Artificial Intelligence in Upper Gastrointestinal Diagnosis.

Abstract

Artificial intelligence (AI) has revolutionized upper gastrointestinal (GI) endoscopy by enhancing the detection, characterization, and management of GI diseases. In this review, we explore the transformative role of AI technologies, including machine learning and deep learning, in improving diagnostic accuracy and streamlining clinical workflows. AI systems such as convolutional neural networks have shown remarkable potential for identifying subtle lesions, assessing tumor margins, and reducing interobserver variability. By providing real-time decision-making support, AI minimizes unnecessary biopsies and improves patient outcomes. We also explore the applications of AI in detecting precancerous conditions such as Barrett's esophagus, atrophic gastritis, and gastric intestinal metaplasia, as well as its role in guiding therapy for early gastric cancer. Non-image-based AI tools such as Raman spectroscopy complement traditional imaging by offering molecular-level insights for real-time tissue characterization. Despite its promise, the adoption of AI in endoscopy faces challenges, including the need for robust validation, user-centric design, and targeted training for endoscopists. Concerns regarding overreliance and deskilling underscore the importance of balancing AI integration with the preservation of clinical expertise. Lastly, we examine the future of AI in upper GI diagnosis and how image-based and non-image-based AI technologies can be integrated to enable comprehensive diagnosis and personalized therapeutic planning. By addressing current limitations and fostering collaboration between clinicians and technologists, AI has the potential to redefine the standards of care for upper GI diagnosis and treatment.