Expedited FDA Review for Novel AI That Detects Diabetic Retinopathy

^{Images courtesy of IDx.}

A diagnostic system driven by artificial intelligence (AI) is undergoing expedited review by the FDA, a sign that the technology could soon help identify diabetic retinopathy in patients, potentially preventing blindness and saving health systems money.

The Iowa-based company IDx announced today that it submitted an application for IDx-DR, its tool that can autonomously diagnose diabetic retinopathy, to the FDA, which designated the system a “breakthrough device.” What that means: The agency thinks IDx-DR could be a major step forward for medicine, a “first-of-its-kind” AI-fueled diagnostic system for clinicians, according to IDx.

“The FDA’s designation of IDx-DR as a ‘breakthrough device’ confirms what we have believed for a long time,” said Michael Abràmoff, MD, PhD, founder and president of IDx. “The healthcare system desperately needs a more efficient and cost-effective way to detect diabetic retinopathy. Too many patients go blind needlessly because they aren’t diagnosed in time.”

When caught early, diabetic retinopathy doesn’t typically lead to blindness or even vision loss. The issue is “almost entirely preventable,” though that fact nudges only half of the 30 million Americans with diabetes to undergo regular eye exams, according to IDx. As a result, diabetic retinopathy claims some level of vision from roughly 24,000 patients each year, making it the leading cause of blindness among people of working age, the company noted.

Diabetes runs the United States healthcare system $322 billion per year, according to the American Diabetes Association. IDx’s AI diagnostic system could help cut that price as it protects patients from diabetic retinopathy and the many costs associated with the disease and its effects, like blindness. An early diagnosis, for instance, spurs regular monitoring, laser treatment, or anti-vascular endothelial growth factor injections to properly and frugally manage the disease.

IDx-DR is designed to foster testing during routine medical visits, scrutinizing retina images for indicators of diabetic retinopathy. The device leverages “classical image analysis, deep learning techniques, and ophthalmic expertise” in seeking out the lesions, hemorrhages, and exudates that signal the existence of the disease, according to IDx.

AI innovators have been eager to tackle the prediction and diagnosis of diabetic retinopathy. Google, for example, has used deep learning to scan hundreds of thousands of images for the disease, a technique borrowed from the tech giant’s face- and object-scanning AI. Late last year, a team of international researchers published a study highlighting the efficacy of their machine learning method, revealing sensitivity and specificity levels above 90% in identifying diabetic retinopathy and similar diseases. Other investigators have also developed AI-driven models to predict the disease.

Clinical practices in Europe use IDx-DR after the technology was approved for sale there in 2013, according to IDx. The business has partnered with IBM Watson Health to distribute the system.

The company expects the FDA to make a decision regarding IDx-DR this year. Clinical trials spanning 10 primary care facilities and 7 states began last year, according to IDx.