An AI model developed at Mayo Clinic accurately predicts pancreatic cancer risk up to three years before diagnosis using longitudinal health data.
Key Details
- 1The AI model analyzed health records and lab results from nearly 40,000 patients.
- 2Validation included 6,066 pancreatic cancer cases and 33,396 controls, with 7.5 to 19 years of data.
- 3Model achieved AUROC of 0.853 and AUPRC of 0.712 for predicting risk three years out.
- 4Users with >50% model-predicted risk had an 88% chance of diagnosis within one year.
- 5The tool is being prospectively validated at Mayo and tested in a non-Mayo health system.
- 6Results will be presented at ACS Clinical Congress 2026; study is not yet peer reviewed.
Why It Matters
Early and accurate risk prediction for pancreatic cancer could transform screening, offering a major advance in catching a deadly cancer while still curable. The study demonstrates practical, large-scale AI integration into EHR-driven screening, with models designed for real-world hospital settings.
Source
EurekAlert
Related News
•EurekAlert
AI Model Predicts Organ Failure from CT in Pancreatitis Hours in Advance
A multicenter deep learning model (ORACLE) predicts organ failure in acute pancreatitis from CT scans with high accuracy, outperforming existing clinical tools.
•EurekAlert
MD Anderson AI Flags Lung Cancer Patients at High Pneumonitis Risk Using CT Scans
MD Anderson researchers developed an AI model that predicts which lung cancer patients may develop serious immunotherapy-induced pneumonitis using routine CT scans.
•EurekAlert
NYU Researchers Launch AI Tool for Stable Molecule Tautomer Prediction
NYU scientists have created an AI tool using crystallographic data to accurately predict stable tautomers in drug-like molecules.