Commercial risk will be a critical catalyst of progress – it’s complicated, but is it possible? We think so.


April 13, 2018

The future is here: FDA approves first AI device that can detect disease without a specialist

Daily Briefing

    FDA on Wednesday announced that it has approved for sale the first medical device that uses artificial intelligence (AI) software to detect certain vision loss in diabetic patients.

    April 19 webconference: How Northside leverages IT to optimize quality reporting

    About the device

    The IDx-DR device was developed by Iowa-based IDx, a company that develops AI tools for health care, to detect diabetic retinopathy—a condition in which high blood sugar levels damage blood vessels in the retina damaging eyesight.

    The device uses AI software to analyze digital images of the eye taken with a special retinal camera. A physician takes and uploads the images to a cloud server containing the software. If the image is readable, the AI software will tell the physician if the patient has tested positive or negative for "more than mild diabetic retinopathy." Patients who test positive are referred to an eye care professional, while those who test negative should be rescreened in 12 months.

    In a clinical trial, the device was able to correctly detect more than mild diabetic retinopathy in 87.4% of cases, and correctly determined individuals did not have a more than mild condition 89.5% of the time.

    According to Medscape, the device is the first FDA has approved that can provide a screening decision that does not require a clinician to interpret the image or results. That distinction means providers who are not normally familiar with eye care, such as primary care physicians who are treating patients with diabetes, can use the device.

    Malvina Eydelman—who oversees FDA's division of ophthalmic, and ear, nose and throat devices—noted that early detection is key for diabetic retinopathy, but she said, "Many patients with diabetes are not adequately screened for diabetic retinopathy since about 50% of them do not see their eye doctor on a yearly basis."

    FDA looks to streamline approval process for AI devices

    The device approval came the same day FDA Commissioner Scott Gottlieb proposed a new pathway for moderate risk devices, such as AI and machine learning, that leverage new technologies to meet premarket notification, or 510(k), requirements.

    Most manufacturers today rely on comparative testing that shows their new device is as safe or effective as an equivalent legally marketed device. But Gottlieb said, in some cases, the existing products are old and do not closely reflect emerging technologies.

    Therefore, FDA is proposing allowing manufacturers of certain moderate risk devices "to use objective performance criteria" when comparing their products against existing devices. According to FDA, "Under this approach, if a predicate device meets certain levels of performance on characteristics relevant to its safety and effectiveness, and a new device meets or exceeds those same levels of performance on the same characteristics, the FDA could find the new device to be as safe and effective as the predicate."

    Gottlieb said, "[T]his new pathway offers a more efficient and less burdensome option to demonstrate that certain new devices are substantially equivalent to ones already on the market" (Berkrot, Reuters, 4/11; Brooks, Medscape, 4/11; FDA release, 4/11; Gottlieb tweet, 4/11).

    Just updated: Your IT cheat sheets on Artificial Intelligence, cybersecurity, and more

    Download our cheat sheets so you can keep track of the fast-changing technologies and capitalize on opportunities for IT-powered innovation. Check out our guides for these topics and more:

    Explore all 14 Cheat Sheets

    Have a Question?


    Ask our experts a question on any topic in health care by visiting our member portal, AskAdvisory.