CURRENT STATUS OF AI-ASSISTED DIAGNOSIS IN EARLY GASTRIC CANCER

Abstract

Gastric cancer（GC）remains a major global health burden, and early detection and early treatment are essential for improving patient outcomes. Deep-learning artificial intelligence（AI）has accelerated the development of endoscopy-based diagnostic support systems for early GC. This review summarizes the current evidence on AI-assisted diagnosis of early GC, focusing on computer-aided detection, computer-aided diagnosis, invasion-depth prediction, delineation of tumor margins, and diagnosis of precancerous lesions. AI systems are useful for supporting non-expert endoscopists, with some performance comparable to that of experts. For lesion detection, randomized controlled trials have demonstrated reduced miss rates with AI assistance. In Japan, an AI-based upper gastrointestinal lesion detection system received regulatory approval in 2022, an early GC-specific detection AI in 2023, and an invasion-depth prediction AI in 2024. and clinical implementation is now progressing. However, several issues must be addressed before broader social implementation, including insufficient external validation, uncertain cost-effectiveness, reimbursement challenges, and questions regarding explainability and responsibility sharing.