Abstract

Study Design: Prospective-comparative study
Purpose: To assess the effectiveness of an AI-integrated smartphone fundus camera in detecting Referable Glaucoma (RG) at various glaucoma severity levels compared to specialist diagnoses.
Methods: One image per/eye/subject was taken with a validated, portable non-mydriatic fundus camera. The AI tool's diagnostic capability to identify RG was compared to specialist diagnoses (clinical assessment, SD-OCT and HVF (Glaucoma severity based on HAP criteria).
Results: In a sample of 213 subjects, the AI system achieved 92.02% accuracy, with a sensitivity of 91.4% (95% CI 85.9-95.2) and specificity of 94.1% (83.8-98.8) for RG. The AI's sensitivity for detecting mild, moderate, and advanced glaucoma was 86.9%, 90.3%, and 94.7%, respectively.
Conclusion: The Glaucoma AI tool demonstrated promising performance in identifying Referable Glaucoma, with higher accuracy in detecting advanced glaucoma, followed by moderate and early.