
0:00
21:39
A hospital CEO said AI could already replace “a great deal” of radiologists. Another called it “actually better than human beings.” Dr. A says both statements ignore the same inconvenient number: the false negative rate.In this episode, Dr. A breaks down the five hurdles that stand between today’s AI imaging tools and an actual radiologist replacement: radiologists trusting the technology enough to sign off on it, clinicians trusting an AI read without a second opinion, patients accepting an AI diagnosis, radiologists’ own resistance to being replaced, and the biggest fight of all: who takes on the liability when an AI misses a cancer. Running the clock on each hurdle, he lands on a prediction: real replacement doesn’t begin before January 2040.The claims are premature, they’re profitable for the hospitals making them, and none of the executives making them are radiologists.
We Also Cover
● For patients: An AI “read” on a mammogram or X-ray isn’t a diagnosis on its own, a radiologist’s second read is still the standard of care, and patients should ask whether a human physician reviewed their scan before acting on it.
● For clinicians: The technology’s real bottleneck isn’t accuracy alone, it’s trust, liability, and regulation. Current AI tools are decision-support, not a replacement for radiologist interpretation, no matter what a hospital CEO claims publicly.
● System-level: No one has settled who’s liable when an AI misses a cancer, the software vendor, the hospital, or no one and that unresolved question, not the technology itself, is what’s actually slowing deployment.
We Also Cover
● For patients: An AI “read” on a mammogram or X-ray isn’t a diagnosis on its own, a radiologist’s second read is still the standard of care, and patients should ask whether a human physician reviewed their scan before acting on it.
● For clinicians: The technology’s real bottleneck isn’t accuracy alone, it’s trust, liability, and regulation. Current AI tools are decision-support, not a replacement for radiologist interpretation, no matter what a hospital CEO claims publicly.
● System-level: No one has settled who’s liable when an AI misses a cancer, the software vendor, the hospital, or no one and that unresolved question, not the technology itself, is what’s actually slowing deployment.
More episodes from "They Did What?"



Don't miss an episode of “They Did What?” and subscribe to it in the GetPodcast app.








