Skip to main content
. Author manuscript; available in PMC: 2019 Jan 1.
Published in final edited form as: Nature. 2017 Jan 25;542(7639):36–38. doi: 10.1038/nature21492

Figure 1 |. Accurate cancer diagnosis by computer.

Figure 1 |

a, Esteva et al.3 trained a computer to recognize skin diseases using a deep-learning approach in which the computer was presented with 129,450 digital images of medically diagnosed samples. b, To test the diagnostic accuracy level of the trained machine, the computer and 21 skin specialists were presented with previously unseen images representing verified examples of benign or malignant skin disease arising from two cell types. The machine was as successful as or better than the doctors at diagnosing whether or not the images represented conditions that might need further medical attention.