For years, the AI conversation in radiology has focused on one question: Will AI replace radiologists? But what if we've been asking the wrong question this whole time?In this episode of Contrast & Clarity with the JACR, Maddi and Jeff sit down with Tessa Cook, MD, PhD to explore one of the biggest unsolved challenges in healthcare AI:Who watches the algorithms after they are deployed? AI models don't behave like CT scanners or medications. They can drift, degrade, and encounter new populations, protocols, and real-world conditions long after FDA clearance. That's where Assess-AI, the ACR Data Science Institute's fist national AI monitoring initiative, comes in. Using large language models to extract findings from radiology reports, Assess-AI creates a scalable way to compare AI predictions against what radiologists actually report. In other words: Radiologists read the scans. AI reads the scans. And now AI is reading the radiologists. Join us as we discuss AI governance, post-deployment monitoring, model drift, and why radiology may be uniquely positioned to become healthcare's AI control tower. Because building AI is one thing. Building systems we can trust is another. Find the full JACR article here: https://www.jacr.org/article/S1546-1440(26)00231-0/fulltextLink to the 2026 SIIM-ACR Data Science Summit: https://www.acr.org/Data-Science-and-Informatics/SIIM-DSI-Summit

Ask this episode

“What did the council decide on item 42?” — answered with the speaker and the moment.

Transcript

Related