The advent of artificial intelligence in medicine has long been touted as a way to deliver care where specialists can be hard to come by. Now, the vision-focused nonprofit Orbis International is putting that to the test, looking to prove the technology can live up to the promise.
Orbis conducted a study in Rwanda on the use of AI to screen for diabetic retinopathy. The condition can lead to blindness if left untreated, though early diagnoses have been shown to reduce vision loss by 98%, according to the group.
Researchers found that the deployment of AI programs—and their ability to deliver immediate feedback—made it much more likely that people would go to a recommended follow-up appointment, compared to receiving a manually completed report that may be returned to the patient days after screening.
"The number of people living with diabetes is growing rapidly worldwide, with the largest projected increase in Africa at an estimated 143 percent by 2045,” said the study’s principal investigator, Ciku Mathenge, a medical advisor for Orbis. “Unfortunately, screening programs for diabetes-related vision loss are often difficult to implement effectively in low-resource settings."
According to the researchers, the immediate feedback also provides an opportunity for providers to educate the patient about the importance of further care, as well as a picture of their own retinas, while they are still in the clinic and first learning about their condition.
And in some cases, they can visit an ophthalmologist on the same day—helping to overcome barriers such as financial costs, travel time, lack of clarity in the referral process and uncertainty about the treatability of the disease.
"Our research findings prove that integrating cutting-edge technology such as artificial intelligence into diabetic retinopathy screenings is not only feasible but is also an effective solution for identifying those at risk and improving access to care for those with the disease, even in the most vulnerable communities," Mathenge said.
In the U.S., the use of a computer program to catch early cases of diabetic retinopathy marked the FDA’s first approval for an AI-powered diagnostic, with the 2018 green light of Digital Diagnostics’ IDx-DR system.
Poorly controlled blood sugar levels among people with Type 1 or Type 2 diabetes can damage the retina’s blood vessels, leading to a gradual loss of vision. Orbis estimates that 224 million people will have some form of diabetic retinopathy worldwide by 2040, with potential blindness threatening as many as 70 million.
Orbis’ randomized study, conducted at four diabetes clinics around the capital Kigali in partnership with the Rwandan International Institute of Ophthalmology, recruited participants during routine visits to their diabetes care providers. Using AI through its Cybersight telemedicine platform, the program also aims to detect common eye diseases such as glaucoma and macular disease.
Of 823 screened participants, about one-third tested positive for diabetic retinopathy. Out of those 275 patients, half received the AI results while the other half received a human-generated report three-to-five days after the initial screening.
The AI group did 30% better compared to the control arm in successful adherence to referrals and typically had their appointments days earlier compared to the control group. The study’s results were published in the journal Ophthalmology Science.