The chances of surviving a diagnosis of pancreatic cancer are horrible. The five-year survival rate is an abysmal 9%, according to the Pancreatic Cancer Action Network. Even sadder is the news that 9% is viewed as relatively good news because it represents the third straight year the 5-year survival rate has climbed. In the absence of noninvasive screening technology and or telltale symptoms, by the time a tumor is detected it probably has already metastasized.

When the pancreas is compromised (pancreatitis), there can be a rise in bilirubin. While this increase can have many causes, from gallstones to decreased liver function, it could also serve as an early warning system for pancreatic cancer. Alex Mariakakis, a doctoral student at the University of Washington‘s  Paul G. Allen School of Computer Science & Engineering, is presenting a paper later this month at Ubicomp 2017, the Association for Computing Machinery’s International Joint Conference on Pervasive and Ubiquitous Computing. The paper describes BiliScreen, a smartphone app that uses a smartphone camera, computer vision algorithms, and machine learning to detect heightened bilirubin levels in a human’s sclera, the eye’s white part. Mariakakis and his team designed two inexpensive tools to reduce the impact of external lighting on the tests: a 3D-printed box and paper glasses with colored calibration squares. In testing the box worked better than the glasses. In a clinical study involving 70-people, the app and algorithms used with the box had a significantly high correlation in bilirubin level detection.

The next steps for BiliScreen are testing with more subjects who have a wider range of conditions and improving the box. Possibly the app can be advanced to the point that neither the glasses, the box, or anything else is needed other than a smartphone camera. With effective screening, people with pancreatic cancer that is detected earlier should have a higher survival rate than current technology allows.