There are now several hundred medical AI devices approved by the FDA for clinical usage. I will discuss how these devices are evaluated prior to approval and their deployment patterns across the U.S. We identify several key challenges in the evaluation process that could mask important limitations of the AIs. Motivated by this, I will share learning from a recent study conducting the first blinded, randomized clinical trial evaluating an AI (EchoNet) for assessing cardiac conditions. 

MIA Talks Search