Artificial intelligence can spot skin cancer as well as a trained doctor
Scientists at Stanford University have developed an , the researchers fed the program tens of thousands of images gathered from all over the world, in addition to the labels determining the kind of cancer they showed, or if they were benign.
“There’s no huge dataset of skin cancer that we can just train our algorithms on, so we had to make our own,” Brett Kuprel, a co-author of the paper stated in a Stanford University post on the topic. “We collected images from the internet and worked with the medical school to produce a nice taxonomy out of information that was really unpleasant– the labels alone remained in a number of languages, including German, Arabic and Latin.”
The team ended up with a database of 129,450 images covering 2,032 different diseases. The deep neural network then scanned these pixel by pixel, trying to find the characteristics common to each medical diagnosis. By the end of its training duration, the network was able to recognize illness “on par with all tested experts,” state the researchers. With melanomas, for instance, the human skin specialists properly identified 95 percent of deadly lesions and 76 percent of benign moles. In the very same tests, the AI was right 96 percent of the time for the malignant samples, and 90 percent of the time for safe sores.
The Stanford team say the aim of developing their program is not to replace human skin doctors, however to offer individuals an economical alternative for early screening. The hope is that an advanced variation of the algorithm can be become an app and used at home. This would take extra training for the AI (it’s used to working with premium medical images– not the sort of shots that would be produced on a smart device) and more extensive assessments of its safety would require to be made before such a program could go public.