Next-generation AI model for visually occult pancreatic cancer detection in a low-prevalence setting

Failure of conventional imaging to detect pancreatic ductal adenocarcinoma (PDAC) at is visually occult prediagnostic stage is a primary barrier to improving its otherwise poor rate of survival. Researchers at three US centers developed and validated the Radiomics-based Early Detection Model (REDMOD), an AI framework to identify subvisual radiomic signatures of prediagnostic PDAC on standard-of-care CT. REDMOD was trained on a cohort of 969 individuals (156 prediagnostic, 813 control) and tested on an independent set of 493 individuals (63 prediagnostic, 430 control) simulating a low prevalence early detection paradigm. On the independent test set, REDMOD identified occult PDAC (AUC 0,82; 73.0% sensitivity) at a median 475-day lead time. This represented nearly twofold higher sensitivity than radiologists (38.9%; p<0.001), which grew to nearly threefold (68.0% versus 23.0%) at > 24-months lead time. REDMOD showed strong longitudinal stability (90%-92% concordance) and generalizable specificity across multi-institutional and public datasets. The authors conclude that REDMOD is an externally validated AI model that surpasses radiologists for PDAC detection at its visually occult prediagnostic stage.