Radiology AI Podcast | RSNA

The Radiology AI Podcast comes from RSNA. Both co-hosts, Paul Yi and Ali Tejani, are radiologists. It shows in the questions. The conversations keep coming back to what happens after an algorithm leaves the lab: does it catch what clinicians actually see, does it change a decision, and who answers for what follows? In 'From Prototype to Patient Care', two guests explain why tools built around the definition of a nodule do not always detect or segment T3 and T4 lung cancers properly. They describe brain metastasis tools that start seeing lesions at five or six millimeters, and they argue that the real battle is at one millimeter, where metastases could be found months or perhaps a year earlier. The Hot Take series works differently. It is presented by Satyajit Tripathi, a training editorial board member and associate editor for the podcast, and it sets out to bring radiology's hallway conversations into the open. 'Hot Takes: Radiology Reimagined with AI' measures agentic AI against an old rule: technology should work for its users, not the other way around. It also asks a harder question. If AI surfaces more borderline abnormalities and leads to more follow-up imaging, more testing and more patient anxiety without clear outcome benefits, who is accountable? 'Hot Takes: Rapid Questions on the Future of Radiology AI' covers radiomics in clinical reports, which has to be developed in-house because institutions differ so much. It sets out a three-step test: first you have information, then you ask whether it is actionable, then whether it actually leads to action. It also asks what a useful LLM co-pilot inside PACS or an EMR system would look like, as opposed to just another button on the screen. Then there is the live room. 'Radiology: AI RSNA2025 Fireside Chat LIVE' puts the co-hosts in front of an audience with a panel, and the talk turns to human-AI interaction as the technology moves past single-finding detection toward more holistic interpretation and report writing. One panelist points out that healthcare usually adopts technology late, which normally means someone else has already paved the road. Here, nobody has. Other episodes widen the frame. 'BRATS Africa: Building Inclusive AI in Radiology' features Udonna Anezodo, who works on imaging without contrast at McGill University in Montreal and had recently been working in Lagos, Nigeria, along with Marouf Adewale. 'Radiology Meets Venture Capital' turns to careers: when can a radiologist who is interested in entrepreneurship, but not ready to found a company, help out as a consultant or domain expert? The guest's answer is that there are many entry points, including for people just out of medical school. Some episodes open with the co-hosts catching up with each other, which keeps things collegial. For clinical specifics, start with 'From Prototype to Patient Care'. For debate, start with either Hot Takes episode.