In 2016, Geoffrey Hinton predicted computers would replace radiologists within five years. They haven't. Radiology's ranks are growing, with practitioners expected to expand by 26 percent over the next three decades. Yet Hinton's prescience holds: AI now matches or exceeds human performance in many imaging tasks. As of early 2026, about three-quarters of the 1,400 AI-enabled medical devices cleared by the FDA are for radiology.

AI tools draft reports, flag urgent images, and sometimes spot abnormalities invisible to the human eye. For example, an analysis of 43 clinical trials found AI-assisted colonoscopies reveal more polyps. Human error rates for diagnostic images range from 3 to 5 percent, causing about 40 million errors worldwide yearly. The solution isn't replacing humans but combining AI precision with human experience.

Designing collaboration is complex. Even if AI is more reliable on average, it still errs where humans wouldn't, says radiologist Curtis Langlotz of Stanford. Radiologists must evaluate each AI decision, a "whole mental rewiring" per Paul Yi of St. Jude Children's Research Hospital. Unlike rules-based alerts, AI neural networks are "black box" systems, making oversight harder.

The goal is to merge AI's technical accuracy with human flexibility for patient benefit.