saliency maps
Visualization techniques used to reverse-engineer which image regions a neural network relies on, part of explainable AI.
As defined in 313 | Eric Topol on the Changing Face of Medicine and Aging
Where it comes up · 1
Most important claims · 1 of 1Search all claims for “saliency maps” →
- Once trained on hundreds of thousands of images, AI's 'digital eyes' can interpret medical images far more accurately than expert physicians, detecting not only the targeted disease but findings invisible to humans, such as inferring heart attack, stroke, kidney, and liver disease risk from a retinal scan.313 | Eric Topol on the Changing Face of Medicine and Aging