As we’ve written several times earlier, diabetic retinopathy (DR) is a significant concern for the estimated 30 million Americans who live with diabetes. A recent CDC study indicated that nearly one-third of all Americans with diabetics aged over 40 suffer to some extent from DR. Diabetes is the leading cause of blindness among adults. As the incidence of diabetes increases, the numbers of people blinded by DR will follow. Many companies and institutions are working on tools and treatments for early detection and treatment of more than mild diabetic retinopathy (mtmDR), including IBM, Google’s DeepMind Team, and the Madras Diabetes Research Foundation.
Privately-held IDX, an Indiana AI diagnostics company, recently announced the FDA’s De Novo request to market IDx-DR. The De Novo program is a fast-track approval process for novel technology with relatively low risk. The IDx-DR system’s FDA go-ahead is the first approval for an autonomous, AI-based DR-diagnostic system. The system is appropriate for primary care office patient screenings because specialists are not required to administer the test or analyze the results. Primary care patient screening via the exam, which according to IDx takes a few minutes, could discover early DR symptoms and enable clinical staff to counsel patients without delay or specialist referral. Even though DR is a major threat, the FDA stated about 50% of adults with diabetes don’t see an eye doctor each year. The FDA clearance followed a clinical trial at ten primary care sites across the U.S. An FDA spokesperson reported IDx-DR correctly identified mtmDR 87.4 percent of the time and also correctly identify patients without mtmDR 89.5 percent of the time.
Autonomous screening for diseases with dire consequences is a major win and first step for AI-based health-system benefits. As AI develops, we can hope, and possibly assume, that technology such as IDx-DR will work with increased granularity and with growing accuracy.