Objectives: Electronic health record (EHR) phenotyping often relies on noisy proxy labels, which undermine the reliability of downstream risk prediction. Active learning can reduce annotation costs, but typical heuristics do not ... ...

请注册登录后继续浏览