Bullock deployed a breast risk model no one had run before — and still says the radiologist owns the report
Darren Bullock counted roughly 1,200 FDA-approved AI algorithms across healthcare at RSNA in December, about 800 of them in radiology. His practice went live last month as the first outpatient imaging center to deploy a de novo–cleared risk score that reads the DICOM image itself, trained on a diverse population — replacing reliance on Tyrer-Cuzick, a questionnaire-based score built largely on European white women. His line under all of it: "AI does not practice medicine." The radiologist signs, so the radiologist owns it.
Go deeper
PodcastSolving Radiology Screening Bias and Ignoring the Shiny AIHow he and the vendor invented the workflow from scratch — image routing, results return, how findings land in radiologist and tech queues — with no prior deployment anywhere to copy.
Darren Bullock, COO & CIO, Envision Sally Jobe / Radiology Imaging Associates
Article"AI Does Not Practice Medicine"Why he refuses to hold every AI tool to one efficiency metric: the risk model actually slows the read, while MRI acceleration speeds the scan and moves the bottleneck onto the technologist.
Darren Bullock, COO & CIO, Envision Sally Jobe / Radiology Imaging Associates


