Kwon YS, Park TY, Kim SE, Park Y, Lee JG, Lee SP, Kim KO, Jang HJ, Yang YJ, Cho BJ. Deep learning-based localization and lesion detection in capsule endoscopy for patients with suspected small-bowel bleeding. World J Gastroenterol 2025; 31(27): 106819 [PMID: 40741104 DOI: 10.3748/wjg.v31.i27.106819]
Reader's ID:
08359836
Submitted on:
July 21, 2025, 16:11
Reader Expertise:
Reader’s expertise on the topic of the manuscript
Conflicts-of-Interest Statement:
Does the reader have a conflict of interest?
Reader Comment Standards for Published Articles:
1 Title
Does the title reflect the main subject/hypothesis of the manuscript?
2 Abstract
Does the abstract summarize and reflect the work described in the manuscript?
3 Key Words
Do the key words reflect the focus of the manuscript?
4 Background
Does the manuscript adequately describe the background, present status and significance of the study?
5 Methods
Does the manuscript describe methods (e.g., experiments, data analysis, surveys, and clinical trials, etc.) in adequate detail?
6 Results
Are the research objectives achieved by the experiments used in this study?
Has the study made meaningful contributions towards research progress in this field?
7 Discussion
Does the manuscript interpret the findings adequately and appropriately, highlighting the key points concisely, clearly and logically?
Are the findings and their applicability/relevance to the literature stated in a clear and definite manner?
Is the Discussion accurate and does it discuss the paper’s scientific significance and/or relevance to clinical practice sufficiently?
8 Illustrations and Tables
Are the figures, diagrams and tables sufficient, good quality and appropriately illustrative of the paper contents?
Do figures require labeling with arrows, asterisks, etc., or better legends?
9 Biostatistics
Does the manuscript meet the requirements of biostatistics?
10 Units
Does the manuscript meet the requirements of use of SI units?
11 References
Does the manuscript appropriately cite the latest, important and authoritative references in the Introduction and Discussion sections?
Does the author self-cite, omit, incorrectly cite and/or over-cite references?
12 Quality of manuscript organization and presentation
Is the manuscript concisely and coherently organized and presented?
Are the style, language and grammar accurate and appropriate?
13 Ethics statements
For all manuscripts involving human studies and/or animal experiments, author(s) must submit the related formal ethics documents that were reviewed and approved by their local ethical review committee. Did the manuscript meet the requirements of ethics?
Scientific Quality:
The overall quality of the manuscript, based on the above-listed criteria, should be evaluated and classified according to the following five categories
Language Quality:
Language quality (style, grammar, and spelling) should be evaluated and classified according to the following five categories.
Reader Comments:
This study aims to address the issues of time-consuming interpretation and reliance on the experience of readers when using small bowel capsule endoscopy (SBCE) to evaluate obscure gastrointestinal bleeding (OGIB). An AI model was developed to automatically distinguish the small intestine, stomach, and colon, and diagnose small intestinal abnormalities such as erosions, ulcers, vascular malformations, and bleeding. Using 87,005 images from 101 SBCE videos as the dataset, the researchers built models based on DenseNet161 and DenseNet201 convolutional neural networks, trained them through preprocessing and data augmentation, and validated their clinical utility on 32 external videos. The results showed that the model achieved over 99% accuracy in organ localization (AUC > 0.99) and 99.4% to 99.9% accuracy in lesion detection in internal testing. The AI-assisted reading time was significantly reduced to an average of 8.7 minutes (compared to 53.9 minutes traditionally), and the diagnostic performance was comparable to traditional reading. The discussion pointed out that the model significantly improved efficiency while maintaining high accuracy, but limitations included the dataset mainly consisting of Asian patients, being limited to a specific platform (MiroCam), having a small external validation sample (32 cases), incomplete coverage of lesion types, and the lack of advanced image enhancement techniques. The conclusion is that AI-assisted SBCE reading can reduce time and maintain performance, making it suitable for clinical implementation to improve consistency and efficiency. Readers believe this can help clinicians improve work efficiency and reduce misdiagnosis. Future research should expand the sample size, cover more lesion types, adapt to different platforms, and integrate advanced technologies to optimize diagnostic capabilities.
Reply from the Editorial Office:
First, thank you very much for your professional comments on the article published in World Journal of Gastroenterology.
Second, we read your comments with great interest. You are welcome to format your valuable comments into a Letter to the Editor and submit it online to World Journal of Gastroenterology at https://www.f6publishing.com. There are no restrictions on the number of words, figures (color, B/W) or authors for a Letter to the Editor. In addition, the article processing charge will be exempted for this Letter to the Editor. As with all articles published by the Baishideng Publishing Group, the Letter to the Editor will be published online after completing peer review. The guidelines for a Letter to the Editor can be found at: https://www.wjgnet.com/bpg/GerInfo/219.
Finally, we look forward to receiving your high-quality Letter to the Editor, which will promote academic communication and lead the development of this discipline.