Artificial Intelligence–Assisted Endoscopy in Diagnosis of Gastrointestinal Tumors: A review of Systematic Reviews and Meta-Analyses

Introduction: Early and accurate detection of gastrointestinal (GI) tumors is crucial to reduce their associated high morbidity, mortality, and financial burden. Recent advancements in health technology have led to the development of AI-assisted endoscopy, which has been studied as potential tool to increase early and accurate GI tumors detection. This review aims to summarize systematic reviews and meta-analyses on AI-assisted endoscopy in diagnosing GI tumors. Methods: A comprehensive search was performed using PubMed/MEDLINE, Google Scholar, DOAJ, AJOL, and the Cochrane Library, along with manual searches. Systematic reviews and meta-analyses focusing on AI-assisted endoscopy for GI tumor diagnosis were included based on pre-specified criteria. Relevant data were extracted and summarized to present different metrics AI endoscopy in GI tumors diagnosis. Results: From 569 studies identified, 23 systematic reviews with meta-analyses were included. 6 focused on detection rates, and 17 examined diagnostic accuracy. AI-assisted endoscopy showed a significantly higher detection rate for GI tumors compared to conventional endoscopy. Additionally, AI-assisted endoscopy showed high diagnostic accuracy in detecting different types of GI tumors. However, variability in detection rate and accuracy results across different AI algorithms and studies was noted. Conclusion: AI-assisted endoscopy offers high diagnostic accuracy and improved tumor detection than conventional endoscopy. Meanwhile, rigorous evaluation of AI models is necessary to ensure clinical reliability. Additionally, AI should enhance clinical decision-making rather than replace clinicians. Addressing ethical concerns and further research are essential for optimizing AI in healthcare.