An AI model may help predict who is more likely to develop pancreas cancer, Elizabeth Tracey reports

Podcast: Download (Duration: 1:07 — 1.6MB)

Subscribe: RSS

Pancreas cancer is associated with a very poor prognosis, with many people diagnosed with the disease given only months or a couple of years of predicted life remaining. Now an AI model may discern who is at risk to develop the disease more than a year in advance. Johns Hopkins Kimmel Cancer Center director William Nelson explains.

Nelson: They try to train AI model using a cohort from many different institutions. 969 folks, 156 of them had images before they were later discovered to have pancreas cancer. They tested it on an independent validation cohort, 93 folks, 63 of them they ultimately developed pancreas cancer. What they came up with had about a 73% sensitivity to predict who was going to end up with pancreas cancer 475 days ahead. 68% or some more than two years ahead.            :34

Nelson notes that this prediction was not based on identifying a specific lesion. At Johns Hopkins, I’m Elizabeth Tracey.