AI Models for EGFR Prediction in Lung Cancer Show Ancestry-Based Performance Gaps
A cohort study reveals that open-source AI pathology models predicting EGFR mutations in lung adenocarcinoma demonstrate variable performance across ancestral groups, with notably lower accuracy in Asian patients (AUC 0.68) compared to European (AUC 0.84) and African (AUC 0.85) populations, raising important considerations for equitable clinical deployment.