Only 5% of AI medical evaluations use real patient data, study finds. A new framework adapted from @StanfordCRFM
's HELM aims to bridge the gap between theoretical and practical AI performance in healthcare. @StanfordMed

→ View original post on X — @stanfordhai