November 15, 2018
3 min read

Artificial intelligence zeroes in on DR

The IDx-DR AI system can diagnose diabetic retinopathy with high sensitivity and specificity without human input.

You've successfully added to your alerts. You will receive an email when new content is published.

Click Here to Manage Email Alerts

We were unable to process your request. Please try again later. If you continue to have this issue please contact

IDx-DR, an autonomous artificial intelligence-based diagnostic system for the detection of diabetic retinopathy in primary care offices, outperformed all prespecified superiority endpoints of sensitivity, specificity and imageability rates, according to the results of a clinical trial.

After the results were released, the FDA cleared the system in April for use by health care providers to detect more than mild diabetic retinopathy and diabetic macular edema, Michael D. Abramoff, MD, PhD, told Ocular Surgery News. Abramoff is founder and CEO of IDx, the company that developed the system.

“This is one of the most important causes of blindness. It’s preventable if caught early, but we’re not catching it early. Less than 50% of patients get the retinal exam they need. They’re not getting an exam, and they’re going blind needlessly when we have excellent treatments to prevent this. To do better, one way to do this, and we think the best way, is autonomous AI. That’s special because it makes a clinical decision for a patient without a human being involved in the diagnostic process. An operator is taking the images, but the clinical decision is made by the system without human input. This is the first time the FDA has approved something like this anywhere in medicine,” he said.

Michael D. Abramoff, MD, PhD
Michael D. Abramoff

Designed for primary care

The system was designed for use in the primary care setting, operated by a user with no prior experience in ophthalmology diagnostics or capturing images of the retina. The system includes a robotic camera that autofocuses and auto-aligns. The system allows the operator to take consistent, high-quality pictures of patients during every exam, Abramoff said.

In October, IDx reached an exclusivity agreement with Topcon to use the Topcon NW400 robotic fundus camera in the IDx-DR system. The agreement also grants exclusive rights to IDx in the United States as the only autonomous AI company permitted to sell its products with the NW400.

While prior AI systems have been developed and have shown diagnostic success, the IDx-DR system evaluated real-life patients at 10 primary care study sites and compared results with the Wisconsin Fundus Photograph Reading Center (FPRC) widefield stereoscopic photography and macular OCT, obtained by FPRC-certified photographers.

The FPRC is the gold standard for trials that require grading of the severity of diabetic retinopathy, including the Diabetic Retinopathy Clinical Research Network and Epidemiology of Diabetes Interventions and Complications/Diabetes Control and Complications Trial studies, Abramoff said.

“Nobody has ever gone to primary care offices and said we will evaluate all these people with diabetes and compare them with the Wisconsin Reading Center, with high-quality images taken by highly experienced certified clinical photographers, as well as OCT. That has never been done. We want the highest standard for what everyone will agree upon,” he said.


Exceeding superiority endpoints

The clinical trial organizers raised $22 million to complete the study with the FPRC. The clinical trial included 900 patients with no history of diabetic retinopathy examined at primary care clinics. All AI system operators underwent a standardized training protocol before beginning the study and capturing patient images.

More than mild diabetic retinopathy was defined as ETDRS level 35 or higher and/or clinically significant and/or center-involved diabetic macular edema in at least one eye, according to the study.

The AI system exceeded all the prespecified superiority endpoints, achieving sensitivity of 87.2%, specificity of 90.7% and imageability of 96.1%, Abramoff said.

“All of those [the prespecified endpoints] were superior to what we had agreed on with the FDA. We considered this a monumental accomplishment, given the robust reference standard used for the trial, including both a widefield view of the retina and OCT imaging,” Abramoff said.

The AI system has important telemedicine implications that could potentially deliver high-quality, low-cost exams for patients in areas of the world that lack access. Unlike telemedicine systems, the images would not have to be read by clinicians offsite but would be able to provide a diagnosis instantly, Abramoff said. – by Robert Linnehan

Disclosure: Abramoff reports he is the founder and CEO of IDx and has patent and patent applications on the technology described.