Skip to Content

Facebook and NYU Want to Make MRIs 10 Times Faster. How? With AI.

It seems there’s no shortage these days of big tech companies swerving out of their lane to enter the quagmire of American health care. Many of these firms, such as Apple and Amazon, haven’t exactly been shy about their ambitions, whether through wide-ranging research platforms like ResearchKit or acquisitions of online pharmaceutical delivery services.

This week, Facebook added itself to the list. The company has set the ambitious goal of making MRI scans up to ten times faster than they currently are through the use of artificial intelligence—a project in collaboration with NYU, which appears set on making news left and right recently.

Facebook and NYU School of Medicine succinctly laid out their rationale in a blog post: “[MRIs] are relatively slow, taking anywhere from 15 minutes to over an hour, compared with less than a second or up to a minute, respectively, for X-ray and CT scans,” they wrote. That slow pace presents both public health barriers to entry and downstream effects; for instance, young children who need MRIs may not be able to stay still for all that long, and certain parts of the country aren’t necessarily abound with available MRI machines, thus creating an unnecessary bottleneck due to a cumbersome process.

“By boosting the speed of MRI scanners, we can make these devices accessible to a greater number of patients,” say NYU and Facebook.

How exactly would that process work? Through a project the organizations dub “fastMRI.” This would—or at least the hope goes—harness AI to change the way MRIs are conducted in the first place. Rather than having to form a full, data-intensive MRI image, artificial intelligence could theoretically be used to “fill in views omitted from the accelerated scan,” the groups say.

This, it should go without saying, isn’t a simple initiative. The neural networks that would ostensibly be used to fill in the imaging gaps would have to be extremely precise so that patients aren’t ultimately harmed. Or, as the researchers themselves put it, “A few missing or incorrectly modeled pixels could mean the difference between an all-clear scan and one in which radiologists find a torn ligament or a possible tumor.”

Subscribe to Brainstorm Health Daily, our newsletter about exciting health innovations.