Jeremy W. Stewart
Bradley A. Barth
Isabel Rojas- Division of Pediatric Gastroenterology, Hepatology and Nutrition, University of Texas Southwestern Medical Center, Dallas, TX, United States
The utilization of artificial intelligence (AI) is rapidly expanding in all areas of medicine. Pediatric gastroenterology is among the fields exploring the use of AI to better visualize the gastrointestinal tract and improve diagnosis, disease subtyping, lesion detection, risk prediction, and treatment optimization for better patient outcomes. AI shows promising developments and applications in complex diseases, such as Crohn's disease, polyposis syndromes, and eosinophilic esophagitis, where diagnosis and initial or subsequent management are impacted by mucosal visualization and analysis. This article summarizes how AI, machine learning, and these complex networks work in addition to addressing the limitations and ethical challenges faced with use of this budding technology. Although most available information on this topic comes from adult literature, this discussion focuses on current and emerging pediatric research and applications of AI in pediatric diagnostic and interventional endoscopy.
1 Introduction
Artificial intelligence (AI) is increasingly integrated into daily life, improving efficiency, creativity, and productivity, from generating art and music to drafting documents and performing complex calculations. Healthcare has rapidly adopted AI, with an exponential rise in PubMed-indexed publications over the past 20–30 years. Between 2020 and 2025, over 150,000 articles included the terms “artificial intelligence, “deep learning,” or “machine learning,” of which 3,142 pertain to gastroenterology and only 220 specifically address pediatric populations (1).
Recently, generative AI (GAI) has emerged as a clinical tool capable of answering medical queries, assisting with documentation, and generating patient-facing materials (2). AI applications show promise in complex gastrointestinal diseases such as Crohn's disease, polyposis syndromes, and eosinophilic esophagitis, where diagnosis and management rely heavily on mucosal visualization and analysis. As in most emerging medical technologies, pediatric research lags behind adult studies. Pediatric gastroenterology, however, presents unique opportunities to apply AI and GAI in both clinical care and research. This article provides an overview of AI, GAI, and machine learning principles, and reviews current literature, limitations, and ethical challenges related to their use in visualizing the pediatric gastrointestinal tract.
2 How it works
A basic understanding of AI, its terminology and processes is essential. The terms artificial intelligence, generative AI, machine learning, and deep learning are often used interchangeably but represent distinct concepts.
Artificial intelligence broadly refers to the simulation of human intelligence by a system or a machine (3, 4). First coined by John McCarthy at a Dartmouth College conference in 1956, AI was initially developed to prove mathematical theorems and solve algebraic problems. AI requires perceptual, cognitive, and decision-making capabilities, traditionally assessed by the “Turing Test,” which evaluates whether a machine's responses are indistinguishable from a human's (5). According to Xu et al, AI development involves data collection, computing power, an AI framework, and machine learning. Generative AI refers to systems that generate new content after training on large datasets (4). Examples include creating an educational animation or video explaining endoscopy (6).
Machine learning (ML), a subset of AI, recognizes data patterns through algorithms such as regression models and decision trees. A generic form of ML is supervised learning; it uses labeled data with adjustable parameters to reach desired outputs, whereas unsupervised learning identifies features without labeled input. Deep learning (DL), a subset of ML, employs multilayered neural networks, e.g., convolutional neural networks (CNN), to create models for more complex pattern recognition, mimicking the human nervous system. Unlike ML, DL models “learn” from data through hierarchical pattern recognition and representation learning (7). Examples of ML include email filters or electronic medical record medications alerts, while DL applications include image or speech recognition. Understanding these distinctions is critical to applying AI effectively.
3 Visualization of the GI tract
As in other medical fields, AI shows promise in enhancing diagnostic precision and treatment accuracy through improved visualization of the gastrointestinal tract. A 2021 systematic review assessed AI and predictive models for detecting malignant and non-malignant lesions in the upper and lower GI tract, hepatobiliary system, and pancreas (8). A meta-analysis of 43 studies involving 15,000 tandem colonoscopies reported miss rates of 26% for adenomas, 9% for advanced adenomas, and 27% for serrated polyps, underscoring the potential of AI to reduce missed lesions (9).
The “Computer-aided detection” (CADe) systems aim to improve real-time identification of polyps and malignancies during colonoscopy. One CADe model, trained on 69,000 chromoendoscopy and narrow band imaging (NBI) images, achieved accuracies of 98% and 96%, respectively, exceeding expert and nonexpert endoscopists (10). CADe systems demonstrate high sensitivity but depend on adequate bowel preparation. Table 1 lists currently FDA-approved CADe and other AI systems in the United States for gastroenterology (11–18).
Table 1. This table includes the currently FDA approved systems for lesion detection available in the United States with the corresponding company and/or manufacturer.
In contrast, “computer-aided diagnosis” (CADx) systems classify detected abnormalities to guide management decisions, such as distinguishing neoplastic from non-neoplastic lesions (19). Limitations include scarce external validation and the absence of FDA- approved systems in the United States.
Although no CADe or CADx models have been studied specifically in pediatrics, their application could benefit polyposis disorders such as juvenile polyposis syndrome, Peutz-Jeghers syndrome, and familial adenomatous polyposis, where routine endoscopic surveillance is essential (20–22). Broader adoption would require multicenter data collection and pediatric-specific model training.
3.1 Video capsule endoscopy
Since its FDA approval in 2001, video capsule endoscopy (VCE) has provided a less invasive method to visualize the gastrointestinal tract. Indications include small bowel bleeding, Crohn's disease activity, surveillance of polyposis syndromes, and evaluation of malabsorption disorders such as Celiac disease (23). The capsule, swallowed or placed endoscopically captures 2–35 images per second over 8–10 h, creating a substantial time burden for gastroenterologists and making it an ideal target for AI-assisted analysis (23, 24). In a multicenter trial of 137 patients across 13 European centers, the Navicam SB system using ProScan (a deep neural network-based AI system) reduced average reading time from 33.7 to 3.8 min and improved lesion detection from 62.4% to 73.7%. Kroner et al. also reviewed AI use in VCE since early ML development, and the review summarizes findings from 31 studies of GI bleeding, angioectasia, small intestinal ulcers, celiac disease, and even Hookworm infection (8, 25).
Ding et al. developed a DL model from over 100 million images from VCE, to identify and categorize findings, achieving 99.9% sensitivity and 99.88% specificity, compared with gastroenterologists’ 74.6% and 76.9% sensitivities for “per-patient” and “per-lesion” analyses, respectively (24). Another striking finding from this study includes the reading time of the CNN model of 5.9 min when compared to 96.6 min of conventional reading (24). Another multicenter CNN study analyzed 66,208 images using RAPID reader software QuickView mode, identifying 44,684 abnormalities, mucosal breaks, angioectasia, protruding lesions, and blood content in alignment with CEST standardized reporting terminology, with detection rates of 95.7%, 75.9%, 98.8%, and 100% respectively, outperforming QuickView (99% vs. 89%) (26, 27). Its adaptive sampling rate allowed more efficient image review. Additional AI-based VCE studies have combined endoscopic, histologic, MRI, and genetic data for IBD detection in adults.
While pediatric comparative studies are lacking, small bowel findings are generally similar between adults and children. Huang et al. retrospectively analyzed VCE from 162 pediatric patients using four DL models (DenseNet121, Visual Geometry Group-16, ResNet50, and Vision Transformer), with DensNet121 and Resnet50 achieving 90.6% and 90.5% accuracy, respectively (28). This remains the only pediatric-specific VCE AI published study to date, highlighting the need for larger pediatric datasets to enhance accuracy and generalizability.
3.2 Inflammatory bowel disease
Inflammatory bowel disease (IBD) can involve the entire gastrointestinal tract and requires accurate histologic, clinical, and endoscopic assessment for diagnosis and staging. AI-assisted technologies have shown potential benefits in imaging, biomarker evaluation, and treatment decisions (29–31). Endoscopic imaging has also been used to develop DL models for evaluating and grading inflammatory bowel disease (IBD) severity. Because grading systems are subject to inter- and intra-observer variability, AI-assisted technology can help standardize assessments. A DL model for ulcerative colitis (UC) severity grading demonstrated sensitivity, specificity, and accuracy comparable to experienced human reviewers (32). Another study developed a multilayered CNN that analyzed 16,514 colonoscopic images from 3,082 patients to distinguish remission (Mayo score 0 or 1) from moderate to severe disease (Mayo score 2 or 3), achieving 83% sensitivity and 96% specificity (33). A meta-analysis of 12 studies evaluating deep ML and CNN algorithms found excellent accuracy, sensitivity, and specificity for UC severity scoring, with the Ulcerative Colitis Endoscopic Index of Severity (UCEIS) outperforming the Mayo Endoscopic Score (MES) (34).
Although multiple adult studies support AI-assisted endoscopic visualization in IBD, similar pediatric studies are lacking. This gap highlights the need for pediatric-specific AI models, particularly given the distinct features and presentation patterns of pediatric and very-early-onset IBD (35).
3.3 Celiac disease
The use of DL and ML in Celiac disease has been explored by comparing expert human readers with machine learning algorithm diagnosis and disease activity monitoring of VCE studies of 63 adult patients. The study found strong agreement between the two groups in identifying villous damage (36). Other DL models have aimed to improve lesion detection accuracy. For example, Ciaccio et al. Developed a “color masking” technique to filter extraneous image features, achieving 80% accuracy in distinguishing villous atrophy compared to normal controls (37). Although pediatric-focused studies are limited, Syed et al. Developed a CNN to differentiate pathological vs. healthy duodenal tissue in 102 children across multiple institutions, reporting a 93.4% detection accuracy (38). These findings highlight the promise of AI-based tools for enhancing diagnostic precision in both adult and pediatric celiac disease, warranting further validation in larger pediatric cohorts.
3.4 Other uses of AI in gastroenterology
In another adult study, a DL algorithm-based diagnostic system, HOPE AI, was developed using 308,887 endoscopic images and 197 videos from 6,207 patients to detect Helicobacter pylori infection. The model demonstrated higher sensitivity than senior endoscopists (85.7% vs. 68%), illustrating the potential of AI to enhance diagnostic accuracy. The study also incorporated external geographic validation across multiple centers and prospectively enrolled patients in later phases, strengthening its generalizability (39).
4 Medical society guidelines
Recognizing the growing role of AI in gastroenterology, several major societies have recently issued guidance. In January 2025, the American Society for Gastrointestinal Endoscopy (ASGE) AI Task Force released consensus statements outlining the current landscape of AI applications, developed by experts in endoscopy, technology, regulatory authorities, and other subspecialties (40). These statements emphasized AI's potential to enhance lesion detection and characterization, data quality and modeling accuracy, diagnostic precision, prognostication, clinical research, and medical education. They also encourage collaboration among engineers, researchers, and gastroenterologists to advance understanding and responsible integration of AI.
In April 2025, the American Gastroenterology Association (AGA) published a living clinical practice guideline on CADe systems, issuing no recommendation for or against their use due to the “close tradeoff between desirable and undesirable effects” and current evidence limitations (41). Similarly, the European Society of Gastrointestinal Endoscopy (ESGE) released a position statement comparing AI to experienced endoscopists, addressing applications such as landmark recognition and completeness in upper GI endoscopy, detection and resection of neoplastic or cancerous lesions, automated reading of small bowel capsule studies, and characterization (CADx) of polyps of ≤5 mm and selection for resection in ≥6 mm (42).
To date, the North American Society for Pediatric Gastroenterology, Hepatology & Nutrition (NASPGHAN) has not published position statements or guidelines on the use of AI in pediatric gastroenterology.
5 Teaching
Artificial intelligence is already influencing the education and training of adult and pediatric gastroenterologist fellows, much like other medical specialties impacted by its integration. Although no studies specifically address its effect on pediatric trainees, Kang et al. reviewed the potential benefits, challenges, and limitations AI presents to medical educators and learners. A key point from this review emphasized that the question is not if AI will be used, but how it should be applied (43). While there are several studies citing improved detection rates of lesions with trainee utilization of AI detection in VCE (44) and reduced reading time (45), misidentification of lesions as neoplastic with increase in unnecessary biopsies or interventions, and thus risk for harm, were noted with the CADe system (10); however, the CADx system previously discussed is a potential tool for trainees in an added layer of diagnostic assessments (46, 47). One concern in medical education (and other areas of teaching) is the potential overreliance on technology at the expense of experiential learning. The impact of AI on trainee education, particularly in pediatric gastroenterology, warrants further exploration to ensure that educational quality and clinical judgement are preserved for the next generation of physicians.
6 Limitations
While AI has certainly proven a valuable technological advance, several limitations remain in its application to gastroenterology imaging. A recurring theme throughout this manuscript is the limited pediatric data and the need for caution when extrapolating adult findings to children. A manual review of the pediatric radiology literature involving AI and machine learning identified small dataset size as the most significant limitation, followed by lack of external validation and other methodological constraints (48). Small sample sizes and preclinical studies hinder robust dataset generation; however, CADe remains one of the most extensively evaluated AI technologies within gastroenterology (49, 50).
A limited number of studies also complicates differentiation between IBD and mimicking pathologies, such as Behçet's disease, gastrointestinal tuberculosis, ischemic colitis, infectious colitis, and distinguishing Crohn's disease from Ulcerative Colitis (51–55). One study evaluating eye-tracking metrics and polyp detection with CADe systems found similar reaction times but increased misinterpretation of normal mucosa, underscoring the importance of maintaining physician oversight rather than deferring to automated systems (56).
Cost-effectiveness is another consideration. Although economic analyses are limited, one microsimulation study estimated AI-assisted colon cancer screening could save $57 per individual screened, an approximately $290 million if applied at the U.S. population level (57). Finally, automation bias is a concern for physicians where the output of an AI system is chosen over the physician's own decisions, even if incorrect; conversely, automation aversion can develop when a physician may not trust the output even if it is correct (50, 58). Additional limitations are evident when considering the ethical and regulatory factors discussed in the following section. Ultimately, AI systems, models, and networks need larger, multicenter studies with external validation before widespread use.
7 Ethics
As gastroenterology specialists explore the use of AI to enhance patient care, practice efficiency, cost-effectiveness, and reduce administrative load, it is essential to recognize the accompanying ethical challenges. Responsible implementation requires maintaining safety, transparency, and ethical integrity. While ethical considerations in AI warrant extensive discussion, this section provides only a brief overview.
Artificial intelligence introduces multiple ethical concerns, including lack of transparency and reliability, potential for bias, data security and confidentiality, inequity, potential for “hallucinations,”, and environmental impact (59, 60). Because system training depends on real patient data, ensuring privacy and compliance with the Health Insurance Portability Accountability Act of 1996 (HIPAA) remains a significant challenge (61). Although training data should be de-identified, the risk of “re-identification” persists, especially in small or rare pediatric datasets where case uniqueness may inadvertently reveal patient identity (62). Bias may arise when training data do not represent the true population, potentially perpetuating inequities in diagnosis or treatment (63, 64). Transparency is undoubtably one of the challenging aspects of AI, the so-called “black box” problem, where the inner workings of complex algorithms are not easily explainable. Proposed strategies to mitigate this issue include transparency of data sources, algorithms, processes, and outcomes, allowing users to interpret and validate results (65, 66).
AI “hallucinations”, a phenomenon seen in large language models, further complicate trust and reliability. These models can generate inaccurate, unreliable diagnostic and therapeutic data from flawed reasoning pathways (67). For instance, a generative AI tool might produce an image of a physician with anatomically inconsistent details, an example of nonsensical but confident output. Mitigation strategies include quantifying hallucination frequency and enforcing continuous oversight and stewardship of large language models (68).
Regulatory efforts toward the use of AI are at a pivotal point as we see a dramatic incorporation of its use in everyday life and in healthcare. Currently the FDA, European Union (EU), World Health Organization (WHO), and United Kingdom (UK) are among the many governing bodies with processes for approval of new or emerging AI technology with attention to the training data, performance, bias, post-marketing plans, changes (if added to an existing approved technology), monitoring, transparency, and safety, among others (69–74).
8 Future direction
The future for AI in pediatric GI is likely to mirror the current research in adults. With reduction in reading time reported by VCE, these robust studies will likely continue to improve datasets and thus the efficiency of the models developed (75). Cloud-based AI detection software is also likely to continue to develop like the OLYSENSE CAD/AI, which has approval by the FDA for CADDIE to assist in detecting (without diagnosing) colorectal polyps (17), but the software has support for SMARTIBD to aid in analyzing ulcerative colitis during colonoscopy (76). Models that combine different aspects of diagnoses, like histological findings with visual findings, particularly in IBD, are also expected to continue to develop. Vision Transformer (ViT) architecture is another emerging technology that evaluates images across multiple models and networks to capture the relationships of complex datasets. ViT or a combination of ViT-CNN could be used to detect and filter image imperfections, such as poor lighting, debris, bubbles or poor prep (77, 78). Utilizing predetermined change control plans (PCCPs), the FDA issued guidance to support a simplified approval pipeline to allow systems to be updated without a new device approval (79).
9 Discussion
Artificial intelligence technologies, including machine learning, deep learning, convolutional neural networks, large language models, and generative AI, will continue to shape healthcare. These models and systems show promise of reducing errors or missed lesions, as in CADe/CADx systems, reduce time spent viewing thousands of images in VCE, better characterize chronic diseases to help create precision medicine in IBD and other diseases of the GI tract, and have potential to help improve efficiency to minimize additional cognitive load. As we recognize the roles AI can play in the care of patients, it is evident that more pediatric data across more centers is needed to create robust and accurate dataset training to prevent bias, among other challenges, to develop better models for Pediatric GI. Being at the forefront of the development of these models and systems, we have a duty to our trainees to not only prevent reliance on AI for patient care but also to instill that AI is best used to assist and augment our current care without replacing our knowledge and skills. We should strive to maintain a boundary of assistance from the tools while stressing the importance of validation and approval of the information, algorithm, or other output generated. Clinicians must also strive to continue preserve patient autonomy, beneficence, non-maleficence, and maintain justice in the ethical use of this developing and emerging technology.
Author contributions
JS: Writing – review & editing, Writing – original draft. BB: Writing – review & editing, Writing – original draft. IR: Writing – original draft, Writing – review & editing.
Funding
The author(s) declared that financial support was not received for this work and/or its publication.
Conflict of interest
The author(s) declared that this work was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
Generative AI statement
The author(s) declared that generative AI was used in the creation of this manuscript. Generative AI was used to assist in language editing, improving clarity, and ensuring style consistency. These tools were not used to generate content or interpret data.
Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.
Publisher's note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
References
1. PubMed search: (([(artificial intelligence) OR (machine learning)] OR (deep learning)) AND (gastroenterology)) AND (pediatric). Available online at: https://pubmed.ncbi.nlm.nih.gov/?term=%28%28%28artificial+intelligence%29+OR+%28machine+learning%29%29+OR+%28deep+learning%29%29+AND+%28gastroenterology%29%29+AND+%28pediatric%29 (Accessed October 28, 2025).
2. Reddy S. Generative AI in healthcare: an implementation science informed translational path on application, integration and governance. Implement Sci. (2024) 19(1):27. doi: 10.1186/s13012-024-01357-9
3. Amisha MP, Pathania M, Rathaur VK. Overview of artificial intelligence in medicine. J Family Med Prim Care. (2019) 8(7):2328–31. doi: 10.4103/jfmpc.jfmpc_440_19
4. Xu Y, Liu X, Cao X, Huang C, Liu E, Qian S, et al. Artificial intelligence: a powerful paradigm for scientific research. Innovation (Camb). (2021) 2(4):100179. doi: 10.1016/j.xinn.2021.100179
5. Turing AM. I.—computing machinery and intelligence. Mind. (1950) LIX(236):433–60. doi: 10.1093/mind/LIX.236.433
6. World Intellectual Property O. Generative AI. Geneva: World Intellectual Property Organization (2024).
7. LeCun Y, Bengio Y, Hinton G. Deep learning. Nature. (2015) 521(7553):436–44. doi: 10.1038/nature14539
8. Kröner PT, Engels MM, Glicksberg BS, Johnson KW, Mzaik O, van Hooft JE, et al. Artificial intelligence in gastroenterology: a state-of-the-art review. World J Gastroenterol. (2021) 27(40):6794–824. doi: 10.3748/wjg.v27.i40.6794
9. Zhao S, Wang S, Pan P, Xia T, Chang X, Yang X, et al. Magnitude, risk factors, and factors associated with adenoma miss rate of tandem colonoscopy: a systematic review and meta-analysis. Gastroenterology. (2019) 156(6):1661–74. doi: 10.1053/j.gastro.2019.01.260
10. Kudo S-E, Misawa M, Mori Y, Hotta K, Ohtsuka K, Ikematsu H, et al. Artificial intelligence-assisted system improves endoscopic identification of colorectal neoplasms. Clin Gastroenterol Hepatol. (2020) 18(8):1874. doi: 10.1016/j.cgh.2019.09.009
11. Food US, Drug A. FDA 510(k): GI Genius (K211951). 510(k) Premarket Notification (2021). Available online at: https://www.accessdata.fda.gov/scripts/cdrh/cfdocs/cfpmn/pmn.cfm?ID=K211951 (Accessed October 29, 2025).
12. Food US, Drug A. FDA 510(k): GI Genius Modules and ColonPRO 4.0 (K233964). 510(k) Premarket Notification (2024). Available online at: https://www.accessdata.fda.gov/scripts/cdrh/cfdocs/cfpmn/pmn.cfm?ID=K233964 (Accessed October 29, 2025).
13. Food US, Drug A. FDA 510(k): EndoScreener (K211326). 510(k) Premarket Notification (2021). Available online at: https://www.accessdata.fda.gov/scripts/cdrh/cfdocs/cfpmn/pmn.cfm?ID=K211326 (Accessed October 29, 2025).
14. Food US, Drug A. FDA 510(k): SKOUT Software (K213686). 510(k) Premarket Notification (2022). Available online at: https://www.accessdata.fda.gov/scripts/cdrh/cfdocs/cfpmn/pmn.cfm?ID=K213686 (Accessed October 29, 2025).
15. Food US, Drug A. FDA 510(k): ME-APDS; MAGENTIQ-COLO (K223473). 510(k) Premarket Notification (2023). Available online at: https://www.accessdata.fda.gov/scripts/cdrh/cfdocs/cfpmn/pmn.cfm?ID=K223473 (Accessed October 29, 2025).
16. Food US, Drug A. FDA 510(k): CAD EYE/EW10-EC02 Endoscopy Support Program (K230751). 510(k) Premarket Notification (2023). Available online at: https://www.accessdata.fda.gov/scripts/cdrh/cfdocs/cfpmn/pmn.cfm?ID=K230751 (Accessed October 29, 2025).
17. Food US, Drug A. FDA 510(k): CADDIE (K240044). 510(k) Premarket Notification (2024). Available online at: https://www.accessdata.fda.gov/scripts/cdrh/cfdocs/cfpmn/pmn.cfm?ID=K240044 (Accessed October 29, 2025).
18. Food US, Drug A. FDA De Novo Classification: NaviCam ProScan (DEN230027) (2023). Available online at: https://www.accessdata.fda.gov/scripts/cdrh/cfdocs/cfpmn/denovo.cfm?id=DEN230027 (Accessed October 29, 2025).
19. Hassan C, Badalamenti M, Maselli R, Correale L, Iannone A, Radaelli F, et al. Computer-aided detection-assisted colonoscopy: classification and relevance of false positives. Gastrointest Endosc. (2020) 92(4):900–4. doi: 10.1016/j.gie.2020.06.021
20. Attard TM, Cohen S, Durno C. Polyps and polyposis syndromes in children: novel endoscopic considerations. Gastrointest Endosc Clin N Am. (2023) 33(2):463–86. doi: 10.1016/j.giec.2022.11.001
21. Phen C, Rojas I. Paediatric polyposis syndromes: burden of disease and current concepts. Curr Opin Pediatr. (2021) 33(5):509–14. doi: 10.1097/MOP.0000000000001044
22. MacFarland SP, Zelley K, Katona BW, Wilkins BJ, Brodeur GM, Mamula P. Gastrointestinal polyposis in pediatric patients. J Pediatr Gastroenterol Nutr. (2019) 69(3):273–80. doi: 10.1097/MPG.0000000000002421
23. Melson J, Trikudanathan G, Abu Dayyeh BK, Bhutani MS, Chandrasekhara V, Jirapinyo P, et al. Video capsule endoscopy. Gastrointest Endosc. (2021) 93(4):784–96. doi: 10.1016/j.gie.2020.12.001
24. Ding Z, Shi H, Zhang H, Meng L, Fan M, Han C, et al. Gastroenterologist-level identification of small-bowel diseases and normal variants by capsule endoscopy using a deep-learning model. Gastroenterology. (2019) 157(4):1044–54.e5. doi: 10.1053/j.gastro.2019.06.025
25. He JY, Wu X, Jiang YG, Peng Q, Jain R. Hookworm detection in wireless capsule endoscopy images with deep learning. IEEE Trans Image Process. (2018) 27(5):2379–92. doi: 10.1109/TIP.2018.2801119
26. Aoki T, Yamada A, Kato Y, Saito H, Tsuboi A, Nakada A, et al. Automatic detection of various abnormalities in capsule endoscopy videos by a deep learning-based system: a multicenter study. Gastrointest Endosc. (2021) 93(1):165. doi: 10.1016/j.gie.2020.04.080
27. Korman LY, Delvaux M, Gay G, Hagenmuller F, Keuchel M, Friedman S, et al. Capsule endoscopy structured terminology (CEST): proposal of a standardized and structured terminology for reporting capsule endoscopy procedures. Endoscopy. (2005) 37(10):951–9. doi: 10.1055/s-2005-870329
28. Huang Y-H, Lin Q, Jin X-Y, Chou C-Y, Wei J-J, Xing J, et al. Classification of pediatric video capsule endoscopy images for small bowel abnormalities using deep learning models. World J Gastroenterol. (2025) 31(21):107601. doi: 10.3748/wjg.v31.i21.107601
29. De Deo D, Dal Buono A, Gabbiadini R, Nardone OM, Ferreiro-Iglesias R, Privitera G, et al. Digital biomarkers and artificial intelligence: a new frontier in personalized management of inflammatory bowel disease. Front Immunol. (2025) 16:1637159. doi: 10.3389/fimmu.2025.1637159
30. Okpete UE, Byeon H. Explainable artificial intelligence for personalized management of inflammatory bowel disease: a minireview of recent advances. World J Gastroenterol. (2025) 31(35):111033. doi: 10.3748/wjg.v31.i35.111033
31. Tariq R, Afzali A. Artificial intelligence in inflammatory bowel disease: innovations in diagnosis, monitoring, and personalized care. Therap Adv Gastroenterol. (2025) 18:17562848251357407. doi: 10.1177/17562848251357407
32. Benjamin GB, Stefan F, Heming Y, Jerome L, Rafal G, Bartosz M, et al. Ulcerative colitis severity classification and localized extent (UC-SCALE): an artificial intelligence scoring system for a spatial assessment of disease severity in ulcerative colitis. J Crohns Colitis. (2025) 19(1):jjae187. doi: 10.1093/ecco-jcc/jjae187
33. Stidham RW, Liu W, Zhu J, Nallamothu BK, Waljee AK, Bishu S, et al. Performance of a deep learning model vs human reviewers in grading endoscopic disease severity of patients with ulcerative colitis. JAMA Netw Open. (2019) 2(5):e193963. doi: 10.1001/jamanetworkopen.2019.3963
34. Jahagirdar V, Bapaye J, Chandan S, Ponnada S, Kochhar GS, Navaneethan U, et al. Diagnostic accuracy of convolutional neural network-based machine learning algorithms in endoscopic severity prediction of ulcerative colitis: a systematic review and meta-analysis. Gastrointest Endosc. (2023) 98(2):145–54.e8. doi: 10.1016/j.gie.2023.04.2074
35. Ozawa T, Ishihara S, Fujishiro M, Saito H, Kumagai Y, Shichijo S, et al. Novel computer-assisted diagnosis system for endoscopic disease activity in patients with ulcerative colitis. Gastrointest Endosc. (2019) 89(2):416–21.e1. doi: 10.1016/j.gie.2018.10.020
36. Zammit SC, McAlindon ME, Greenblatt E, Maker M, Siegelman J, Leffler DA, et al. Quantification of celiac disease severity using video capsule endoscopy: a comparison of human experts and machine learning algorithms. Current Med Imaging. (2023) 19(12):1455–662. doi: 10.2174/1573405619666230123110957
37. Ciaccio EJ, Lewis SK, Bhagat G, Green PH. Color masking improves classification of celiac disease in videocapsule endoscopy images. Comput Biol Med. (2019) 106:150–6. doi: 10.1016/j.compbiomed.2018.12.011
38. Syed S, Khan MN, Moore SR, Sadiq K, Iqbal NT, Ali SA, et al. Assessment of machine learning detection of environmental enteropathy and celiac disease in children. JAMA Netw Open. (2019) 2(6):e195822. doi: 10.1001/jamanetworkopen.2019.5822
39. Hu Y, Xu J, Huang L, Zheng Z, Zhao J, Chen T, et al. Artificial intelligence–assisted endoscopic diagnosis system for diagnosing Helicobacter pylori infection: a multicenter study. BMC Med. (2025) 23(1):1–12. doi: 10.1186/s12916-025-04379-2
40. Parasa S, Berzin T, Leggett C, Gross S, Repici A, Ahmad OF, et al. Consensus statements on the current landscape of artificial intelligence applications in endoscopy, addressing roadblocks, and advancing artificial intelligence in gastroenterology. Gastrointest Endosc. (2025) 101(1):2–9. doi: 10.1016/j.gie.2023.12.003
41. Sultan S, Shung DL, Kolb JM, Foroutan F, Hassan C, Kahi CJ, et al. AGA living clinical practice guideline on computer-aided detection-assisted colonoscopy. Gastroenterology. (2025) 168(4):691–700. doi: 10.1053/j.gastro.2025.01.002
42. Messmann H, Bisschops R, Antonelli G, Libânio D, Sinonquel P, Abdelrahim M, et al. Expected value of artificial intelligence in gastrointestinal endoscopy: european society of gastrointestinal endoscopy (ESGE) position statement. Endoscopy. (2022) 54(12):1211–31. doi: 10.1055/a-1950-5694
43. Kang AJ, Rodrigues T, Patel RV, Keswani RN. Impact of artificial intelligence on gastroenterology trainee education. Gastrointest Endosc Clin N Am. (2025) 35(2):457–67. doi: 10.1016/j.giec.2024.12.008
44. Tomonori A, Atsuo Y, Kazuharu A, Hiroaki S, Gota F, Nariaki O, et al. Clinical usefulness of a deep learning-based system as the first screening on small-bowel capsule endoscopy Reading. Dig Endosc. (2019) 32(4):585–91. doi: 10.1111/den.13517
45. Park J, Hwang Y, Nam JH, Oh DJ, Kim KB, Song HJ, et al. Artificial intelligence that determines the clinical significance of capsule endoscopy images can increase the efficiency of reading. PLoS One. (2020) 15(10):1–12. doi: 10.1371/journal.pone.0241474
46. Nemoto D, Guo Z, Katsuki S, Takezawa T, Maemoto R, Kawasaki K, et al. Computer-aided diagnosis of early-stage colorectal cancer using nonmagnified endoscopic white-light images (with videos). Gastrointest Endosc. (2023) 98(1):90–9. doi: 10.1016/j.gie.2023.01.050
47. Naohisa Y, Ken I, Yuri T, Reo K, Hikaru H, Satoshi S, et al. An analysis about the function of a new artificial intelligence, CAD EYE with the lesion recognition and diagnosis for colorectal polyps in clinical practice. Int J Colorectal Dis. (2021) 36(10):2237–45. doi: 10.1007/s00384-021-04006-5
48. Alkhulaifat D, Rafful P, Khalkhali V, Welsh M, Sotardi ST. Implications of pediatric artificial intelligence challenges for artificial intelligence education and curriculum development. J Am Coll Radiol. (2023) 20(8):724–9. doi: 10.1016/j.jacr.2023.04.013
49. Hogg HDJ, Martindale APL, Liu X, Denniston AK. Clinical evaluation of artificial intelligence-enabled interventions. Invest Ophthalmol Vis Sci. (2024) 65(10):10. doi: 10.1167/iovs.65.10.10
50. El-Sayed A, Lovat LB, Ahmad OF. Clinical implementation of artificial intelligence in gastroenterology: current landscape, regulatory challenges, and ethical issues. Gastroenterology. (2025) 169(3):518–30. doi: 10.1053/j.gastro.2025.01.254
51. Kim JM, Kang JG, Kim S, Cheon JH. Deep-learning system for real-time differentiation between crohn’s disease, intestinal behçet’s disease, and intestinal tuberculosis. J Gastroenterol Hepatol. (2021) 36(8):2141–8. doi: 10.1111/jgh.15433
52. Guimarães P, Finkler H, Reichert MC, Zimmer V, Grünhage F, Krawczyk M, et al. Artificial-intelligence-based decision support tools for the differential diagnosis of colitis. Eur J Clin Invest. (2023) 53(6):e13960. doi: 10.1111/eci.13960
53. Wang L, Chen L, Wang X, Liu K, Li T, Yu Y, et al. Development of a convolutional neural network-based colonoscopy image assessment model for differentiating crohn’s disease and ulcerative colitis. Front Med (Lausanne). (2022) 9:789862. doi: 10.3389/fmed.2022.789862
54. Brodersen JB, Jensen MD, Leenhardt R, Kjeldsen J, Histace A, Knudsen T, et al. Artificial intelligence-assisted analysis of pan-enteric capsule endoscopy in patients with suspected crohn’s disease: a study on diagnostic performance. J Crohns Colitis. (2024) 18(1):75–81. doi: 10.1093/ecco-jcc/jjad131
55. Quénéhervé L, David G, Bourreille A, Hardouin JB, Rahmi G, Neunlist M, et al. Quantitative assessment of mucosal architecture using computer-based analysis of confocal laser endomicroscopy in inflammatory bowel diseases. Gastrointest Endosc. (2019) 89(3):626–36. doi: 10.1016/j.gie.2018.08.006
56. Troya J, Fitting D, Brand M, Sudarevic B, Kather JN, Meining A, et al. The influence of computer-aided polyp detection systems on reaction time for polyp detection and eye gaze. Endoscopy. (2022) 54(10):1009–14. doi: 10.1055/a-1770-7353
57. Areia M, Mori Y, Correale L, Repici A, Bretthauer M, Sharma P, et al. Cost-effectiveness of artificial intelligence for screening colonoscopy: a modelling study. Lancet Digit Health. (2022) 4(6):e436–e44. doi: 10.1016/S2589-7500(22)00042-5
58. Dietvorst BJ, Simmons JP, Massey C. Algorithm aversion: people erroneously avoid algorithms after seeing them err. J Exp Psychol Gen. (2015) 144(1):114–26. doi: 10.1037/xge0000033
59. Leslie D. Understanding artificial intelligence ethics and safety: a guide for the responsible design and implementation of AI systems in the public sector. SSRN Electron J. (2019):11–73. doi: 10.2139/ssrn.3403301
60. Selinger E, Carroll T. The ethics of empathetic AI in medicine. IEEE Transac Technol Soc. (2025) 6(3):276–82. doi: 10.1109/TTS.2025.3563812
62. Dankar FK, El Emam K, Neisa A, Roffey T. Estimating the re-identification risk of clinical data sets. BMC Med Inform Decis Mak. (2012) 12(1):66–80. doi: 10.1186/1472-6947-12-66
63. Cassidy N, John TOB. Assessing the risks posed by the convergence of artificial intelligence and biotechnology. Health Secur. (2020) 18(3):219–27. doi: 10.1089/hs.2019.0122
64. Dang A-H, Vu T, Le-Minh N. Survey and analysis of hallucinations in large language models: attribution to prompting strategies or model behavior. Front Artif Intell. (2025) 8:1622292. doi: 10.3389/frai.2025.1622292
65. Weiner EB, Dankwa-Mullan I, Nelson WA, Hassanpour S. Ethical challenges and evolving strategies in the integration of artificial intelligence into clinical practice. PLOS Digit Health. (2025) 4(4):e0000810. doi: 10.1371/journal.pdig.0000810
66. Goktas P, Grzybowski A. Shaping the future of healthcare: ethical clinical challenges and pathways to trustworthy AI. J Clin Med. (2025) 14(5):1605. doi: 10.3390/jcm14051605
67. Roustan D, Bastardot F. The Clinicians’ guide to large language models: a general perspective with a focus on hallucinations. Interact J Med Res. (2025) 14:e59823. doi: 10.2196/59823
68. Ong JCL, Chang SY-H, William W, Butte AJ, Shah NH, Chew LST, et al. Ethical and regulatory challenges of large language models in medicine. Lancet Digit Health. (2024) 6(6):e428–e32. doi: 10.1016/S2589-7500(24)00061-X
69. Food US, Drug A. AI-Enabled Device Software Functions: Lifecycle Management and Marketing Submission (Draft Guidance for Industry and Food and Drug Administration Staff). Silver Spring, MD: U.S. Food and Drug Administration (2025).
70. World Health O. Ethics and Governance of Artificial Intelligence for Health. Geneva: World Health Organization (2021).
71. World Health O. Ethics and Governance of Artificial Intelligence for Health: Guidance on Large Multimodal Models. Geneva: World Health Organization (2025).
72. Medicines, Healthcare products Regulatory A. Software and AI as a Medical Device (AIaMD): Change Programme Roadmap. London: Medicines and Healthcare products Regulatory Agency (2025).
73. European C. Medical Device Coordination G. Interplay Between the Medical Device Regulation (MDR), the in Vitro Diagnostic Medical Devices Regulation (IVDR), and the Artificial Intelligence Act (MDCG 2025-6/AIB 2025-1). Brussels: European Commission (2025).
74. Food US, Drug A, Health C. Medicines, Healthcare Products Regulatory A. Predetermined Change Control Plans for Machine Learning-Enabled Medical Devices: Guiding Principles. Silver Spring, MD: U.S. Food and Drug Administration (2025).
75. Spada C, Piccirelli S, Hassan C, Ferrari C, Toth E, González-Suárez B, et al. AI-assisted capsule endoscopy Reading in suspected small bowel bleeding: a multicentre prospective study. Lancet Digit Health. (2024) 6(5):e345–e53. doi: 10.1016/S2589-7500(24)00048-7
76. Olympus C. Olympus Announces CE Approval for Three Cloud-based AI Medical Devices and Announces Plans for Launch of AI-Powered Endoscopy Ecosystem in 2025. Tokyo: Olympus Global News Release (2024).
77. Shamshad F, Khan S, Zamir SW, Khan MH, Hayat M, Khan FS, et al. Transformers in medical imaging: a survey. Med Image Anal. (2023) 88:102802. doi: 10.1016/j.media.2023.102802
78. Lodola I, D'Amico F, Danese S, Parigi TL. Artificial intelligence in inflammatory bowel disease endoscopy—a review of current evidence and a critical perspective on future challenges. Therap Adv Gastroenterol. (2025) 18:17562848251350896. doi: 10.1177/17562848251350896
79. Food US, Drug A. Marketing Submission Recommendations for a Predetermined Change Control Plan for Artificial Intelligence-Enabled Device Software Functions: Guidance for Industry and Food and Drug Administration Staff. Guidance Document. Silver Spring, MD: U.S. Food and Drug Administration (2025).
Keywords: artificial intelligence, deep learning, gastroenterology, inflammatory bowel disease, machine learning, pediatric gastroenterology, video capsule endoscopy
Citation: Stewart JW, Barth BA and Rojas I (2026) The utility of artificial intelligence in visualization of pediatric gastrointestinal mucosa. Front. Pediatr. 13:1739000. doi: 10.3389/fped.2025.1739000
Received: 4 November 2025; Revised: 8 December 2025;
Accepted: 15 December 2025;
Published: 12 January 2026.
Edited by:
Muhammad Khan, Nationwide Children’s Hospital, Columbus, United StatesReviewed by:
Brett J. Hoskins, Indiana University School of Medicine, Indianapolis, United StatesCopyright: © 2026 Stewart, Barth and Rojas. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
*Correspondence: Jeremy W. Stewart, SmVyZW15LnN0ZXdhcnRAdXRzb3V0aHdlc3Rlcm4uZWR1