Artificial Intelligence and Deep Learning for Upper Gastrointestinal Neoplasia.

Area of Science:

• Gastroenterology
• Medical Imaging
• Artificial Intelligence

Background:

• Upper gastrointestinal (GI) neoplasia cause 1.5 million deaths annually, representing 35% of GI cancers.
• Diagnostic upper GI endoscopy has a significant miss rate for neoplastic lesions due to recognition failure or navigation issues.

Purpose of the Study:

• To review the development and clinical integration of artificial intelligence (AI) for detecting and characterizing upper GI neoplasia.
• To guide clinicians through the steps of AI development for improved endoscopic diagnostics.

Main Methods:

• Review of AI applications, including computer-aided detection (CADe) and computer-aided diagnosis (CADx), for upper GI neoplasia.
• Analysis of CADe performance for esophageal and gastric cancers.

Main Results:

• Stand-alone CADe demonstrated promising accuracy and sensitivity (83%-93%) for esophageal squamous cell neoplasia, Barrett's esophagus-related neoplasia, and gastric cancer.
• Clinical integration challenges include potential biases, human-AI interaction, and managing false-positive results.

Conclusions:

• AI holds potential to enhance the accuracy of upper GI endoscopy by reducing missed lesions.
• Successful clinical adoption requires careful consideration of algorithm development, validation, and real-world implementation strategies.