Background: Wireless capsule endoscopy is widely used for diagnosing gastrointestinal diseases, but manual interpretation of capsule videos is time-consuming and can vary between clinicians. Artificial intelligence has been increasingly studied to support capsule analysis and reduce clinical workload. This systematic literature review and meta-analysis summarizes current evidence on artificial intelligence methods applied to wireless capsule endoscopy, with a focus on diagnostic performance, validation strategies, and clinical readiness. Methods: A systematic search was conducted in PubMed, Scopus, Embase, Web of Science, and Google Scholar. Original journal articles were included based on predefined eligibility criteria. The reviewed studies addressed multiple artificial intelligence tasks, including detection, classification, segmentation, and localization of gastrointestinal abnormalities. Results: A total of 72 studies were included. Meta-analysis using random effects models showed high pooled diagnostic performance across clinical indications and gastrointestinal tract locations, with the strongest results reported for bleeding and vascular lesions and more variable performance for inflammatory bowel disease and mixed abnormality categories. The review also identified important clinical and technical barriers that may limit reliability and slow clinical adoption. These included limited external validation, small patient cohorts, retrospective study designs, and inconsistent reporting and evaluation practices. Conclusions: Artificial intelligence methods show strong potential to support wireless capsule endoscopy interpretation. Based on the findings, we propose practical recommendations to improve study design and validation. If these recommendations are applied, future studies may report more robust and reliable results, supporting better translation into clinical workflows.
Sahafi et al. (Thu,) studied this question.