A computer algorithm can catch skin cancer with the same accuracy as a doctor, a study has found.
Researchers say the "exciting" discovery could soon result in smartphones being used as cancer scanners.
Scientists at Stanford University in the US trained the deep-learning algorithm, developed by Google and known as a neural net, to visually diagnose potential cancers.
A study, published in the journal Nature, showed that when tested on nearly 130,000 images, it detected malignant cancers with 91 per cent accuracy.
It performed with inspiring accuracy from the first test, said Professor Sebastian Thrun, of Stanford's Artifical Intelligence Laboratory.
News that makes sense
Your trusted source for staying up-to-date with the world around you. Get free daily news updates and analysis, straight to your inbox.
"We realised it was feasible, not just to do something well, but as well as a human dermatologist," Prof Thrun said.
At present, dermatologists use an instrument called a dermoscope to examine a patient's skin.
The algorithm exists on a computer, but the Stanford team would like to make it compatible with smartphones, which could potentially help in the early detection of deadly melanoma.
Melanoma is almost always curable when it is caught early.
In Australia, melanoma is the third most common cancer in Australian women and the fourth most common cancer in men, and the most common cancer in Australians aged 15-44.
Professor of dermatology Susan Swetter says the algorithm needs more testing, but believes it has the potential to help doctors.
"Advances in computer-aided classification of benign versus malignant skin lesions could greatly assist dermatologists in improved diagnosis for challenging lesions and provide better management options for patients," Prof Swetter said.
