AI retinal screening adopted at West Virginia hospital as study shows device‑dependent accuracy

WVU Medicine Uniontown Hospital has implemented an FDA‑approved autonomous AI retinal screening system for diabetic retinopathy. The eight‑minute, no‑dilation exam captures an image that is analyzed instantly by AI, allowing diagnosis, referral and treatment planning in a single visit. Since adoption, screening rates rose from 26 % to 73 %, streamlining appointments, reducing costs and improving access for the region’s high‑Medicaid, rural population. Hospital officials noted that the technology is “quality aligned, revenue positive, and truly helps address equity access challenges” and that early detection costs far less than treatment injections.

A prospective primary‑care study in India evaluated three AI algorithms on two fundus‑camera models. Sensitivity reached 97.5 % on one camera but specificity fell to 62.7 %, while another algorithm achieved a more balanced 95.7 % specificity with 80 % sensitivity. The findings underscore that AI performance varies with specific imaging devices and patient factors, prompting calls for device‑specific validation before wide deployment.