What if we could all carry around in our pockets the technology needed to diagnose potentially deadly skin cancer? Well, that may be closer to reality than you might think. Researchers from Stanford University have developed a new artificial intelligence (AI) that is as accurate as doctors at identifying skin cancer from images, and hope to get everyday smartphones to carry it.
Roughly 5.4 million people a year are diagnosed with skin cancer in the United States, and while the survival rate is good if it is caught in the early stages, this drops to a shocking 14 percent if only detected at the late stage. Having the ability to easily check if a mole or lesion is likely to be cancer could significantly help cut the deaths from this disease. And so researchers decided to see if they could get a computer to do it for them.
The researchers built their program on software that was initially developed by Google in order to accomplish the slightly less vital, but no less significant, task of differentiating between images of cats and dogs. They created a library of 130,000 images of different skin lesions that represented over 2,000 different diseases. They then fed it into the algorithm, and trained it to visually diagnose the conditions from the images, identifying potential cases of cancer.
It usually requires a doctor to identify potential cancer and send the patient for a biopsy. This might now be possible with AI. Matt Young/Stanford
To test how accurate the AI is in picking out cancer, the researchers pitted it against 21 expert dermatologists in a test to identify malignant carcinomas and malignant melanomas, which represent the most common and deadliest skin cancer. Both the doctors and the software were asked whether or not they would send a patient for a biopsy of the lesion based on its appearance, and they found that the AI matched the performance of the dermatologists with impressive accuracy.
“We realized it was feasible, not just to do something well, but as well as a human dermatologist,” explained Sebastian Thrun, who co-authored the study published in Nature this week, in a statement. “That’s when our thinking changed. That’s when we said, ‘Look, this is not just a class project for students, this is an opportunity to do something great for humanity.’”
Having a piece of software that can perform as well as a doctor in identifying potentially cancerous moles could radically change the diagnoses of these cancers. The team is aiming to produce a smartphone-friendly version of the algorithm. “Everyone will have a supercomputer in their pockets with a number of sensors in it, including a camera,” said Andre Esteva, the study's other lead author. “What if we could use it to visually screen for skin cancer? Or other ailments?”
