News Science Technology

AI Used in Specialty-Level, Primary Care Disease Diagnosis

AI Used in Specialty-Level, Primary Care Disease Diagnosis

An ophthalmologist at the University of Iowa has designed a system to safely identify diabetic retinopathy without needing human interpretation of results. The system, which uses artificial intelligence (AI), has earned FDA (Food and Drug Administration) authorization following a clinical trial in primary care offices. Study results, first published online in Nature Digital Medicine in August, offers a look at the data that led to the FDA approval for the IDx-DR system — the first ever health device that utilizes AI for detecting the disease autonomously.

IDx-DR Clinical Trial

The clinical trial compared IDx-DR performance with the world’s gold standard for diagnosing diabetic retinopathy, the Wisconsin Fundus Photograph Reading Center (FPRC). The disease is the foremost cause of sight loss in adults and one of the greatest impediments for 30.3 million Americans suffering from diabetes.

The IDx-DR system surpassed all superiority endpoints detailed in sensitivity thus far: the capacity to accurately identify patients with the disease; the ability to precisely categorize people as free of the disease; and the capability to generate quality retinal images and establish the seriousness of the disease. The results of the study show the safety of the device and the potential to bring specialty diagnostics to primary care at lower costs and with more access.


Early Detection Benefits

Every year, over 24,000 people in the United States go blind because of diabetic retinopathy. Early discovery and treatment can lower the risk of sight loss by 95%, however, fewer than 50% of diabetics book systematic eye care tests with a specialist.

The study examined 900 adult diabetics with no diabetic retinopathy history at 10 primary care locations throughout the U.S. Four quality images of patients’ retina were acquired using a robotic camera aided by an AI operator. After completion, the diagnostic AI took 20 seconds to generate a clinical diagnosis as would clinical experts. According to lead investigator for the study, Michael Abràmoff, MD, Ph.D., it was important to test the device in a real life setting, by clinicians who would use it, and to compare it to the FPRC.

Results from using the AI system compared to FPRC imaging operated without AI were noteworthy. After FDA approval, the IDx-DR screening system was first used at the University of Iowa’s Diabetes and Endocrinology Center. Abràmoff said the aim is to also use AI systems in detecting other diseases early, such as macular degeneration and glaucoma. The aim is to bring specialty diagnostics into retail and primary care where these patients are.