Researchers at the Mayo Clinic have designed an artificial intelligence model that can predict a person's risk of developing pancreatic cancer years before a formal diagnosis.
The results of this research were presented at the 2026 American College of Surgeons Clinical Conference in Washington, D.C., held between September 26 and 29.
Pancreatic cancer is known to be relatively rare, but extremely deadly, as it is often diagnosed at advanced stages due to the lack of clear symptoms at first.
"Pancreatic cancer can be curable, but only when detected early, and fewer than one in five patients are diagnosed in time," says Cornelius Thiels, an oncology surgeon at the Mayo Clinic and a co-author of the research paper. "As a result, many patients' survival after diagnosis is still measured in months, not years."
How does the model work?
Thiels says that comprehensive screening for pancreatic cancer is not possible, so his team sought to develop an AI model that identifies patients at highest risk. He explains, "We know that pancreatic cancer develops over five to seven years, but what the doctor or patient sees only happens too late."
The model, developed by Thiels, lead author Chris Varghese, and their team, used longitudinal patient health records from the Mayo Clinic system, combining them with routine laboratory test results collected over an average of a decade or more.
The study dataset included 6,066 people with pancreatic cancer and 33,396 people in a control group with clinical records ranging from 7.5 to 19 years, with the aim of identifying subtle signs that might indicate an early-stage risk of developing the disease.
To ensure the model could predict pancreatic cancer three years before actual diagnosis, the researchers used three different measures, each answering a specific question about the model's performance.
The first method answers the question: Does the model differentiate between at-risk and non-at-risk patients? It received a score of 0.853 out of 1.0, meaning it is very close to perfect in distinguishing between those who will develop cancer and those who will not.
The second method answers the question: When the model warns of a danger, is it correct? It received a score of 0.712, meaning its warnings are reliable most of the time, and it does not issue many false alarms.
The third method answers the question: When the model states that the risk to someone is 50%, is this figure accurate? It received a score of 1.08 (with an ideal score of 1.0), meaning its figures are very close to reality.
Chris Varghese, the study's lead author, says, "If the model says someone has a risk of developing pancreatic cancer over 50%, there's an 88% chance they'll actually be diagnosed within a year." This means the AI can predict the disease up to three years before diagnosis. The higher the risk the model assigns, the more accurately it predicts the next year. In other words, the prediction within a year reflects how accurate the model is when the risk is high.
According to the researchers, the model accurately distinguishes between those at risk and those not at risk, has few errors in its warnings, and its figures are close to reality. This makes it a promising tool for the early detection of pancreatic cancer.
"We built this to be as generalizable, scalable, and practical as possible," Varghese emphasizes, noting that the data inputs for the model are gathered almost universally from hospital systems around the world. He adds, "If it proves to work, it can be used in virtually any environment."
what is your opinion on this news ?do let us know in the comments on tha comment box
if you liked this news article please share on friends & family & don't forget to follow on website & social media
