AI improves diabetic retinopathy follow-up in underserved regions
Click Here to Manage Email Alerts
Key takeaways:
- The AI system increased DR follow-up rates by more than 30% compared with off-site image screening.
- More than two-thirds of patients preferred AI to human grading.
SAN DIEGO — An artificial intelligence system helped increase rates of follow-up for diabetic retinopathy care in Rwanda, according to a study.
At DOS Digital Day at the American Society of Cataract and Refractive Surgery meeting, Hunter Cherwek, MD, said the Cybersight AI-enabled platform developed by Orbis International includes more than 20 machine learning algorithms designed to detect sight-threatening conditions, including diabetic retinopathy (DR), glaucoma and macular disease.
“What we’re trying to do is give people who need it most access to AI for either diagnostics or training purposes,” Cherwek said. “We’re here to close the digital and data divide and, using our footprint with Cybersight, improve patient diagnosis.”
In the first Orbis randomized controlled trial, known as RAIDERS, researchers enrolled 827 participants from four diabetes clinics in Rwanda. The AI platform increased rates of follow-up for DR care by more than 30% compared with off-site image screening by a specialist.
Additionally, patient satisfaction was more than 99%, and more than two-thirds of patients preferred AI over human grading because of the immediate turnaround for results. Compared with a human grader, the sensitivity of AI for detecting referable DR was 92%, and the specificity was 85%.
Cherwek said Orbis is partnering with the minister of health of Rwanda to take the program nationwide. In addition, there are efforts in Vietnam to adopt countrywide screening with the AI platform.
“It’s been really impressive how much the doctors enjoy an algorithm that isn’t a black box,” Cherwek said. “There’s actually been a decline in human consults because of the machine mentoring on Cybersight.”