AI Software Detects Diabetic Retinopathy Without Physician Involvement

Posted on April 27, 2018 I Written By

Anne Zieger is veteran healthcare consultant and analyst with 20 years of industry experience. Zieger formerly served as editor-in-chief of FierceHealthcare.com and her commentaries have appeared in dozens of international business publications, including Forbes, Business Week and Information Week. She has also contributed content to hundreds of healthcare and health IT organizations, including several Fortune 500 companies. Contact her at @ziegerhealth on Twitter or visit her site at Zieger Healthcare.

The FDA has approved parent company IDx to market IDx-DR, the first AI technology which can independently detect diabetic retinopathy. The software can make basic recommendations without any physician involvement.

Before approving the software, the FDA reviewed data from a clinical study of 900 patients with diabetes across 10 primary care sites. IDx-DR accurately identified the presence of diabetic retinopathy 87.4% of the time and accurately identified those without the disease 89.5% of the time. In other words, it’s not perfect but it’s clearly pretty close.

To use IDx-DR, providers upload digital images of a diabetic patient’s eyes taken with a retinal camera to the IDx cloud server. Once the image reaches the server, IDx-DR uses an AI algorithm to analyze the images, then tells the user whether the user has anything more than mild retinopathy.

If it finds significant retinopathy, the software suggests referring the patient to an eye care specialist for an in-depth diagnostic visit. On the other hand, if the software doesn’t detect retinopathy, it recommends a standard rescreen in 12 months.

Apparently, this is the first time the FDA has allowed a company to sell a device which screens and diagnoses patients without involving a specialist. We can expect further AI approvals by the FDA in the future, according to Commissioner Scott Gottlieb, MD. “Artificial Intelligence and Machine Learning hold enormous promise for the future of medicine,” Gottlieb tweeted. “The FDA is taking steps to promote innovation and support the use of artificial intelligence-based medical devices.”

The question this announcement must raise in the minds of some readers is “How far will this go?” Both for personal and clinical reasons, doctors are likely to worry about this sort of development. After all, putting aside any impact it may have on their career, they may be concerned that patient will get short-changed.

They probably don’t need to worry, though. According to an article in the MIT Technology Review, a recent research project done by Google Cloud suggests that AI won’t be replacing doctors anytime soon.

Jia Li, who leads research and development at Google Cloud, told a conference audience that while applying AI to radiology imaging might be a useful tool, it can automate only a small part of radiologists’ work. All it will be able to do is help doctors make better judgments and make the process more efficient, Li told conference attendees.

In other words, it seems likely that for the foreseeable future, tools like IDx-DR and its cousins will help doctors automate tasks they didn’t want to do anyway. With any luck, using them will both save time and improve diagnoses. Not at all scary, right?