Imagine that instead of visiting a doctor for a skin abnormality, you could perform a self-examination with your phone, and artificial intelligence would diagnose you. That’s the goal for researchers at the Stanford Artificial Intelligence Laboratory. They developed a machine-learning AI that taught itself to diagnose skin cancer the same way a human dermatologist would in an exam, and it worked.
The researchers used one of Google’s AI systems that was taught to identify different objects. They trained the algorithm on images of different skin ailments to help build a data bank that would be used to compare conditions. The algorithm was tested against dermatologists and it performed just as well.
While this technology is still in its experimental phase, having a method for patients to check their skin and proactively track changes can assist dermatologists and also help detect skin cancer earlier.
Written by IEEE on March 8, 2017