Artificial intelligence may increase doctor productivity

Michael D. Abramoff

Earlier this week, the FDA granted marketing approval for IDx-DR, the first artificial intelligence medical device to detect more than a mild level of diabetic retinopathy in adults with diabetes, as reported by Ocular Surgery News.

In March, the IDx inventors were awarded the first-of-its-kind U.S. patent for an algorithm to automatically determine an arteriolar-to-venular diameter ratio (AVR) in blood vessels.

Primary Care Optometry News spoke with Michael D. Abramoff, MD, PhD, who oversees the IDx leadership team and is a fellowship-trained retina specialist with a PhD in image analysis and the Robert C. Watzke Professor of Ophthalmology and Visual Sciences at the University of Iowa Hospitals and Clinics.

Abramoff explained how he and his co-inventors harness artificial intelligence (AI) to allow for increased patient access to identify health risks while saving time and money.

PCON: How does the AVR algorithm work for evaluating blood vessels?

Abramoff: The patent describes methods to quantify the diameters of arteries and veins in tissue based on images and then calculates from there the diameters of the trunk artery and vein feeding that organ, which allows comparisons of these diameters among individuals. We know that these trunk diameters, which in the retina are called central retinal artery equivalent diameter and central retinal venous equivalent diameter, correlate with micro- and macrovasculopathy, including the risk of stroke, the risk of hypertension, the risk of developing diabetic retinopathy, etc. The goal of such methods is to create AI-based systems that can calculate these risks for individual patients.

PCON: How about the diabetic retinopathy system? How would you envision this being incorporated into optometry practices?

Abramoff: IDx-DR is an autonomous AI diagnostic system developed for use in the front lines of care to give more patients access to diabetic retinopathy testing. The system incorporates a robotic fundus camera to capture images of the patient’s eyes and AI to determine image quality and assess for disease, often without requiring dilation. The system uses an autonomous AI diagnostic algorithm to immediately detect more than mild diabetic retinopathy or macular edema.

In most cases, the test results do not show presence of the disease, so it’s appropriate to delay further testing for 1 year. In cases in which test results are positive for diabetic retinopathy, it’s recommended that the provider refer the patient for further evaluation. For patients, IDx-DR has the potential to save time and money because in most cases patients can be evaluated for diabetic retinopathy without needing a separate office visit.

PCON: What does the future of AI on the front line of health care look like?

Abramoff: Artificial intelligence holds tremendous potential to lower costs, increase quality and improve access as it becomes more integrated into the front lines of care. Autonomous AI, which makes a clinical decision without expert oversight, could unlock massive, tangible economic value in the form of productivity gains as clinicians are freed from tasks that can be done just as effectively by a self-directed system.

From the perspective of the patient, as well as the provider, an obvious advantage of AI-based algorithms is the immediacy of the results. That means that generally no additional communication of test results is required, saving staff and patient time and removing costs from the health care system.

By reducing the need for human evaluation, IDx-DR could possibly provide more cost-effective and convenient access to diabetic retinopathy testing, which in turn could lead to a reduction in the world’s top cause of preventable blindness. – Interviewed by Abigail Sutton

Disclosure: Abramoff reports he has been funded continuously since 2004 by the National Eye Institute, the Veterans Administration, the Beckman Foundation and other federal, state and philanthropic funding agencies in the U.S. and Europe.