Could a computer replace a dermatologist? When it comes to detecting skin cancer, maybe so.
According to Stanford researchers in a new paper published in the journal Nature, the future of skin cancer diagnosis may be a simple smartphone app, which patients can use themselves to work out if their moles or other skin abnormalities are malignant.
“Our objective is to bring the expertise of top-level dermatologists to places where the dermatologist is not available,” said Sebastian Thrun, adjunct professor at Stanford University and senior author of the study.
By coaching a computer to develop pattern-recognition skills using an algorithm-based technique known as “deep learning,” Thrun and his team created an automated dermatologist who is able to identify images of skin cancer moles and lesions as accurately as a suitably qualified human.
The process involved presenting the artificially intelligent computer with a huge dataset of 129,450 images from 18 doctor-curated online repositories representing more than 2,000 skin diseases. By feeding the computer the diagnosis of each mole or abrasion, it effectively learned its own rules for how to make its own automatic diagnosis.
When the team presented previously unseen images to the artificial intelligence system, it didn’t disappoint, recognizing the most common type of skin cancer (malignant carcinomas) as well as the most deadly (melanomas.)
“This algorithm performed as well as board-certified dermatologists at several key diagnostic tasks,” said Andre Esteva, first author of the new paper.
Skin cancer is the most common of all cancers in the U.S. (melanoma is the deadliest form of cancer for young adults,) but if detected early, the five-year survival rate is 98 percent. For more information on skin cancer risks and detection, visit the Skin Cancer Foundation.