Detecting pancreatic cancer early with AI

A pancreatic cancer diagnosis is one of the most dreaded because it carries a very high mortality rate.

One reason is that symptoms, including fatigue, stomach pain, back pain and weight loss, are vague enough to go unnoticed.

For that reason, almost 80% of patients are diagnosed after the cancer has spread.

But now, artificial intelligence may help doctors know who is more at risk of this deadly disease and diagnose it before it’s too late.

From the prep work to the riding, Don Wexler loves everything about biking. He’s been riding 30 to 60 miles, three to four days a week for the past four decades, even after being diagnosed with stage four pancreatic cancer.

“I was just maybe just bullheaded and ignorant and just plowed on. I kept riding my bike,” he said.

But doctors found a tumor in his pancreas and another in his liver. Wexler was told he had 12 to 18 months to live.

“One of the reasons that there are so many deaths from pancreatic cancer is because it’s often diagnosed at a late stage,” said Brian Wolpin, Dana-Farber Cancer Institute oncologist.

Wolpin said another problem is that there’s no regular screening for pancreatic cancer like there is for breast and colon cancers, but Wolpin believes harnessing the power of AI could change that.

Wolpin’s team used AI to find patterns in over nine million patients, looking at medical records, family history and genetic mutations. The algorithm was able to correctly predict pancreatic cancer in a high percentage of cases.

“We’ve learned that if we screen patients who have genetic mutations or strong family histories, we can find pancreatic cancer early,” Wolpin said.

Wexler didn’t know he had a genetic mutation that made him a high risk, but with a new medication, chemo and radiation, eight years later, he has beat the odds, is cancer-free and hopes to be an inspiration to others.

“You gotta just keep doing what you do, keep living, keep exercising, ’cause I think doing that was one of the things that really helped me,” Wexler said.

Doctor Wolpin believes AI will be an even more powerful tool for diagnosis when a patient’s full medical record, including current medications and scans, vitals, and medical history, is included in assessing high-risk patients.