Episode 61: Radiology: AI RSNA2025 Fireside Chat LIVE

Episode 61: Radiology: AI RSNA2025 Fireside Chat LIVE

by Radiology: Artificial Intelligence

Applied Sciences & Engineering Medicine & Health MBMedicineMBFMedical and health informatics
🎙 Radiology: Artificial Intelligence 👥 446 📅 February 13, 2026 ⏱ 61 min 👁 64 📄 debate 🧭 2026-08-16
Available in: English (current) Français

Keywords

radiology AIfoundation modelsagentic AIclinical deploymentdraft reporting

Summary

This live fireside chat, recorded at RSNA 2025, brings together experts in radiology AI to discuss the current state and future of the field. The panel, moderated by Drs. Paul Yi and Ali Tejani, includes Charles Kahn, Nina Kottler, Woojin Kim, and Linda Moy. The conversation begins with a lighthearted demonstration of AI-generated images and a game distinguishing AI from human-created content, highlighting the realism of modern generative models. The main discussion focuses on the clinical deployment of AI, particularly foundation models and agentic AI. Panelists agree that AI is already being used in practice, especially for automated draft reporting, with some reporting significant efficiency gains (e.g., 28% average improvement, up to 76% in some cases). They emphasize the importance of human-AI collaboration, noting that radiologists remain in control and that AI can reduce cognitive burden while improving detection. Challenges such as potential over-reliance and the need for ongoing vigilance are discussed. The panel also touches on regulatory aspects, with some foundation models receiving FDA clearance. Overall, the chat provides an optimistic yet cautious outlook on the integration of AI into radiology, stressing the need for careful implementation and continued research.

191 words

Critical Evaluation

Value of the Information & Strength of the Argument

The value of the information is high, as it comes from leading experts actively involved in deploying AI in radiology. They provide concrete examples of current clinical use, such as automated draft reporting at Radiology Partners, and cite specific efficiency gains. The argumentation is solid, grounded in practical experience and recent research (e.g., MedVersa, Merlin, MedGemma). The panelists present a balanced view, acknowledging both benefits and risks, such as the potential for radiologist complacency. They argue that AI is not replacing radiologists but augmenting them, and they stress the importance of human oversight. The discussion is persuasive because it combines real-world data with reasoned predictions, though some claims are anecdotal and lack formal citations.

Scientific Rigor, Source Quality, Title Accuracy

The scientific rigor is moderate: the discussion is informal and lacks formal citations, but the panelists reference specific studies and products (e.g., FDA clearances, papers from Northwestern). The sources mentioned are credible but not systematically cited. The title accurately reflects the content, as it is a live fireside chat at RSNA 2025. The content is well-aligned with the title, covering the year’s developments and future directions in radiology AI. No comments were provided for analysis.

204 words

Title / Content Match

The title accurately reflects the content: a live fireside chat at RSNA 2025 discussing AI in radiology.

Quality & Reliability

7/10

The discussion features expert panelists with direct experience in deploying AI in radiology, providing credible insights. However, it is a conversational fireside chat without formal citations or peer-reviewed references, and some claims (e.g., efficiency gains) are anecdotal or based on unpublished data.

Key Moments

Markers derived by PSI from the transcript: the creator did not define chapters.

Cited Sources

  • MedVersa — Mentioned as a research project from Harvard for automated report generation
  • Merlin — Mentioned as a Stanford project for automated report generation
  • MedGemma — Mentioned as a Google model for medical AI
  • JAMA Network Open paper from Northwestern — Referenced for efficiency gains with vision-language models

Concurring Sources

  • Radiology: Artificial Intelligence — The journal where the podcast is based, providing peer-reviewed research on radiology AI.

Contribution & Novelties

The video provides an insider perspective on the current state of AI in radiology, particularly the shift from narrow AI to foundation models and agentic AI. It offers real-world examples of clinical deployment, such as automated draft reporting, and discusses the nuanced human-AI interaction. The panelists’ predictions and concerns add value for practitioners.

Pour aller plus loin :

  • Foundation models in radiology — Discusses the potential of foundation models in medical imaging.
  • Agentic AI in healthcare — Explores the role of agentic AI in clinical workflows.
  • FDA clearance for AI in radiology — Overview of regulatory pathways for AI devices.

100 words

Radar Profile

The radar profile shows high scores in quantity of information and global reliability, reflecting the expert panel and concrete examples. The technical level is moderate, suitable for a broad audience. The quality of information is strong but not perfect due to the informal nature and lack of formal citations.

Reliability 7/10