Researchers are encouraging the greater use of AI to support eye tests and treatment after a study demonstrated its accuracy in evaluating fundus photographs.
A release supporting the US study, which was conducted at Johns Hopkins University School of Medicine and Hospital in Baltimore and reported in the Journal of the American Medical Association, said researchers wanted to know if an AI algorithm could estimate best-corrected visual acuity (VA) from fundus scans.
“Determining spectacle-corrected VA is essential when managing many ophthalmic diseases,” it said.
“If artificial intelligence (AI) evaluations of macular images estimated this VA from a fundus image, AI might provide spectacle-corrected VA without technician costs, reduce visit time, and facilitate home monitoring of VA from fundus images obtained outside of the clinic.”
The study was a retrospective cross-sectional evaluation of deidentified fundus photographs matched to spectacle-corrected VA determined by technicians.
They involved eye charts of patients with a history of diabetic macular edema (DME) based on optical coherence tomography and at least two visits within one to six months of each other at a university-based clinic between January 2014 and December 2022.
“In this cross-sectional study of 282 eyes with spectacle-corrected VA of 20/80 or better, with or without centre-involved DME among 141 patients, the mean absolute error of AI-estimated VA was about one line on a standardised eye chart from the actual spectacle-corrected VA, and estimations were within two lines in about 80% of estimated visual acuities.”
For eyes with less than two lines of change in VA with spectacle correction between the two visits, AI was able to estimate such a change in 230 of 243 eyes (95%).
The researchers said the results supported the use of AI evaluation of fundus photographs to determine spectacle-corrected VA among patients with DME globally, “beyond ophthalmology offices”.
And they said there was potential for “eliminating the need to refract a patient and then ask the patient to read a vision chart to determine distance VA”.
“These findings might facilitate determining VA with spectacle correction for monitoring in a medical clinic evaluating patients with diabetes where fundus images can be obtained without the need of technicians or space or equipment to measure VA, or potentially from smartphones that can obtain such images.
“If home monitoring of VA from a fundus image obtained on a smartphone can be perfected, AI algorithms evaluating these images also may assist with determining when patients with diabetes may need referral to ophthalmologists for treatment of DME or when patients receiving anti-VEGF treatments for DME by ophthalmologists can extend follow-up intervals when VA is stable or resume treatment when VA worsens.”
The researchers believed if this was used for patients with retinal diseases in which frequent or home monitoring was important, it could potentially improve management and access to care around the world.
The report said further investigations were planned about the application of AI.
More reading:
Artificial intelligence: eyecare’s friend of foe?
Optometrists back Aussie-developed AI system
