Mayo Clinic AI Model Detects Pancreatic Cancer Up to 3 Years Before Clinical Diagnosis on Routine CT Scans — Landmark ASCO 2026 Validation Study

A landmark validation study presented by Mayo Clinic at ASCO 2026 and subsequently featured in post-meeting analysis found that an artificial intelligence model trained on routine abdominal CT scan data can detect subtle radiologic signs of pancreatic cancer up to 3 years before the cancer becomes clinically apparent or radiologically visible to human radiologists. The model was validated on a large retrospective cohort from Mayo Clinic and external sites. Pancreatic ductal adenocarcinoma is one of the deadliest cancers precisely because it is typically diagnosed at Stage III–IV, when curative surgery is no longer possible (only approximately 10–15% of PDAC patients are diagnosed at surgically resectable Stage I–II). The AI model identifies changes in pancreatic tissue texture, ductal morphology, and surrounding fat attenuation that precede frank tumor visibility — changes that fall below the threshold of human perceptual detection on routine clinical review. If validated prospectively and integrated into clinical workflows, this technology could shift PDAC diagnosis to earlier stages where curative resection is feasible, dramatically improving the 12% 5-year survival rate. The Mayo Clinic AI detection breakthrough arrives in the same week as daraxonrasib's historic ASCO plenary result — together representing a convergence of AI-driven early detection and precision-targeted therapy that could fundamentally reshape the prognosis of this historically fatal disease.