Vascular Network Classification in Retinal Fundus Images
Diabetes has become the leading cause of blindness in adults, and as the number of people with diabetes increases, more people are at risk of losing their vision. Diabetic retinopathy (DR) is a disease that affects about one-third of adults with diabetes. Initially, DR has no symptoms, but if it goes untreated, DR can cause permanent blindness. This makes early detection of the disease crucial to preventing blindness, which requires screenings performed by experts. The screenings involve first taking pictures of the retina and then having a trained analyst look through the image for certain abnormalities. To make screenings more accessible to people with diabetes, this project seeks to automate the role of the image analyst.
Our approach determines rules using a shape grammar framework to construct retinal vasculature. By better understanding of how rules used in retinal vasculature differ between healthy and unhealthy patients, we hope to provide an automated but explainable method to diagnose patients.