Eye AI? Aye-aye, FDA tells AEYE

The FDA has cleared an artificial intelligence program to help screen for the early signs of a major cause of blindness among people with diabetes. 

Developed by AEYE Health, the automated program scans images taken by retinal cameras during a standard annual eye exam to spot cases of diabetic retinopathy—a procedure that was granted reimbursement in the U.S. in September 2021 by the Centers for Medicare & Medicaid Services.

AEYE Health—not to be confused with AEye, which develops smart lidar sensors for cars and other applications—estimates that more than 35 million people with Type 1 or Type 2 diabetes in the U.S. are at risk of developing diabetic retinopathy, the result of damaged blood vessels feeding the retina.

More than 90% of vision loss caused by the disease can be treated with early detection, the company said in a release.

Even in the nascent sector of healthcare AI, programs for diabetic retinopathy are something of an elder statesman: The FDA’s first green light for an AI diagnostic tool was for the eye disease, with a 2018 de novo clearance for an algorithm developed by IDx, a company since renamed Digital Diagnostics.

AEYE’s offering requires only a single image per eye, and the company said it has delivered diagnostic results for over 99% of patients while rarely requiring pupil dilation—making it a practical choice for primary care offices. Clinical trials demonstrated a false-negative rate of 7% and an 8.6% rate of false positives. 

The program was granted an indication for use with the Topcon NW-400 desktop retinal camera, with total scan times taking about one minute, according to AEYE, which is also pursuing an FDA green light for the AI’s use in a more-portable camera.

The company is also developing a follow-on program that screens for glaucomatous optic neuropathy, where the loss of cells damages the optic nerve, and plans to launch pivotal trials in the near future.