Donate to Science & Enterprise

S&E on Mastodon

S&E on LinkedIn

S&E on Flipboard

Please share Science & Enterprise

FDA Clears A.I. Software Detecting Diabetic Eye Disease

Fundus photograph

Fundus photograph of a healthy right eye (OptometrusPrime, wikimedia Commons)

11 April 2018. The Food and Drug Administration authorized a system for detecting diabetic retinopathy, a condition threatening the vision of people with diabetes, that uses artificial intelligence to analyze images of patients’ retinas. The system known as IDx-DR is a product of the company IDx LLC, in Coralville, Iowa, that developed the technology with University of Iowa.

High levels of blood sugar can cause serious complications in people with diabetes, including vision problems. Among these issues is damage to the fine blood vessels in the retina, located in the back of the eye that detects and converts light to signals sent through the optic nerve to the brain. Diabetic retinopathy is the name given to this condition that can range from mild leakage in the eye to swelling and distortion of the blood vessels, and proliferation of new blood vessels to compensate for the damage.

One consequence of this damage is swelling in the macula, the part of the retina that provides the sharpest vision, a disorder known as diabetic macular edema. Other outcomes include cataracts, glaucoma, or permanent vision loss if the retina detaches from the eye. Centers for Disease Control and Prevention says diabetes is the leading cause of new blindness among adults in the U.S. age 20 to 74. CDC estimates 4.1 million people have diabetic retinopathy, including 899,000 individuals with advanced stages of the condition threatening their vision.

The IDx-DR system analyzes images of patients’ retinas taken by a specialized fundus camera that provides high-resolution color medical images of the eyes. The images are stored on a local system and uploaded to a cloud-based server where IDx-DR software analyzes the images, and according to the company returns results in one minute or less. The analysis is based on algorithms the company says combine image analysis and deep learning that first assess image quality, then inspect for lesions and location of irregularities in the retina, resulting in an assessment of diabetic retinopathy. Assuming images of sufficient quality, IDx-DR either returns a positive result for moderate or more severe diabetic retinopathy, or a mild case of the condition, calling for a retest in one year.

FDA’s clearance is based in part on results from a clinical trial where retinal images from 900 individuals with diabetes were analyzed by IDx-DR software, compared to the original professional diagnosis. The results show the system could accurately detect more than mild cases of diabetic retinopathy 87 percent of the time, and identify cases where individuals had mild or no diabetic retinopathy 90 percent of the time. FDA’s authorization applies only to detection of more than mild cases diabetic retinopathy, and not to other eye disorders. The agency previously gave the system a Breakthrough Device designation, a new program that gives expedited reviews for devices that offer more effective diagnostics or treatments, with no cleared or approved alternatives.

IDx-DR is designed  for detection of diabetic retinopathy by primary care clinicians rather than specialists. The system is an invention of Michael Abramoff, a practicing ophthalmologist who developed the system with clinicians, researchers, and engineers at the nearby University of Iowa.

More from Science & Enterprise:

*     *     *

Comments are closed.