An artificial intelligence algorithm is highly accurate for detecting endoscopic features of eosinophilic esophagitis

Area of Science:

• Gastroenterology
• Medical Imaging
• Artificial Intelligence

Background:

• Endoscopic features of eosinophilic esophagitis (EoE) can be subtle and missed during routine examinations.
• Objective assessment of EoE endoscopic findings is crucial for accurate diagnosis and management.

Purpose of the Study:

• To develop and evaluate an AI algorithm for detecting and quantifying endoscopic features of EoE using white light images.
• To compare the AI algorithm's performance against human endoscopists with varying experience levels.

Main Methods:

• An AI algorithm (AI-EoE) was trained on endoscopic images to differentiate EoE from normal esophagus.
• A second algorithm (AI-EoE-EREFS) incorporated the EoE Endoscopic Reference Score (EREFS) criteria for enhanced feature quantification.
• AI algorithms were validated on an external dataset and compared with endoscopist performance.

Main Results:

• The AI-EoE algorithm achieved high performance: 0.93 sensitivity, specificity, accuracy, and 0.986 AUC.
• The AI-EoE-EREFS algorithm showed improved performance: 0.96 sensitivity, 0.94 specificity, 0.95 accuracy, and 0.992 AUC.
• Both AI algorithms significantly outperformed beginner and senior endoscopists in detecting EoE features.

Conclusions:

• AI algorithms can be effectively trained to detect and quantify endoscopic features of EoE with high accuracy.
• Integrating EREFS criteria into the AI algorithm further improved diagnostic performance.
• AI demonstrates potential as a valuable tool to aid endoscopists in diagnosing EoE, especially for those with less experience.