Diabetic retinopathy takes sight slowly and quietly. Caught early, it is treatable. Caught late, it is not. The difference is whether someone looked at the right photograph in time.
The IMPRESS II project has built an artificial intelligence that assesses retinal photographs for signs of the disease with accuracy up to 97%. It is a research collaboration between Odense University Hospital and the University of Southern Denmark, with Steno Diabetes Center and SDU Robotics, funded through the BETA.HEALTH platform.
The audience was the hardest kind: peers who can evaluate the method, and a funder deciding whether it continues. Precision was not optional.
We delivered three versions of one concept film, built around the researchers and around the mechanism itself. What the algorithm sees, and why finding it early is the entire point.
