Approving AI for medical use: 7 things to know

The FDA grants approval for drugs and medical devices, but as artificial intelligence continues to grow in healthcare, the agency also is tasked with ensuring computer algorithms are both safe and effective, NPR reports.

In 2018, the FDA approved IDx-DR, an AI-powered software that can identify whether a patient's vision is beginning to erode due to diabetic retinopathy. The agency said it was the first time it approved a device that can make a screening decision without a clinician's interpretation, according to the report.

Seven things to know:

1. Medical software often undergoes a less "rigorous pathway," for FDA approval than medical devices, NPR reports.

2. IDx-DR CEO and Founder Michael Abramoff, MD, PhD, told NPR that when he started working with the FDA for software approval "they were uncomfortable," and "so we started working together on how can we prove that this can be safe."

3. To showcase the software's effectiveness, and its ability to work with a diverse population, Dr. Abramoff, an ophthalmologist at University of Iowa Health Care in Iowa City, and the FDA tested it on 900 people at 10 different primary clinics.

4. IDx-DR test results surpassed the FDA's requirement, which indicated that "eye disease needed to be correct at least 85 percent of the time, while those finding no significant eye damage needed to be correct at least 82.5 percent of the time," according to NPR.

"It's better than me, and I'm a very experienced retinal specialist," Dr. Abramoff said.

5. While the FDA has experience approving software that is part of medical devices, new algorithms create challenges, such as being wary of approval of algorithms that are based on a particular patient population, according to the report.

6. The FDA is testing a new system called precertification to approving algorithms. The system shifts the agency's focus to continued monitoring of companies' processes to develop their products, instead of examining every new product update or tweak.

7. It's important for the FDA to continue monitoring algorithms once they're deployed in medical practice because AI is constantly evolving, said Christina Silcox, PhD, a researcher at Durham, N.C.-based Duke University-Margolis Center for Health Policy.

"Right now, we might see an update to a medical device every 18 months," Dr. Silcox said. "In software you might expect to see one every two weeks or every month."

More articles on artificial intelligence:
Google quietly disbands another AI advisory board
Partners HealthCare pledges to make AI tools accessible to clinicians, researchers
6 trends in blockchain and AI

© Copyright ASC COMMUNICATIONS 2019. Interested in LINKING to or REPRINTING this content? View our policies by clicking here.


Top 40 Articles from the Past 6 Months