Rethink Imaging

The episode delivers substantial, non-obvious insights about AI in radiology, particularly the shift from quality-focused detection tools (2016) to capacity-focused workflow integration (2026), the distinction between narrow AI and foundation models, and the 10x annual imaging growth vs. radiologist capacity problem. However, there are stretches of explanation that repeat concepts (e.g., multiple definitions of agentic AI) and some conversational throat-clearing that dilute density.

“Every year, the amount of imaging is an order of magnitude that's 10x bigger than the amount of capacity that's being generated every single year.”

“the worst quality exam is an exam that you never get to. And that's the problem of today.”

The episode offers genuinely fresh framing - particularly the recontextualization of AI priorities from quality enhancement to capacity management, and the precise definition of agentic AI with self-monitoring as a differentiator. The analogy to handlebars and basal ganglia for workflow change is novel. However, much of the AI taxonomy (deep learning, foundation models, multimodal) is now standard discourse, and frameworks around data quality and model generalization are well-trodden.

“This is not a cycle. This is like a direct line that started to go up like this with a 10x order of magnitude difference in the next 10 years.”

“if all of a sudden you then say, yeah, I'm going to turn your handlebars now so that when you turn to the right, you're actually going to the left, all of a sudden you have to think a lot about what buttons to push.”

Dr. Nina Kottler is exceptionally well-credentialed and operating at genuine scale. She is CMO of Mosaic (Radiology Partners' tech arm), a practicing radiologist with 20+ years of experience, holds a graduate degree in applied mathematics/optimization theory, has deployed AI across 20+ million patient exams, serves on ACR/SIIM/RSNA committees, and won a 2018 Trailblazer Award. This is a true practitioner-operator, not a pure theorist or career podcast guest. She speaks with authority born from real implementation challenges.

“she's a practicing radiologist with more than 20 years reading cases”

“She spent years building and deploying clinical AI at a scale very few people have worked at. North of 20 million patient exams.”

The episode includes concrete numbers (10x growth gap, 15,000 radiologist shortfall vs. 35,000 total, 16-minute vs. 45+ minute ER turnaround times, 85% imaging usage rate, 20+ million exams deployed, 9-week delays cited) and specific clinical tools (named models like ChatGPT, Gemini, Claude; mentions narrow AI binary outputs; references FDA authorization). However, many claims lack detailed examples: specific hospital ROI numbers, named AI tool performance comparisons, or granular data on the Mosaic system are absent. The evidence supports claims but rarely goes deep.

“Every year, the amount of imaging is an order of magnitude that's 10x bigger than the amount of capacity that's being generated every single year.”

“We are expected that we would need another 15,000 radiologists and we only have 35,000 in the industry to begin with.”

The host, Chris St. John, asks solid foundational questions (AI taxonomy, clinical applications, data validation, trust, implementation barriers) and occasionally pivots well (e.g., requesting a concrete workflow description). However, follow-ups are often surface-level confirmations rather than probing challenges. The host rarely pushes back on claims, doesn't press for contradictions, and misses opportunities to stress-test the guest's framing. Some segments feel like the guest is lecturing rather than being interrogated. The host does ask about the softer ROI measurement challenge effectively.

“I was wondering if you could like, give us a bit of like, a system or like, what's the term? I'm looking for, like a classification of these different buckets?”

“How do you think about quantifying ROI in, in this context? Right. Like, obviously there's some cases where, you know, you could say, okay, we've done this many more exams since implementation.”