Scientists create algorithm that recognizes skin cancer as well as a dermatologist

Scientists at Stanford University have developed a machine-learning computer algorithm that can recognize skin cancers just as well as dermatologists. The researchers expect that in the long term anyone with a smartphone will be able to perform this test themselves.

The developed algorithm is based on the existing Google algorithm TensorFlow, which is able to recognize what can be seen in an image. The Stanford researchers’ algorithm works in a similar way. The researchers collected images of all kinds of skin cancers on the Internet and classified them. Ultimately, 130,000 images of skin conditions were collected, but only high-quality images of the deadliest forms of skin cancer were used to test the algorithm’s accuracy.

The algorithm scored several percentage points better than the doctors, when the test was limited to recognizing three types of skin cancer, according to a publication in the scientific journal Nature. If the number of recognizable forms of skin cancer was increased to nine, the dermatologists scored slightly better. Therefore, the researchers concluded that the algorithm achieves a similar result in recognizing skin cancers as dermatologists.

An advantage of the developed algorithm is that it can be made more or less sensitive, allowing researchers to adapt the algorithm. The researchers are already speculating about the possibilities of using this artificial intelligence in smartphones, with the phone’s camera sensor combined with the algorithm allowing users to visually test themselves for skin conditions.