Digital Gonioscopy Based on Three-dimensional Anterior-Segment OCT: An International Multicenter Study.

Area of Science:

• Ophthalmology
• Medical Imaging
• Artificial Intelligence

Background:

• Primary angle-closure glaucoma (PACG) is a leading cause of blindness globally.
• Early detection of narrow angles and peripheral anterior synechiae (PAS) is crucial for preventing vision loss.
• Current gonioscopy methods can be subjective and time-consuming.

Purpose of the Study:

• To develop and evaluate a 3D deep-learning-based automated digital gonioscopy system (DGS).
• To assess the DGS's performance in detecting narrow iridocorneal angles (Task I) and PAS (Task II).
• To determine the DGS's utility in identifying eyes with suspected PACG.

Main Methods:

• A multicenter study utilizing 1.112 million OCT scans from 2294 patients.
• Development of a 3D deep-learning model for automated gonioscopy.
• Evaluation of the DGS against gonioscopic records for narrow angles and PAS detection.

Main Results:

• The DGS achieved high diagnostic performance in detecting narrow angles (AUC 0.943) and PAS (AUC 0.902).
• External testing demonstrated strong sensitivity and specificity for both tasks.
• The system showed robust performance across different evaluation levels (quadrant and clock-hour).

Conclusions:

• The 3D DGS is effective in identifying eyes with suspected PACG.
• The automated system shows significant potential for screening individuals at high risk.
• Widespread adoption in primary eye care could improve early detection and management of PACG.