CONVERSATIONS IN RETINA: A Conversation With Michael Abramoff, MD, PhD

Automating the screening process for diabetic retinopathy.


To call Dr. Michael Abramoff a “Renaissance Man” would probably be an understatement. The Netherlands-born University of Iowa ophthalmology professor has earned a PhD in computer science, been instrumental in the creation of 3 retinal imaging companies, been a successful playwright, and served as an interviewer of Holocaust survivors for Steven Spielberg’s Visual History of the Shoah project. However, his greatest achievement may have occurred just recently when his decades of groundbreaking work in artificial intelligence culminated in the FDA approval of IDx-DR, which uses artificial intelligence for the highly accurate diagnosis of diabetic retinopathy. The IDx-DR is the first autonomous diagnostic system whose findings do not have to be validated by a physician. Retinal Physician caught up with the very busy Dr. Abramoff for a compelling question-and-answer session.

Dr. Abramoff providing training on IDx-DR.

Q. What led you to choose ophthalmology, and specifically the retina subspecialty, as your career choice?

A. As long as I can remember, I was always interested in medicine, especially vision, the brain, and neuroscience. I went to medical school in the Netherlands and chose a residency in ophthalmology because I felt it was the most exciting area of medicine. I also did a fellowship in Japan at the RIKEN Neural Networks Lab in Tokyo.

Q. But you also had a second passion for computer science. How did you begin to combine these interests in a creative and productive way?

A. I loved being a practicing retina specialist. I still do it. I love interacting with patients and performing intricate retinal microsurgical procedures. But in life you have to make choices. I often think that there should be a measure called a “P index” (P for patient) that could quantify how many patients have been helped by your science. I thought I could make my best contribution to the most patients by combining a knowledge of retinal imaging with my interest in computer science to develop easy-to-use diagnostic and imaging devices that could play a role in identifying and combating retinal disease.

Q. You were instrumental in starting 3 retinal imaging companies. How did you learn how to navigate the business world? Did you have any mentors who helped you?

A. At the first 2 companies — EyeCheck and i-Optics — I had partners but I never had a real mentor until I founded IDx, where Gary Seamans, our CEO and chairman, has shown me how valuable a true mentor can be. I really needed that. It is hard for a doctor to succeed in business without a mentor who has a wealth of business experience.

Q. What were your basic goals and priorities in starting IDx?

A. We felt that the best way to combat the growing incidence of diabetic retinopathy was to identify it at its earliest stage in individuals who might otherwise not be diagnosed. Our goal was to develop the software for an automated method that could make autonomous diagnostic decisions based on algorithms developed from artificial intelligence research.

We wanted to make it available and easy to use, especially at primary care providers and even in retail clinics so that it could be efficient, effective, and accurate even where an ophthalmologist is not present. Our challenge was to make the process so simple that an operator with a high school education can perform the test and identify disease biomarkers after only 4 hours of training. Though ophthalmology practices are not our primary target market, I think they could also develop diagnostic programs for diabetic retinopathy as part of community outreach. Right now, we’re seeing customer interest across the board of all our potential markets.

Q. The FDA had never cleared an autonomous artificial intelligence diagnostic system before. How did you work with the FDA to develop a large-scale clinical trial that would validate IDx technology?

A. We had many discussions with the FDA. We knew we would have to meet the very highest standard for patient safety, so we chose to compare the IDx-DR output with findings from the highly respected Wisconsin Reading Center. We needed to prove a high level of accuracy in identifying the presence or absence of “more than mild” diabetic retinopathy. We also needed to prove that our findings would be accurate for individuals of all ethnicities. The results of the clinical trial validated all of our goals.

Q. The IDx-DR has only been recently approved in the United States (it was approved earlier in Europe) but could you give us an idea of how much the system costs and what are the opportunities for reimbursement?

A. The biggest expense is the cost of the Topcon NW400 retinal camera, but we can provide a camera to help a health care provider get started with the program. As far as reimbursement, it is still very early and we are having discussions with insurers. We are working to provide our customers with clear guidance on what codes they could use to apply for reimbursement.

Q. Now that you have your first approval for autonomous artificial intelligence, are other initiatives in the IDx pipeline?

A. Yes. Our glaucoma detection program is furthest along. We are also developing AI algorithms that can detect disease across a number of imaging modalities, with a focus on fundus photography and optical coherence tomography. Current prototypes also exist for macular degeneration, Alzheimer disease, cardiovascular disease, and stroke risk. RP