Diabetic Retinopathy Detection with Machine Learning

Abigail Woolf was reading a research paper in her AI for Healthcare class about the success of a convolutional neural network — artificial neurons used to analyze visual imagery — that could detect referable diabetic retinopathy, a preventable but major cause of blindness around the world. The paper impressed her, but it was mum on actually utilizing an algorithm with so much potential in clinical settings. “I asked in class why the technology hadn’t been deployed,” says the Berkeley master of development engineering student, “and the professor said that it was complicated to standardize the data and processes behind everything.”


Her aunt, who has diabetes, has to make frequent treks to the doctor’s office to get her eyes checked. Woolf also knew there were cheap lenses that could be attached to iPhones for use in clinical settings. What if she could combine these powerful algorithms for detecting diabetic retinopathy — which can be more accurate than doctors — with these lenses that diabetics could use at home? It would save folks like her aunt time and money, while allowing ophthalmologists to spend more time on treating cases and less on diagnostics. Woolf, a 2022 Big Ideas Contest finalist, envisions “a data/camera package that can be sold or donated as a single unit to clinics for automated DR diagnostics.”


Through Big Ideas and the CoLab, Woolf will continue her research and look for partners for her diagnostic innovation. She’s reaching out to some of the researchers who worked on that paper from her AI for Healthcare class and asking them why the algorithm “hasn’t been deployed in the wild.”


“I am waiting for someone to tell me, ‘this idea is impossible because…’” Abigail says. “But until then, I will believe that it is possible.”

More Information?

Fill out your details and we will keep you updated on HealthTech CoLab news.