Pancreatic cancer patients have one of the lowest five-year survival rates, due in large part to the disease going undiagnosed in its early and intermediate stages. There are no overt symptoms during the critical early period, and non-invasive screening tools for identifying early pancreatic tumors before they metastasize have yet to be developed and translated into clinical use.
A team of researchers at the University of Washington has recently developed a novel smartphone app, which enables straightforward, non-invasive screening for pancreatic cancer and other diseases based on image analysis. The app, named BiliScreen, combines a smartphone camera, machine learning tools and computer vision algorithms to identify early jaundice, a yellowing of the skin and eyes caused by increased bilirubin levels, which is otherwise undetectable at minimally elevated levels. Blood tests that measure raised levels of bilirubin are administered in instances of concern, such as when jaundice is already visibly apparent. BiliScreen is designed to identify jaundice in the whites of the eyes (sclera) at an early stage, before such changes are visible to the patient or health care provider, thereby acting as an early screen for pancreatic cancer, hepatitis, or other diseases.
The app analyzes images of an individual’s eye using a computer vision system that analyzes color metrics from the sclera, and correlates this information with bilirubin levels using artificial intelligence. To date, BiliScreen has been tested in 70 individuals, and has been shown to accurately identify instances of concern 89.7% of the time, showing proof of principle that this technology is promising and relevant. Future work will investigate the utility of the app in a larger cohort to develop a new screening program for individuals at risk of developing pancreatic cancer, hepatitis, and other related diseases.
Here’s a University of Washington video about the BiliScreen app: