Usefulness of Automated Analysis System for Microvascular Blood Flow Rate Using Magnifying Endoscopy in Early Gastric Cancer: A Multicenter Retrospective Study (With Video)

Background and Aim Early gastric cancer (EGC) diagnosis after Helicobacter pylori eradication and differentiation from patchy redness (PR) remains challenging. We developed a novel automated analysis system for microvascular blood flow rate using magnifying endoscopy. This study evaluated its diagnostic performance in differentiating EGC from PR compared with endoscopists in a multicenter setting. Methods We analyzed 76 magnifying endoscopy videos, including 36 EGCs and 40 PRs recorded at our institution. Lesions were diagnosed as cancer (EGC) or noncancer (PR) by the analysis system, based on the cut‐off value calculated from the validation data set. Additionally, 30 endoscopists (15 experts and 15 trainees) from three institutions, all skilled in magnifying endoscopy diagnosis, participated. We compared the diagnostic performance between the analysis system and endoscopists. Results Diagnostic accuracy, sensitivity, specificity, positive predictive value, and negative predictive value (NPV) were 84.2%, 86.1%, 82.5%, 81.6%, and 86.8%, respectively, for the analysis system, and 84.2%, 80.2%, 87.3%, 83.2%, and 84.9%, respectively, for the endoscopists. Sensitivity and NPV of the analysis system tended to be higher than those of endoscopists. Additionally, the accuracy of the analysis system was comparable to that of all endoscopists and tended to be superior to that of trainees (84.2% [95% CI: 74.0%–91.6%] vs. 81.7% [95% CI: 79.4%–83.8%], respectively). Conclusions The automated analysis system for microvascular blood flow rate showed diagnostic performance equivalent to or superior to that of endoscopists skilled in magnifying endoscopy, suggesting its utility as a new diagnostic modality for differentiating EGC and PR in clinical practice.