pricing (be sure to provide your manuscript number and title). Alternatively, you should make your own arrangements to improve the language quality and provide details in your response letter. – PeerJ Staff
(1)The English used in this paper needs significant improvement. I suggest the manuscript be proof-read by an expert in the field of deep learning to check grammar and improve readability.
(2)The Introduction does not show context well. The objective and specific contribution of this paper are not clearly described.
(3)The raw data is not clearly introduced. For example, what’s the number of the original cases/images before augmentation? How are the CT and MRI data used in the proposed method?
(1)The major problem with experiments is the lack of comparison to baseline methods. As a paper proposing a new algorithm/classification-framework, it is essential to show that the proposed method is superior to existing methods.
(2)Ablation study is also needed to show the importance of key components of the proposed framework and the rationality of major design choice.
(3)Experiments on more datasets are needed and a dataset is usually divided into train/validation/test sets instead of only train-test sets.
The proposed method lacks of technical novelty because it is only an integration of several existing widely-used methods. The experimental design has some flaws, such as the lack of a validation set, and the finding is not very sound and robust.
This paper proposes a framework to detect and classify retinoblastoma. From a technical point of view, the proposed framework is a simple combination of existing algorithms and lacks of novelty. The experimental results are not sufficient to support the conclusion because of the flaws in experimental design, the lack of comparison to existing methods and the limited datasets used. The English has big room of improvement.
Placement of Data Augmentation in Pre-processing: Regarding the organization of the paper's content, I propose relocating the discussion of data augmentation techniques to the pre-processing section. This adjustment would align the paper's structure with common practices in presenting the sequence of data manipulation steps. By addressing data augmentation alongside other pre-processing methodologies, readers can gain a holistic understanding of the entire preparatory pipeline before model implementation.
1. Ablation Experiment for Feature Fusion Clarification: To further reinforce the significance of the fusion between hand-crafted and deep-learning features, I recommend the authors consider conducting an ablation experiment. By systematically removing one feature type at a time and assessing the impact on performance metrics, the study can provide a more robust justification for the integration of these feature categories. This analysis would enrich the discussion around the interplay between hand-crafted and deep-learning features, contributing to a deeper understanding of their individual and combined contributions to the model's success.
2. Elaboration on Dataset Splitting and Training Procedure: It would be valuable for readers to gain a comprehensive understanding of the paper's experimental setup. Could the authors elaborate on the methodology employed for splitting the dataset into training, validation, and test subsets? Additionally, delving further into the training process—such as the optimization algorithm used, learning rate schedule, and convergence criteria—would provide a clearer picture of how the model was fine-tuned. Providing this level of detail would enhance the reproducibility and rigor of the study.
3. Thorough Analysis of Results, Comparisons, and Overfitting Concerns: The reported 100% accuracy is indeed remarkable, prompting the need for a more in-depth discussion surrounding this achievement. To contextualize the findings, I suggest the authors undertake a thorough comparative analysis with similar research in the field. Comparisons with state-of-the-art models or benchmarks would not only demonstrate the paper's contributions but also shed light on its performance relative to existing solutions. Furthermore, addressing potential overfitting concerns and detailing how the model's generalization capabilities were ensured during training would augment the validity of the outcomes.
Open-sourcing of Dataset and Code: In the spirit of promoting transparency and facilitating further research in the domain, I encourage the authors to consider open-sourcing both their dataset and code. By making these resources publicly available, the wider research community can verify and build upon the work, fostering collaboration and accelerating advancements in the field. This step would contribute to the paper's impact and assist in replicating the experiments.
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.