MRIs are slow, and sometimes a patient can wait for up to an hour for their scan to finish. NYU’s FastMRI project aims at significantly decreasing imaging processing wait time. Started in 2015, FastMRI generates a MRI scan that only features a portion of what was imaged, and uses machine learning systems to quickly fill in the unprocessed sections, so that the patient is only in the imaging room for half the amount of time. The expedited process allows more patients to undergo imaging procedures in a given day, and images will be cheaper to process. The AI-created images not only need to reflect the patterns and structures that would have been in the original image, but must produce the same kind of abnormalities that would appear in the scan. The program will likely require 10,000 MRI scans to generate sufficiently accurate images.
As a way to expedite the project, which is still in its early stages, NYU is collaborating with Facebook AI researchers to strategize problem solving, record data, and implement baselines and metrics to track their results. The hope is that other organizations and institutions will join them in advancing the technology.
“We have some great physicists here and even some hot-stuff mathematicians, but Facebook and FAIR have some of the leading AI scientists in the world. So it’s complementary expertise,” said Dan Sodickson, MD, PhD, NYU’s director of the Center for Advanced Imaging Innovation and Research.