Mayo Clinic AI Spots Pancreatic Cancer Up to Three Years Before Diagnosis

Researchers at the Mayo Clinic and the University of Texas MD Anderson Cancer Center have developed an AI model called REDMOD (Radiomics-based Early Detection Model) that can detect signs of pancreatic cancer on CT scans up to three years before a clinical diagnosis is made. Rather than looking for a visible tumour, the system analyses radiomic patterns — subtle disruptions in tissue texture and structure that are often too minor for the human eye to spot. The team trained REDMOD on 969 CT scans before testing it on a separate set of 63 scans from patients who were later diagnosed with the most common form of pancreatic cancer, all taken before their diagnosis, alongside 430 healthy controls. REDMOD correctly flagged 46 of the 63 pre-diagnostic scans, a detection rate of 73%, at around 16 months before diagnosis on average. Human radiologists had previously cleared all 63 scans as normal; when two radiologists reviewed them alongside REDMOD, they identified early signs in only 38.9% of cases. Of the 430 healthy controls, 81 were incorrectly flagged as suspicious. The AI produced consistent results across two further dataset tests using different equipment at different hospitals. Pancreatic cancer is on course to become the second-leading cause of cancer-related death in the US by 2030, with 85% of cases not caught until the disease has already spread. The researchers plan to test REDMOD across larger and more diverse groups and explore how it can be integrated into existing clinical workflows. The study was published in the journal Gut.