Our 3D deep-learning system performed well in both primary and external validations, suggesting that it could potentially be used for automated detection of glaucomatous optic neuropathy based on SDOCT volumes. Screening with the deep-learning system is much faster than conventional glaucoma screening methods (ie, by experienced specialists), can be done automatically, and does not require a large number of trained personnel on site. Further prospective studies are warranted to estimate the incremental cost-effectiveness of incorporating this artificial intelligence-based model for screening for glaucoma, both in the general population and among at-risk people.
To read more: https://www.thelancet.com/journals/landig/article/PIIS2589-7500(19)30085-8/fulltext