Overview
- An AI model from Mayo Clinic and MD Anderson, described in Gut, spotted early pancreatic cancer signals on routine CT scans up to three years before diagnosis.
- Across retrospective tests, the tool identified early signs in about 73–75% of future cases, on average 16 months sooner than standard detection, while radiologists found far fewer.
- REDMOD analyzes radiomic patterns, or tiny texture changes in pancreatic tissue that the human eye typically cannot see on a scan.
- The team trained the system on roughly 1,000 scans and then showed it flagged 46 of 63 people who later developed cancer, but it also produced 81 false positives among 430 healthy people, prompting follow-up testing.
- The model is now in prospective trials that will track participants for several years to measure real-world accuracy, false-alarm burden, and whether earlier detection leads to more curative treatment in a disease with a 13% five-year survival rate and no routine screening.