Report: AI Predicts Cancer 3 Years Prior to Diagnosis

A new study led by Harvard Medical School and the University of Copenhagen, in collaboration with VA Boston Healthcare System, Dana-Farber Cancer Institute, and the Harvard T.H. Chan School of Public Health, found that an AI tool could identify individuals at the highest risk of pancreatic cancer up to three years before diagnosis, using only medical records. The research, published in Nature Medicine, indicates that AI-based population screening could help in identifying those at risk and in speeding up diagnosis of pancreatic cancer, which is often detected at later stages when treatment is less effective. “An AI tool that can zero in on those at highest risk for pancreatic cancer who stand to benefit most from further tests could go a long way toward improving clinical decision-making,” said co-senior investigator Chris Sander.

The lack of population-based screening tools for pancreatic cancer means that only those with a family history and certain genetic mutations are screened. However, the targeted approach may miss other cases outside those categories, as the researchers pointed out. Soren Brunak, the other co-senior investigator, highlighted that “AI-based screening is an opportunity to alter the trajectory of pancreatic cancer, an aggressive disease that is notoriously hard to diagnose early and treat promptly when the chances for success are highest.”

The researchers trained the AI algorithm on two separate datasets consisting of 9 million patient records from Denmark and the US. Based on combinations of disease codes and their timing, the AI model identified patients who are likely to develop pancreatic cancer in the future. The researchers tested various versions of the AI models for their ability to detect people at risk within different time scales. Each version of the AI algorithm was significantly more accurate than current population-wide estimates of disease incidence. The researchers believe the model is at least as accurate in predicting disease occurrence as current genetic sequencing tests, which are only available for a small subset of patients.