Diabetes Blog – Happy-Medium.net

It is kinda cool, at least for starters.

Wednesday, the U.S. Food and Drug Administration (FDA) approved marketing of a device designed to use artificial intelligence to detect retinopathy in people living with diabetes.

There are several things to cover here. First of all, the FDA announcement is that it “permits marketing” of the device. In other words, it can be sold to practices as an artificial intelligence diagnostic tool. There was a fair amount of testing and back-and-forth in the development and rollout of this device, and I’ll get to that in a minute.

Simply put, this is a way for non-eye doctors to use a tool (actually, two tools) to take a photo, upload the photo, and get a retinopathy or not retinopathy diagnosis. It’s not designed to avoid the more detailed examination an ophthalmologist can perform. But it’s a way to perhaps get an early diagnosis of a potential issue with your eyes.

The tools used by the doctor (presumably, a primary care physician, or GP) are 1) The Topcon NW400 retinal camera. The doctor will take a photo with the retinal camera, then upload it to the cloud, where it will be analyzed by 2) The newly-marketing-permitted IDx-DR software to determine one of two outcomes (from the FDA press release… see link below):
(1) “more than mild diabetic retinopathy detected: refer to an eye care professional” or
(2) “negative for more than mild diabetic retinopathy; rescreen in 12 months.”

So, where’s the “artificial intelligence” angle here? I’m glad you asked.

This is the first device (the software is considered a device in FDA parlance) allowed to be marketed to provide a decision without getting a clinician involved in interpreting the results. In other words, the software, not a human, tells you whether there’s an issue or not.

How did the FDA arrive at the level of confidence needed to arrive at this decision? There were 900 diabetes patients in the clinical study, in ten different locations. The study revealed that the software made the right decision on “more than mild diabetic retinopathy” 87.4 percent of the time, and “not more than mild diabetic retinopathy” 89.5 percent of the time. That may not seem like a lot, but to this observer, that sounds pretty good for first generation software.

One of the other things I’m encouraged by is the way this rollout was handled through FDA’s De Novo premarket review pathway, designed to help new devices that are seen as being low to moderate risk get marketing approval. Through this, FDA granted the software Breakthrough Device designation, which provided a framework for greater interaction between developers and the FDA on guidance, testing, and documentation, so the developers could work more efficiently on FDA’s concerns, and the FDA could more effectively work on review.

Hey, look… I don’t know if this software and the retinal camera needed to make it work are going anywhere. But I think this story is important to pay attention to for the way it was handled by the developers and the FDA, particularly FDA’s Center for Devices and Radiological Health; and the fact that this is software that removes the human element from at least the initial diagnosis process.

How you feel about such a development is your choice.

Read the full FDA press release HERE.

Find out about the Expedited Access Pathway Program for Breakthrough Devices HERE.