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.
Thank you for your contribution to our journal.
[# PeerJ Staff Note - this decision was reviewed and approved by Shawn Gomez, a PeerJ Section Editor covering this Section #]
nothing to report
nothing to report
nothing to report
nothing to report
Please address all requests and comments of the reviewers thoroughly.
Good
Good
Good
The reviewer's comments were incorporated nicely
I believe the authors did a good job in addressing the comments by the reviewers. However the file I was sent for review didn't contain any figure or table. If this is part of the authors' idea to streamline the paper, it is too much.
The authors addressed the experimental issues.
The authors addressed the experimental issues.
I believe the authors did a good job in addressing the comments by the reviewers. However the file I was sent for review didn't contain any figure or table. If this is part of the authors' idea to streamline the paper, it is too much.
Related work and comparison with other published materials are considered the backbone of the paper. Through related work, readers get insight into the authors’ work and other closely related materials.
I suggest adding a related work section as a separate section consisting of two or three pages to make the paper more valuable.
Experiments are sound
Findings are adequate
Additionally, please resubmit the cover letter with line numbers and page numbers so I can see exactly where your previous comments have been addressed.
Submit the same cover letter with previous comments, along with line and page numbers
Also, add a related work section.
The proposed hybrid CNN-ViT architecture is used to detect leukemia in white blood cells. The hybrid architecture is designed to leverage the spatial localization capabilities of the CNN and the attention module in ViT, while also being more lightweight than current approaches in the literature. Please highlight the contribution and follow the reviewers' requests.
**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.
-
-
-
1. Clear and unambiguous, professional English used throughout
2. Good introduction with recent related works. The authors are appreciated.
3. The abstract should be updated with pertinent statistics in addition to the number of evaluation parameters and prediction accuracy.
4. Each block, which was previously created and made public, is described in great depth. Provide as much detail as possible using pertinent mathematics and references.
5. Transformers are capable of anticipating everything. There is no rationale for combining CNN and Transformer, save for accuracy. In fact, it makes the model more complex and raises the system's computing cost.
6 There are no sample images given from each data set for reference
7. Data imbalance is there, which will lead to improper results
8. There is a mismatch in the number of input images and the confusion matrix values of images
9. The results are overfitting. We cannot come to a conclusion by running just 30 epochs.
10. Why do the authors use six effective transformers when they only need one to achieve 100% accuracy?
11. Recall and other performance metrics are not strictly mentioned as REC …
The authors propose a hybrid CNN-ViT architecture for the purpose of detecting leukemia in white blood cells. The hybrid architecture is designed to take advantage of the spatial localization abilities of the CNN and the attention module in ViT, at the same time being more lightweight than current approaches in the literature.
I believe one of the major flaws of the paper lies in the unclear definition of the novelty: it is unclear whether the efficient modules described in the paper (efficient multi-head self-attention, Efficient MLP Block) are novel or are taken from previous works.
The second major flaw lies in the experimental evaluation: the models chosen as a comparison in Table 4 do not reflect the state of the art for leukemia detection. I would suggest that the authors compare their approach with more recent methods, designed for leukemia detection. The authors did a partial evaluation in Table 14, though only on D2 and without the most recent approaches (2024-2025). E.g.,
A. Genovese, V. Piuri, and F. Scotti, "A Decision Support System for Acute Lymphoblastic Leukemia Detection based on Explainable Artificial Intelligence", in Image and Vision Computing, vol. 151, no. 105298, November, 2024. ISSN: 0262-8856.
A. Genovese, V. Piuri, K. N. Plataniotis, and F. Scotti, "DL4ALL: Multi-task cross-dataset transfer learning for Acute Lymphoblastic Leukemia detection", in IEEE Access, vol. 11, 2023, pp. 65222-65237. ISSN: 2169-3536.
**PeerJ Staff Note:** It is PeerJ policy that additional references suggested during the peer-review process should only be included if the authors are in agreement that they are relevant and useful.
Also, a cross-fold validation is necessary to produce more meaningful results.
It would also be interesting to see results on the C-NMC database and on ALL-IDB patches: https://ieeexplore.ieee.org/document/10193429
-
The novelty must be more clearly defined. At the current stage, it might appear as a mere application of an existing CNN-ViT hybrid method. Moreover, such a hybrid architecture has to be put in context with other hybrid-based architectures proposed in the literature.
I believe the experimental part could be shortened by removing the loss plots and the confusion matrices, and showing the most significant results (e.g., the comparison between different variants and the comparison with the literature).
The paper presents a promising model, and it's a well-written paper.
The experimental design needs to be compared with other models, which is why it has not achieved the same accuracy as the proposed model.
The validity of the results is very accurate; therefore, it requires strong justification.
The paper presents a hybrid CNN-transformer model that obtained superior results on two publicly available datasets.
Most of the comments have been addressed, but the related work section is still missing. The author needs to add related work after the introduction and compare the proposed hybrid CNN-transformer model with other published CNN-transformer models. Also, they need to justify why it only applies to two datasets.
Please cite the equation in the text.
The requirements of algorithm one need to be explained with an example, with the objective function and the acquisition function.
It is very impressive that the proposed model achieved 100% accuracy, but the author may justify why other models, like VGG19, have less accuracy.
The discussion part can be added with work limitations.
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.