Artificial intelligence can detect breast cancer on screening mammograms more accurately than individual expert radiologists in some study settings. In a 2020 Nature study using large U.K. and U.S. datasets, Google Health’s AI system matched or exceeded the performance of six radiologists and, in the U.S. test set, reduced false negatives by 9.4% and false positives by 5.7% compared with the original clinical reads.
This evidence has sparked a debate that is stuck in a false choice. On one side are tech evangelists who think algorithms can simply replace clinical judgment. On the other are skeptics who defend “the human touch” as if every part of medicine requires it equally.
Both sides miss the point. The frontier is not whether to use AI in medicine, but how to design systems where algorithms and clinicians each do what they genuinely do better.
This article is exclusive to STAT+ subscribers
Unlock this article — plus in-depth analysis, newsletters, premium events, and news alerts.
Already have an account? Log in
