Research Shows High Accuracy in Artificial Intelligence Screening for Age-Related Macular Degeneration

An effective tool to assist primary care clinics in referring patients with age-related macular degeneration (AMD) to ophthalmologists has the potential to increase access to screening for individuals at high risk, and expand opportunity for preventive care.

Research presented at the American Academy of Ophthalmology 2020 details results of a trial analyzing the proficiency of artificial intelligence (AI) to categorize fundus images into 2 sets — photographs displaying referable intermediate or advanced AMD, and photographs showing healthy retinas or early AMD. The algorithm classified 503 fundus images of 266 subjects, 50 years of age or more, who were tested at the New York Eye and Ear Infirmary of Mount Sinai affiliated eye clinics. A team of 3 ophthalmologists also evaluated and graded the images, finding 161 eyes referable and 342 not referable.

The deep learning AI system achieved 88.7% accuracy to identify subjects with referable AMD. It reached a sensitivity score of 86.3%, indicating true positives; and specificity rate of 89.8%, suggesting true negatives. 

In machine learning, algorithms are generated from data input, allowing a computer to make calculations, whereas deep learning takes the next step, with multiple layers of additional input that continually improve predictions. The prospective trial’s AI system was initially developed with the data from the Age-Related Eye Disease Study (AREDS), comprising 4757 participants.

AI is now also being examined to classify images for stages in retinopathy of prematurity (ROP).

Reference

Alauddin, S, Otero-Marquez O, Smith RT. An automated ai-based telemedicine platform for screening patients with referable amd: a prospective trial. Presented at: American Academy of Ophthalmology 2020 Annual Meeting; November 13-15, 2020. Session: PA057