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.
CRC diagnosis will be critical, and your DL contribution is highly appreciated.
[# PeerJ Staff Note - this decision was reviewed and approved by Mehmet Cunkas, a PeerJ Section Editor covering this Section #]
Good
OK
Just fine
None
-
-
-
Please follow the requests, criticisms, and suggestions of the reviewers assiduously.
No further issues.
Data preprocessing is lightly touched.
Evaluation methods, assessment metrics, and model selection methods are adequately described.
Experiments and evaluations performed adequately in my view, though no assessment of the impact or novelty was available.
Conclusions identify unresolved questions/limitations/ future directions to a fair extent.
The authors have addressed my outstanding comments.
-
-
**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.
**Language Note:** 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
-
Evaluation and methods as described by the authors seem adequate. Discussion is somewhat suboptimal. This may still be acceptable. Citations are fine.
Small sample as compared to some of the other models, such as UNI, CONCH, and Virchows, so the authors would require adjusting the claims.
The majority of my comments have been addressed in the revised manuscript submitted by the authors. However, several of the clarifications relating to the figures/tables seem to have been overlooked, e.g.:
- Figure 2 – it remains unclear why there are two labelling steps – the distinction between what is meant by ‘labelling’ in step 2 and ‘labelling’ in step 4 should be explained in the legend or within the figure, or a different term other than ‘labelling’ should be used for one of these steps to make the distinction.
- Table 3 – There should be consistency in the number of decimal places displayed. It is also unclear why precision/recall are not reported for some models (an explanation in the legend or text would be helpful).
- Figures 6 and 7 still appear to show identical images.
-
The text addition “It is important to note that the comparison presented in Table 3 may not be entirely equitable…” seems to have been repeated in the revised manuscript (pp. 17-18).
**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.
**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
1. While the authors summarize previous work, adding a brief critical analysis of each study's limitations or contributions would add more depth. This would also help to position the current study's novelty and contribution more effectively.
2. The paper should explicitly identify the gaps in research that the current study aims to fill. What specific limitations of previous studies does this research address?
1. The description of the ensemble model is clear. However, providing a visual representation of the architecture (e.g., a diagram) would be very helpful for the reader. The explanation of EfficientNetv2 and DenseNet121 is good but consider adding a sentence or two about why these specific architectures were chosen.
2. The training details are sufficient. However, consider adding details about the validation strategy (e.g., k-fold cross-validation) and any regularization techniques used.
1. The results are clearly presented, and the tables and figures are generally effective. However, ensure that all table and figure captions are self-explanatory.
2. The performance metrics are impressive, consider adding statistical significance testing to support the claim that the proposed model "significantly outperformed" other state-of-the-art methods if that's the case.
3. A more in-depth analysis of the model's errors. What types of images were misclassified? Why might this have happened?
1. The paper is generally well-written, but a thorough proofreading is recommended to eliminate any minor grammatical errors or typos.
2. Include a more detailed discussion of the limitations of the study and potential sources of bias.
This manuscript describes the use of a deep learning-based ensemble model for classifying whole slide images from colorectal resections into malignant or benign. In general, the manuscript provides a clear, coherent description of the authors’ approach and findings. However, some essential details are lacking in several places which would help improve clarity and aid interpretation/replication. This includes the following:
- Abstract:
o Include the dataset size, i.e., number of images/cases.
- Related work:
o Bilal et al., (2023) PMID: 37890902 may be another relevant work to cite.
- Methods:
o Additional dataset details would be informative, e.g. from what time period/dates were the cases acquired? And how many separate cases/patients were the 2000 images acquired from?
o More details on the scanning step e.g. scanner model used and image acquisition settings.
o Was the expert pathologist who classified the slides blinded to the original diagnosis?
- Figures/Tables:
o Figure 2: more detail in this figure would be useful, e.g. the number of cases, images, and patches at each stage. It is also unclear why there seem to be two labelling steps?
o Table 3: there should be consistency in number of decimal places displayed. It is also unclear why precision/recall are not reported for some models (an explanation in the legend would be helpful).
o Figure 5: the axes are cut off in the figure.
o Figure 6 and 7 are identical (presumably this should be just one single figure, not two).
o It would be preferable for viewing/comparison if Figures 8 and 9 were combined into one figure.
o All figure/table legends: more descriptive detail in the legends would aid interpretation.
There are several statements in the manuscript which seem to unequivocally suggest that utilising AI for diagnostic tasks in histopathology is proven to be an effective approach (examples below). However, it should be acknowledged that while there is strong potential in this area, from a clinical perspective, this field is still in its infancy, and it has not been fully proven that DL approaches can be effective for diagnosis in a routine clinical setting. The language should be toned down to take this into account:
- Abstract:
o “The integration of deep learning architectures in digital pathology proved to be an efficient and effective tool for assisting pathologists in diagnosing cancerous tissues”.
- Introduction:
o “The digital pathology field has had numerous breakthroughs over the years with the advancements in Artificial Intelligence (AI) and Deep Learning (DL) algorithms, where these histopathological images are utilized by such algorithms for an effective diagnosis [6]."
o "The recent innovations in DL architectures presented significant enhancements in CRC diagnosis by making it an efficient and reliable tool for pathologists to diagnose the cases effectively [7]."
o "DL architectures have proven to be effective in extracting complex and high-level features from images, making them suitable and reliable in diagnosing cancerous tissues [8]."
Minor clarification:
- “…where the tumors undergo several phases before placing it in microscopic slides”. It is not clear what is meant by ‘phases’ in this sentence – more specific language would improve the understanding here.
Missing methodological details noted above.
The authors compare the performance of their model on the test set, with the performance of other published methodologies (Table 3). However, it should be clearly acknowledged that this is not an entirely fair comparison, as the current model was trained on data from this same location, whereas the other models were not.
Some discussion and/or description of the samples that were misclassified as benign would be beneficial to understand where there are areas for improvement in the model.
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.