Abstract
Background: Endoscopic gastric cancer diagnosis has evolved from conventional endoscopy (white light) to combined use of image-enhanced endoscopy (IEE) and artificial intelligence (AI)-based lesion detection. The accuracy of gastric cancer screenings conducted by endoscopic specialists has the potential to improve by using IEE and detection AI. Therefore, we evaluated the efficacy using process indicators over a 1-year period.
Subjects and Methods: Between April 2023 and March 2024, 1,223 individuals underwent upper gastrointestinal endoscopy as part of a health check-up. Endoscopists performed examinations using FUJIFILM’s ELUXEO 7000 endoscopic system, utilizing the AI function “CADEYE” for IEE observation and real-time lesion detection with the AI software EW10-EG01. All examinations were conducted using a transnasal endoscope (EG-840N) by endoscopists certified by the Japan Gastroenterological Endoscopy Society.
Results: Sixty of the 1,223 subjects required further examination, resulting in a recall rate of 4.91%. Biopsies were performed for 38 subjects with gastric lesions. Four cases of early gastric cancer were identified, with a gastric cancer detection rate of 0.33% and a positive predictive value of 6.67%.
Conclusions: Since the necessity of detailed examinations for lesions detected by AI was determined by certified endoscopists, the recall rate decreased while false positives might have been minimized. On the other hand, the combination with IEE also contributed to reducing false negatives, and the improved accuracy suggested the potential to reduce disadvantages for patients undergoing screening.
