Artificial Intelligence Outperforms Doctors in Diagnosing Skin Cancer

Researchers developed a new artificial intelligence (AI) tool known as a deep learning convolutional neural network (CNN), and found it is able to diagnose skin cancer (specifically melanoma) more accurately than a group of 58 international dermatologists with varying levels of expertise. The tool was created using Google’s AI technology and was trained using 100,000 skin images.

The CNN improved the doctors’ accuracy of diagnosis for skin cancers from a sensitivity of 86.6 to 95 percent (i.e., the doctors missed fewer true cancerous moles) and from a specificity of 71.3 to 82.5 percent (i.e., the doctors misdiagnosed fewer noncancerous moles as cancerous).

Therefore, the tool can prevent fewer cancer patients from being misdiagnosed as normal early on in their disease progression, as well as, prevent patients without cancer from undergoing unnecessary diagnostic procedures and treatments.


Haenssle HA, Fink C, Schneiderbauer R, et al. Man against machine: diagnostic performance of a deep learning convolutional neural network for dermoscopic melanoma recognition in comparison to 58 dermatologists. Ann Oncol. 2018.