Effectiveness of Artificial Intelligence in Screening Esophagogastroduodenoscopy

This is the authors' abstract. We don't add key points for this paper.

Background: This retrospective study investigated the usefulness of the EW10-EG01 artificial intelligence (AI) application for screening esophagogastroduodenoscopy (EGD). Methodology: A total of 7,655 subjects (4,863 men, 2,792 women; mean age 54.9±10.1 years) who underwent EGD during a medical checkup were enrolled in the study. The number of diagnosed upper gastrointestinal tumors was compared between EGD examinations performed with and without the AI system. Results: EGD examinations with and without the AI system were performed on 3,841 and 3,814 subjects, respectively. Biopsy procedures were more frequently performed, and examination time was longer in EGD with AI applications than in those without. Upper gastrointestinal tumors diagnosed by EGD with and without AI were 39 (1.02%) and 24 (0.63%), respectively (P = 0.062). There was a significant difference in the detection rate of esophageal and gastric tumors between EGD with (30, 0.78%) and without (14, 0.37%) the AI system (P = 0.017). When endoscopists were divided into three groups based on their experience with the EW10-EG01 application, higher detection rates of esophageal and gastric tumors were observed in each group when using EGD with AI. Conclusions: Usage of the EW10-EG01 AI system may be useful for screening EGD due to the increased esophageal and gastric tumor detection rate.