Artificial Intelligence for the Detection of Small Bowel Lesions and Its Potential Implications for Neoplasia Detection: A Scoping Review

Background/Objectives: Artificial intelligence (AI) has emerged as a promising tool to improve the detection and characterization of small bowel (SB) lesions through endoscopic imaging. This scoping review aimed to map and synthesize the available evidence regarding the diagnostic performance of AI systems applied to capsule endoscopy (CE) and device-assisted enteroscopy for the identification of SB lesions with potential malignant or premalignant relevance. Methods: A scoping review was conducted following the frameworks of Arksey and O’Malley, Levac, and Joanna Briggs Institute recommendations, and reported according to PRISMA-ScR guidelines. Searches were performed in PubMed, Scopus, and Embase. Original studies evaluating AI systems on SB endoscopic images or videos and reporting quantitative diagnostic metrics were included. Data extraction covered study characteristics, imaging modality, AI task, diagnostic performance, and methodological limitations. Results: A total of 13 studies were included, with a predominance of retrospective designs (9/13; 69.2%), followed by one prospective study (1/13; 7.7%), and one pilot study (1/13; 7.7%). Most studies originated from China (4/13; 30.8%, including an international collaboration with Denmark) and Japan (4/13; 30.8%). CE was the predominant imaging modality (9/13; 69.2%), followed by device-assisted enteroscopy (2/13; 15.4%) and upper gastrointestinal endoscopy (1/13; 7.7%). Automated lesion detection was the main application of artificial intelligence, reported in most studies (11/12; 91.7%), frequently combined with diagnostic classification tasks (8/12; 66.7%). Most models were based on convolutional neural networks, including architectures such as ResNet50 (29), Single Shot Multibox Detector (28), YOLOv5 (30), and nnU-Net (32). Diagnostic performance was consistently high, with sensitivities ranging from 81.2% to 98.6%, specificities from 88.6% to 99.8%, and area under the ROC curve values approaching 1.0 in several studies (22–30,33). Conclusion: AI improves the detection of SB lesions and significantly reduces endoscopic reading times, particularly in CE. Nevertheless, current evidence remains insufficient to support the use of AI as a screening tool for SB cancer, mainly due to the predominance of retrospective studies and the lack of robust prospective multicenter validation.