The Facebook Artificial Intelligence Research group (FAIR) has teamed up with the NYU School of Medicine on a new project that could make magnetic resonance imaging (MRI) scans 10 times faster.

"MRI scanners provide doctors and patients with images that typically show a greater level of detail related to soft tissues — such as organs and blood vessels — than is captured by other forms of medical imaging," FAIR's Larry Zitnick and NYU's Drs. Daniel Sodickson and Micheal Recht wrote in a Facebook blog post. "But they 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. These long scan times can make MRI machines challenging for young children, as well as for people who are claustrophobic or for whom lying down is painful. Additionally, there are MRI shortages in many rural areas and in other countries with limited access, resulting in long scheduling backlogs. By boosting the speed of MRI scanners, we can make these devices accessible to a greater number of patients."

FAIR, founded in 2013, is a Facebook-affliated research group that partners with universities to study machine learning and artificial intelligence, often in areas with no direct connection to social networking. According to the post, the research group was actively looking for a high-impact project, while the NYU School of Radiology's Center for Advanced Imaging Innovation and Research (CAI2R) was looking for additional AI expertise to actualize a theory the group has been investigating since 2016.

Part of what makes an MRI machine slow is that it must capture many different scans individually that are then reconstructed into cross-sectional or three-dimensional images. Researchers believe they can teach the MRI to take fewer scans and then extrapolate the missing data using machine learning. 

"This approach is similar to how humans process sensory information," Zitnick, Sodickson, and Recht wrote. "When we experience the world, our brains often receive an incomplete picture — as in the case of obscured or dimly lit objects — that we need to turn into actionable information. Early work performed at NYU School of Medicine shows that artificial neural networks can accomplish a similar task, generating high-quality images from far less data than was previously thought to be necessary."

Researchers are using a dataset from NYU that has been stripped of identifying information. It includes 10,000 clinical cases and around 3 million magnetic resonance images of the knee, brain, and liver. No data from Facebook is involved in the study.

The researchers plan to open source the work so others can benefit from the research immediately. In the future, this same method could be adapted to accelerate or otherwise improve other kinds of medical imaging.