Mean absolute error was within 10 letters of actual best-corrected visual acuity.
Artificial intelligence (AI)-determined best-corrected visual acuity (BCVA) from remotely obtained fundus photographs is feasible for evaluating diabetic macular edema (DME), according to a study published online June 8 in JAMA Ophthalmology.
William Paul, from the Johns Hopkins University School of Medicine in Baltimore, and colleagues evaluated the potential application of AI techniques for estimating BCVA from fundus photographs. The analysis included 459 participants with DME (7,185 macular color fundus images of the study and fellow eyes).
The researchers found that the baseline BCVA score for the study eyes ranged from 73 to 24 letters (approximate Snellen equivalent 20/40 to 20/320). In the testing set (641 images), the mean absolute error (MAE) was 9.66, with one-third of the values within zero to five letters and 28 percent within six to 10 letters. For BCVA of >80 to ≤100 letters (20/10 to 20/25; 161 images) the MAE was 8.84 letters, and for BCVA >55 to ≤80 letters (20/32 to 20/80; 309 images), the MAE was 7.91 letters.
“This investigation suggests AI can estimate BCVA directly from fundus photographs in patients with DME, without refraction or subjective visual acuity measurements, often within one to two lines on an Early Treatment Diabetic Retinopathy Study chart, supporting this AI concept if additional improvements in estimates can be achieved,” the authors write.
Several authors disclosed ties to the pharmaceutical industry.
Abstract/Full Text (subscription or payment may be required)
