April 10, 2025
Hari Trivedi, MD | Emory University
As artificial intelligence (AI) systems transition from development to real-world deployment, early experiences underscore that successful integration demands vigilance beyond initial implementation. While pre-deployment evaluations can mitigate known risks, post-deployment realities reveal unforeseen challenges related to model drift, workflow misalignment, and user engagement. In this presentation, Dr. Trivedi will talk about a multi-specialty AI governance process developed at a large academic institution and will explore lessons learned from the evaluation of over a dozen radiology AI models across diverse clinical scenarios. This talk will also focus on the need to perform rigorous subgroup evaluation of AI model performance to understand impact on clinical workflows.
