Mayo Clinic AI Detects Pancreatic Cancer 3 Years Early on CTs

Researchers at the Mayo Clinic have developed an artificial intelligence model called REDMOD that detects signs of pancreatic cancer on routine abdominal CT scans up to three years before clinical diagnosis, according to a study published in the journal Gut.

The model, known formally as the Radiomics-based Early Detection Model, identifies subtle tissue texture changes associated with preclinical pancreatic ductal adenocarcinoma, the most common form of the disease. These changes, described by researchers as "imaging-occult" signatures, are invisible to human radiologists even on expert review. REDMOD analyzes nearly 1,000 radiomic features from each pancreas, narrowing them to 40 key signals, primarily from wavelet-filtered images that reveal microscopic disruptions in pancreatic tissue. It employs an ensemble of three machine learning algorithms—logistic regression, random forest, and XGBoost—with a soft-voting mechanism for classification.

Tested on nearly 2,000 CT scans from multiple institutions, including prediagnostic scans from 219 patients later diagnosed with cancer and matched controls, REDMOD achieved 73 percent sensitivity in identifying preclinical cases at a...