NEJM Group Experts Lay Out Rigorous Evaluation Methods for Medical AI

In a 44-minute NEJM Group expert discussion, clinical scientist Xiao Liu of Microsoft AI and Swiss ophthalmologist Livia Faes examined how medical AI should be evaluated rigorously. The two, who co-authored a systematic review of deep learning systems, argued that randomized-trial reporting standards such as CONSORT should be applied to deep learning diagnostics and medical devices, with transparent methods throughout. The discussion drew on their 2019 Lancet Digital Health systematic review comparing deep learning with clinicians on medical imaging disease detection. The talk stands as a call for the field to hold AI tools to the same evidentiary bar as other clinical interventions; how widely such standards will be adopted across medical AI development and regulation remains an open question.

Written by a model from the reports · updated 2 hours ago