Mayo Clinic has developed an advanced AI model named REDMOD capable of detecting pancreatic cancer years before clinical diagnosis. The AI analyzes routine CT scans, identifying subtle, pre-clinical signatures of the disease. In a validation study, REDMOD achieved a 73% detection rate on scans initially interpreted as normal, with an average lead time of 16 months before a clinical diagnosis.
This AI capability significantly surpasses that of human specialists. REDMOD's detection rate of 73% nearly doubles the 39% rate achieved by radiologists reviewing the same scans. For cases detected over two years prior to diagnosis, REDMOD showed 68% accuracy compared to 23% for human experts.
"The greatest barrier to saving lives from pancreatic cancer has been our inability to see the disease when it is still curable," stated Dr. Ajit Goenka, senior author of the study. "This AI can now identify the signature of cancer from a normal-appearing pancreas, and it can do so reliably."
Pancreatic cancer is projected to be the second-leading cause of cancer-related death in the U.S. by 2030. Early detection is critical, as over 85% of patients are diagnosed after the cancer has already spread, limiting treatment options.
Mayo Clinic is currently evaluating REDMOD's clinical integration through a prospective study called AI-PACED, aiming to translate its laboratory success into real-world patient care.