5
votes
In breast cancer screening, deep neural networks use different features than radiologists
@Taro Makino:
DNNs perform well on a range of medical diagnosis tasks, but do they diagnose similarly to humans?In breast cancer screening, DNNs use different features than radiologists. Some are spurious, while others may represent new biomarkers.https://t.co/kyMiLtSxw0 1/9 pic.twitter.com/akpIH1OpYo
Apparently it's important to look at different subgroups separately to figure out what's going on. Background: Simpson's paradox.