Mayo Clinic AI spots pancreatic cancer up to three years before diagnosis

Mayo Clinic researchers have developed an AI tool that can spot hidden pancreatic cancer signs on routine CT scans. The findings suggest the disease could be caught much earlier, giving one of the deadliest cancers a better chance at treatment.

In Short

• Some warning signs appeared over two years ahead of diagnosis
• REDMOD flagged about 73% of hidden cancers around 16 months earlier
• More than 85% of pancreatic cancer cases are found after progression

A new artificial intelligence (AI) tool developed by Mayo Clinic could help doctors detect one of the deadliest cancers much earlier, sometimes up to three years before it is usually diagnosed.

Pancreatic cancer is often called a “silent” disease because it rarely shows clear symptoms in its early stages. By the time it is found, it has already spread, making treatment difficult.

Early detection has long been the biggest challenge.

Now, researchers say this new AI model may help change that.

FINDING WHAT THE HUMAN EYE MISSES

The tool works by analysing routine abdominal CT scans, which are tests that many patients already undergo for other health reasons. What makes it different is its ability to pick up very subtle changes in the pancreas that are too small or too faint for doctors to notice.

In the study, published in Gut, researchers analysed nearly 2,000 scans, including those of patients who were later diagnosed with pancreatic cancer.

These scans had originally been reported as normal.

The AI model, called REDMOD, correctly flagged about 73% of these hidden cancers, often around 16 months before diagnosis.

In some cases, it detected warning signs more than two years in advance, almost three times better than specialists reviewing the same scans without AI support.

WHY THIS MATTERS

Pancreatic cancer has one of the lowest survival rates among cancers, largely because it is found too late. According to experts, more than 85% of cases are diagnosed after the disease has already progressed.

“This has been the biggest hurdle, being able to see the disease when it is still treatable,” said Ajit Goenka, the study’s senior author.

By identifying early changes before a tumour even forms, the AI could give patients a critical head start—when treatments are more likely to work.

A TOOL THAT FITS INTO EVERYDAY CARE

Another advantage is that the system does not require new tests. It can analyse scans that patients are already getting, especially those at higher risk, such as people with newly diagnosed diabetes.

The AI runs automatically and has shown consistent results across hospitals, machines, and different patient groups, suggesting it could work in real-world settings.

WHAT COMES NEXT

Researchers are now testing how this tool can be used in everyday clinical care through a new study aimed at high-risk patients.

While more testing is needed, the findings offer hope that AI could one day help catch pancreatic cancer earlier when it is still possible to treat and save lives.

- Ends