Facebook and NYU take ‘important step’ toward AI-enabled, 5-minute MRI scans

Facebook AI and NYU Grossman School of Medicine have created an artificial intelligence system capable of producing rapid MRI scans that are just as diagnostically accurate as images made via traditional techniques.

That’s according to a recent study published in the American Journal of Roentgenology, which trained a neural network to generate images using a fraction of the data typically required for exams. Radiologists reading the resultant fastMRI scans spotted the same abnormalities regardless of whether they analyzed AI-created images or standard MRIs.

What’s more, all imaging experts involved in the research said fastMRIs were higher quality compared to traditional scans. The findings move both parties ever closer toward their long-stated goal of producing MR images up to 10 times faster than current technologies. 

“We are highly encouraged by these results,” said Daniel K. Sodickson, MD, PhD, director of the NYU-operated Center for Advanced Imaging Innovation and Research, which is supported by the National Institutes of Health. “We also encourage others to use the fastMRI data and open-source code to build upon our findings. Together, we will continue to push the boundaries of medical imaging, using AI not merely to replicate tasks performed by humans, but to generate entirely new capabilities—like ultrafast MRI—that enhance the care of patients.”

As part of their research, six musculoskeletal rads read over two sets of knee MRIs from patients evaluated at NYU Langone Health. In total, 108 participants were included and two sets of images were created from each patient: one using standard approaches and another via the fastMRI AI model, which was trained using the “world’s largest” open-source dataset of knee MRIs.

All readers were not told which images they were interpreting with AI and traditional images spaced at least one month apart. Those produced with the accelerated tech proved interchangeable with legacy images, the authors noted.

By using about four times less data, the 3T fastMRI technique can image much quicker and requires less time spent inside the scanner.

Going forward, the collaborators plan to use accelerated MRI with other organs, such as the brain.

“This study is an important step toward clinical acceptance and utilization of AI-accelerated MRI scans because it demonstrates for the first time that AI-generated images are essentially indistinguishable in appearance from standard clinical MRI exams and are interchangeable in regards to diagnostic accuracy,” Michael P. Recht, MD, chair of NYU Langone’s radiology department, said in a statement. “This marks an exciting paradigm shift in how we are able to improve the patient experience and create images.”

Read the entire study published in AJR here.