A new application of artificial intelligence (AI), for detecting gonioscopic angle closure on 360˚ swept-source OCT (SS-OCT), demonstrated good diagnostic performance in a study from the Singapore Eye Research Institute. The validation and reliability analysis included an independent test of 39,936 SS-OCT scans from 312 phakic subjects (128 SS-OCT meridional scans per eye).

Patients included in the study were recruited from glaucoma clinics, were older than 50 and had no previous history of intraocular surgery (80% Chinese, 57.4% female). The researchers performed indentation gonioscopy and dark room SS-OCT. They defined gonioscopic angle as non-visibility of the posterior trabecular meshwork in ≥180˚ of the angle. Images were analyzed by the deep learning network based on an AI architecture for gonioscopic angle-closure detection.

The researchers reported a 20.2% prevalence of gonioscopic angle closure in the hospital-based study. From the 39,936 scans, the area under the curve (a statistical measure of predictive accuracy) of the deep learning algorithm was 0.85 for classifying gonioscopic angle closure with an optimal cut-off value of >35% of circumferential angle closure. The researchers also reported the algorithm had had 83% sensitivity and 87% specificity.

The researchers concluded that the algorithm demonstrated good diagnostic performance in a glaucoma clinic setting. “Such an algorithm, independent of the identification of the scleral spur, may be the foundation for a non-contact, fast and reproducible ‘automated gonioscopy’ in the future,” they wrote in their paper.