Eyenuk EyeArt outperforms dilated exam
- Eyenuk’s EyeArt system was shown in a multicenter study to catch diabetic retinopathy far more often than ophthalmologists’ routine dilated eye exams. - On the same patient cohort, EyeArt’s sensitivity for more-than-mild diabetic retinopathy was 96.4%, versus 27.7% for dilated ophthalmoscopy by clinicians. - EyeArt is FDA-cleared for both referable and vision-threatening disease, but U.S. use of AI screening remains limited. (fda.gov)
Diabetic retinopathy is damage to the retina’s tiny blood vessels caused by diabetes, and it can progress before a patient notices blurred vision or blind spots. Screening is meant to catch those changes early enough for referral and treatment. (fda.gov) (jamanetwork.com) Eyenuk’s EyeArt system was evaluated against ophthalmologists’ dilated exams in a multicenter study, and the artificial intelligence system detected more cases of diabetic retinopathy on the same cohort. Ophthalmology Times reported the study’s headline numbers from that comparison. (ophthalmologytimes.com) (ophthalmologyscience.org) For more-than-mild diabetic retinopathy, EyeArt posted 96.4% sensitivity, while ophthalmologists’ dilated exams reached 27.7% sensitivity in that study. The tradeoff ran the other way on specificity: ophthalmologists were at 99.6%, versus 88.4% for EyeArt. (ophthalmologytimes.com) That matters because screening tools are usually tuned to miss as few disease cases as possible, then send positive patients for a full eye workup. In plain terms, EyeArt is built as a front-door triage test, not as a replacement for retina specialists treating advanced disease. (eyenuk.com) (jamanetwork.com) EyeArt’s FDA labeling already covers autonomous detection of both more-than-mild diabetic retinopathy and vision-threatening diabetic retinopathy in adults with diabetes who have not previously been diagnosed with the disease. FDA clearance documents from August 2020 and June 2023 list those indications and the camera models cleared for use. (fda.gov 1) (fda.gov 2) Eyenuk says the system returns a screening report in less than 60 seconds after retinal images are submitted, using two color fundus photos per eye. The company says the software is cleared for Canon CR-2 AF, Canon CR-2 Plus AF, and Topcon NW400 cameras. (eyenuk.com) The longer record behind the product is broader than this single comparison study. A 2026 systematic review and meta-analysis in *Ophthalmologica* pooled 17 studies and reported an area under the summary receiver operating characteristic curve of 0.932, a measure that reflects high overall diagnostic accuracy. (karger.com) Adoption is still the bottleneck. A 2024 report on U.S. use of artificial intelligence diabetic retinopathy screening found that less than 5% of diabetic patients in the study received ophthalmic imaging from 2019 to 2023, and only a small fraction underwent AI-based screening. (ophthalmologytimes.com) (jamanetwork.com) So the practical question is not whether EyeArt can flag disease on retinal photos. It is whether primary care clinics, diabetes centers, and eye practices build the referral, repeat-imaging, and follow-up systems that turn a positive screen into treatment before vision is lost. (eyenuk.com) (ophthalmologytimes.com)
