Computers aren’t quite ready to see patients completely independent of a human doctor, but they can certainly help diagnose. Think of them as bionic residents. They could, potentially, spot some disease states that human doctors miss. It couldn’t hurt to have a second set of eyes on a patient, even if they are electric eyes.

According to a California-based research team, deep learning—a subset of artificial intelligence that can learn unsupervised and without pre-scripted prompts—can effectively detect gonioscopic angle closure and primary angle closure disease (PACD) based on automated analysis of anterior segment OCT images. The investigators suggest the methods developed could be used to automate clinical evaluations of the anterior chamber angle, a capability that could open up a whole new avenue of care, especially in underserved and high-risk populations.

The researchers used data from the Chinese-American Eye Study out of Los Angeles to train the artificial intelligence. It analyzed 4,036 OCT images of 791 subjects, 640 of which were designated the test data set while the remaining 3,396 images were used as a cross-validation data set. The researchers paired the deep-learning outcomes with the corresponding gonioscopy grades in addition to a variety of other data points.

Xu B, Chiang M, Chaudhary S, et al. Deep learning classifiers for automated detection of gonioscopic angle closure based on anterior segment OCT images. Am J Ophthalmol. August 22, 2019. [Epub ahead of print].