For the first time, US health regulators have given the green light to sell a medical device that bypasses doctors and uses artificial intelligence (AI) software to detect an eye disease affecting more than 30 million Americans living with diabetes.
Dubbed the IDx-DR, the device (no doubt named after its parent company IDx LLC in Iowa) comes in the form of a software program. Any doctor, not just an eye specialist, takes images of a person’s eye with a special retinal camera and then uploads digital photos to a cloud server where the software is installed. IDx-DR indicates whether or not the image is of high enough quality to get a result and, if it is, the program then uses an AI algorithm to analyze the images.
It produces two results. The first will refer a patient to an eye care professional if a “more than mild diabetic retinopathy” is detected. If the results come back negative, IDx-DR refers the patient to get another screening in 12 months.
The FDA evaluated data from a clinical study of retinal images from 900 patients with diabetes at 10 facilities in order to test IDx-DR’s accuracy. They found in mild cases it was correct 87.4 percent of the time, and 89.5 percent in “more than mild” cases.
A doctor isn’t needed to interpret the results, which means any health care provider can recommend next steps for the patient.
IDx-DR only detects diabetic retinopathy, which affects the eye when too much blood sugar damages blood vessels in the back of it. It’s the most common vision complication for people with diabetes, with around 200,000 cases in the US each year. But it doesn’t work for everyone: People with a history of laser treatment, surgery or injections in the eye, or a list of other conditions (persistent vision loss, blurred vision, floaters, to name a few) should not be screened using IDx-DR.
The new device joins the ranks of other AI technology moving into the medical field, like this smart software that can diagnose prostate cancer or this other algorithm trained to recognize conditions like age-related vision loss as well as diabetic retinopathy.
While a recent study suggests consumers are more comfortable using AI in their healthcare than in retail or with their financial institutions, the use of AI raises a few ethical questions, such as who becomes responsible for an improper diagnosis, how to protect privacy or prevent hacking, and even the threat of bioterrorism through nanotechnology.