CLINICAL APPLICATIONS AND FUTURE PERSPECTIVES OF AI-ASSISTED ENDOSCOPY

Abstract

The development of endoscopic AI advanced rapidly during the third wave of artificial intelligence due to the progress in deep learning technology. In 2018, the computer-aided diagnosis of endocytoscopy was approved as the first AI-equipped medical device in Japan. Since 2024, additional reimbursements have been introduced. Endoscopic AI is broadly classified into lesion detection and differential diagnosis.

In colonoscopy, meta-analyses have shown that computer-aided detection improves adenoma detection rates by 8-10% and significantly reduces miss rates. Prospective studies have also been conducted on gastric and esophageal squamous cell carcinomas.

However, in clinical practice, issues such as false positives, physician dependence on AI (deskilling), and optimization of human-AI interactions have become apparent. Although the clinical implementation of endoscopic AI is steadily progressing, further prospective studies are required to establish its true clinical utility.