Most people still hear “AI in medicine” and picture a robot doctor. What is shipping in hospitals is quieter and more useful: software that sits in the imaging pipeline, reads CT scans and X-rays alongside the humans, and raises a hand when something looks urgent.
Mercy’s rollout of Aidoc’s aiOS platform is a clean example. Starting in January 2025, the health system put imaging AI across more than fifty hospitals. Public reporting from Mercy describes millions of cases analyzed, hundreds of thousands flagged, and a large cut in turnaround time — with serious findings pushed forward for earlier human review. About a quarter of analyzed cases were “enhanced” by the system: a priority flag, a stronger read of the primary finding, or an additional finding clinicians were not originally hunting. That is not a demo day. That is radiology workflow at health-system scale.
OSF HealthCare’s expansion of RapidAI to all eighteen of its hospitals tells the same story from another network: stroke and broader clinical imaging AI as standard infrastructure, not a pilot in one shiny building.
The applied science here is pattern recognition under clinical constraints. The model does not get the last word. It compresses time-to-attention for people who do. Patients never see the algorithm. They feel it when treatment starts sooner.
What to watch next is governance as much as accuracy — Joint Commission–style responsible-AI certification and how health systems measure false alarms versus missed finds. For appliedscience.com, the hook is simple: the science is already in the reading room.