Artificial intelligence is a healthcare and technology buzzword right now, but IDx Founder and President Michael Abràmoff is not a Johnny-come-lately to this phenomenon. His journey and that of the company’s lead product began over two decades ago in the Netherlands.
The product, IDx-DR, is an AI-based diagnostic system meant to be used as a standalone screening tool for diabetic retinopathy. In February, the company announced that it submitted an application with the FDA, which the agency accepted with a “breakthrough device” designation. In other words, based on its ability to address an unmet medical need, FDA will provide a faster review.
It’s the same unmet need that Abràmoff set out to solve in 1997.
Early detection
In a recent phone interview, Abràmoff recalled that he was working as an ophthalmology resident in Amsterdam when the first seeds were planted. Among his patients was an artist with late-stage diabetic retinopathy, an eye condition caused by damage to the small blood vessels and neurons of the retina.
It’s a cruel diagnosis: Early screening and treatment of diabetes-related eye damage can prevent blindness in more than 90 percent of cases. Yet it’s often asymptomatic in the early stages and once it progresses, there’s no going back.
Regular screening is, therefore, a critical part of managing the disease. But that’s easier said than done. Patients with diabetes are usually treated in a primary care setting. The doctor will examine their feet and kidneys for signs of damage; they’ll check their blood sugar and insulin levels, and write any necessary prescriptions.
“All these complications are evaluated,” Abràmoff said. “What was always lacking was a way to look at the retina at the back of the eye.”
To screen for retinopathy, patients need to be referred out to a specialist. The challenge isn’t the science, it’s ensuring patients can access and afford an annual eye test on top of their healthcare needs.
Abràmoff’s idea was to develop an algorithm that could interpret images of the retina and diagnose the disease without the need for a trained retinal specialist.
The work was initially academic, funded with grants from the NIH and the Juvenile Diabetes Research Fund (JDRF). Some years and many patents later, Abràmoff decided to advance the science himself by founding IDx and creating the IDx-DR diagnostic system using the algorithms. The company eventually moved into clinical trials in the United States with backing from angel investors.
If approved, the company believes IDx-DR would be the first autonomous, AI device intended for use in the front lines of healthcare. Unlike other medical imaging technologies that use AI, the system made by the Iowa company, is designed to make a directive, not a recommendation.
“The system has an artificial intelligence algorithm that evaluates the images coming from the camera and determines whether or not there’s diabetic retinopathy in them,” Abràmoff explained. “So it makes a clinical decision, without someone like me being involved — it’s fully autonomous.”
This is the most contentious bit about AI in that in some case, the technology is designed to replace humans. Not surprisingly, Abramoff is reluctant to talk about the implications of the technology replacing eye care providers. He stressed that the system isn’t designed to take patients away from trained professionals. Instead, it’s dealing with the reality that many patients with diabetic retinopathy still aren’t being screened before vision loss occurs — just like the artist he treated in 1997.
Peter McDonnell, director of the Wilmer Eye Institute at Johns Hopkins University School of Medicine, shared a similar sentiment in an editorial on the topic in 2010. It came in response to a paper by Abramoff and his team that claimed machines may now be as good as people when it comes to detecting diabetic retinopathy. McDonnell wrote:
Some ophthalmologists may disagree, resent, or even fear the replacement of human eyes and minds by computers to inspect retinal photographs for evidence of disease. But I don’t.
My view is that ophthalmologists are best employed using their highly specialized skills and talents to counsel and treat patients with diseases that need our expertise. Performing screening exams in asymptomatic diabetic patients who turn out to have no disease, and continuing to have so many diabetics in this country go years without any eye exams, do not make sense to me.
In other words, it’s a choice between autonomous AI or nothing. And it’s about to get worse.
By 2030, close to 55 million Americans are projected to have diabetes (both Type I and Type II), a 54 percent increase from 2015 levels. Many will no doubt live in rural areas like they do today, several hours drive from the nearest ophthalmologist.
It’s here that devices like the IDx-DR device can come as a boon, assuming it is cleared by the FDA.
Abràmoff claims that it can be operated by “anyone with a high school degree,” given four hours of training. That makes it scalable and cost-effective as demand for screening grows.
Another bonus is that results are delivered at the point-of-care: The algorithm takes 20 seconds and the total system runs in a matter of minutes, Abràmoff said. This allows it to be used almost like a blood pressure cuff in a primary care setting. The patient can have their retinas screened prior to their appointment. The doctor can then immediately follow up with an appropriate referral or even a change to the patient’s treatment plan if there are signs that the disease is not being adequately controlled.
Most patients, however, will get the all-clear. And that’s valuable, too. It saves them an unnecessary trip to see a specialist every year, Abràmoff believes and frees up the ophthalmologist for those that do have signs of the disease.
Photo: chombosan, Getty Images