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.
According to reviewers' comments and your reply, after comprehensive consideration, it is decided to accept your manuscript.
[# PeerJ Staff Note - this decision was reviewed and approved by Carlos Fernandez-Lozano, a PeerJ Section Editor covering this Section #]
I believe the manuscript meets the journal’s publication standards and recommend acceptance.
N/A
N/A
Please revise the paper according to the reviewer's comments.
The authors have proactively addressed the initial review comments by supplementing the dataset description and expanding related work. However, the following recommendations are proposed for further enhancement:
(1) Incorporate a comparative analysis with general segmentation models to highlight the methodological advantages in medical scenarios. For instance, the authors may supplement performance comparisons or task adaptation analyses with SAM in either the methodology or experimental sections.
(2) While the current experiments are primarily based on too few public datasets, it is recommended to include test sets from diverse medical institutions or public medical image repositories (e.g., TCIA, MSD) to validate the model's robustness across different scenarios and pathological conditions.
(3) Although partial model comparisons have been presented, it is suggested to introduce representative methods from recent years (e.g., TransUNet, nnUNet, Swin-UNet, FocalNeXt) and conduct quantitative analyses using consistent key metrics (e.g., Dice, IoU, AUC) to strengthen the demonstration of the model's advantages.
Overall, the manuscript demonstrates publication potential. The incorporation of these suggestions would significantly enhance its research depth and practical value. A minor revision is recommended prior to acceptance.
Please refer to Review 1.
Please refer to Review 1.
none
Please revise the paper according to the reviewer‘s comments.
**PeerJ Staff Note:** It is PeerJ policy that additional references suggested during the peer-review process should only be included if the authors agree that they are relevant and useful.
This paper proposes an integrated artificial intelligence system based on deep learning for the classification and segmentation of pancreatic cancer. The system uses the enhanced UNet model segmentation technology for identification, the ResNext model for the classification of pancreatic cancer, and finally, the membrane group optimization algorithm is utilized to optimize the hyperparameters of the improved ResNext model, providing a new idea for improving the diagnostic performance of pancreatic cancer. However, there are some issues in the article that cannot be ignored.
My specific opinions are as follows.
Recommendation: Minor revision
1. Some references may have a role in advancing this manuscript, for example, “Oral multi-pathology segmentation with Lead-Assisting Backbone Attention Network and synthetic data generation”” From patches to WSIs: A systematic review of deep Multiple Instance Learning in computational pathology”, the descriptive details of the dataset need to be further refined, such as: dataset size, image size, etc.; In the related works section, some redundant content needs to be deleted
**PeerJ Staff Note:** It is PeerJ policy that additional references suggested during the peer-review process should only be included if the authors agree that they are relevant and useful.
Please consult the Basic Reporting for details.
Please consult the Basic Reporting for details.
Please consult the Basic Reporting for details.
Please revise the paper according to the reviewer's comments.
**PeerJ Staff Note:** Please ensure that all review and editorial 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.
**Language Note:** The review process has identified that the English language must be improved. PeerJ can provide language editing services - please contact us at [email protected] 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
The authors have addressed the reviewer's comments and I have no further comment.
The authors have addressed the reviewer's comments and I have no further comment.
The authors have addressed the reviewer's comments and I have no further comment.
While the manuscript presents valuable content, there are minor grammatical issues, occasional repeated words, and some instances where words are inadvertently joined without spaces. A careful proofreading would further improve the readability and presentation quality of the paper.
The current evaluation of the classification and segmentation models lacks sufficient comparison with recent state-of-the-art (SOTA) approaches. In particular, the classification model is primarily compared against traditional machine learning models, which does not provide an adequate benchmark given the advancements in deep learning-based methods. The authors are encouraged to incorporate comparisons with the latest relevant models to provide a more comprehensive assessment of their method's performance.
The images and tables in the manuscript would benefit from more descriptive and informative captions. Clear and detailed captions will help readers better understand the content and context of the presented results without having to refer extensively to the main text.
Please revise the paper according to the reviewer's comments.
1. What are the advantages of ResNext Model over ResNet Model? How to reflect the advantages of ResNext model in medical image segmentation?
2. What are the differences between the UNet model used in this article and existing methods? The author should make clear the main contribution of the paper.
3. The introduction and analysis of related work are not comprehensive enough. Some related works maybe useful for this paper: -- BDAL: Balanced Distribution Active Learning for MRI Cardiac Multistructures Segmentation, -- Local Variance-driven Level Set Model with Application to Segment Medical Images.
4. Experiment validations are not convincing. Some additional experiments need to be conducted to make its conclusion stronger.
no comment
no comment
Please revise the paper according to the reviewer's comments.
1. What are the advantages of ResNext Model over ResNet Model? How to reflect the advantages of ResNext model in medical image segmentation?
2. What are the differences between the UNet model used in this article and existing methods? The author should make clear the main contribution of the paper.
3. The introduction and analysis of related work are not comprehensive enough.
4. Experiment validations are not convincing. Some additional experiments need to be conducted to make its conclusion stronger.
no comment
no comment
The manuscript is generally well-written with clear, professional English throughout. The introduction provides a comprehensive background on pancreatic cancer detection and its challenges, effectively establishing the research context. However, some sentences in the methods section contain awkward phrasing that could benefit from revision. The literature review is thorough and appropriately cited, covering relevant recent work in the field. The paper's structure follows standard scientific reporting conventions, with a logical flow from introduction through conclusions. The motivation for developing an improved pancreatic cancer detection system is clearly articulated, though the specific gaps in existing approaches could be more explicitly stated.
The study's content aligns well with the journal's scope of computer science applications in medical imaging. The investigation demonstrates strong technical rigor, particularly in the model architecture design and evaluation methodology. The methods section provides detailed information about the EUNet and MResNext architectures, though some implementation details about the training process could be more specific (e.g., optimization parameters, hardware used). The data preprocessing steps are well described, including the hybrid filtering approach and contrast enhancement techniques. The evaluation metrics chosen (accuracy, sensitivity, specificity, IOU, DSC) are appropriate and comprehensive for the tasks. The authors have properly credited previous work and clearly distinguished their novel contributions.
The experimental results appear sound and well-documented, with appropriate statistical measures and comparisons to existing methods. The performance claims are supported by extensive quantitative results and ablation studies. The authors' arguments effectively address the goals outlined in the introduction, demonstrating improved accuracy in both segmentation and classification tasks. The confusion matrices and ROC curves provide convincing evidence of the model's effectiveness. However, the discussion of limitations could be more thorough - while technical limitations are mentioned, clinical implementation challenges are not fully addressed. The future directions section would benefit from more specific recommendations for extending this work. Additionally, while the dataset size is substantial (10,870 images), information about data diversity and potential biases could be more detailed.
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.