You’re back from a fun weekend at the beach and want to tag all your friends in the photos you took. That’s easy: Facebook’s facial recognition software does it automatically.
Or say, your doctor just took CAT-scan images of your chest and wants to tag everything that might be a lung tumor. Can computers do that, too?
Jeremy Howard says they can. The data scientist is the CEO of a just-launched company called Enlitic. He’s leveraging the power of a concept dubbed “deep learning” to develop ways for computers to spot injuries and diseases in X-rays, CAT scans, MRIs and other medical imaging, according to MIT Technology Review.
In the quest to make machines smarter, computers scientists have turned away from simply pumping the devices full of more and more facts and toward programming them to look for patterns and make connections on their own. This is deep learning — teaching computers to work the way the human brain does.
Deep learning is helping Google create self-driving cars and IBM make a computer that can win Jeopardy. Enlitic wants to use it to help doctors do their jobs better and faster.
Enlitic’s approach is to feed into a computer hundreds of pictures of, say, malignant liver growths, allowing the computer to identify the common characteristics they share. The computer can then begin to automatically tag tumors in medical images, highlighting areas for doctors to examine more closely.
Part of the venture’s appeal, for Howard, is that it’s a chance to apply data-mining techniques to a problem that’s more worthwhile than helping businesses figure out who your friends are or what you’ll want to watch next on Netflix.
“Data science is a very sexy job at the moment,” he tells Wired. “But when I look at what a lot of data scientists are actually doing, the vast majority of work out there is on product recommendations and advertising technology and so forth.”
Howard isn’t the only one working on ways to apply data science to medicine. Stanford University computer scientist Daphne Koller is training computers to analyze breast cancer scans. Her invention, Computational Pathologist (C-Path), also learns as it goes, figuring out which characteristics are the most important to look for.
Not only can Koller’s C-Path catch breast cancer tumors, it can also analyze them to predict patient survival rates.
Howard says his goal is not to replace radiologists and other doctors who examine medical images, but to help them be more efficient.
With deep learning computers on the rise in medicine, it may only be a matter of time before computers are tagging not only your vacation photos but your X-rays, too.