Facebook AI
“If you were sitting in an MRI, you’d heard that loud noise it makes when it collects data,” Dr. Dan Sodickson, a researcher at NYU Langone Health, told Engadget, “it’s the raw data that a picture of magnetic resonance is off … and that raw data actually resembles this fascinating starburst ”(See above.) That k-space data is stored in temporary storage and once it is full, the scan is complete and the data undergoes a Fourier -transform to actually plot the spatial frequency and generate an MRI image (below).
Facebook AI
“The MRI collects information about the entire image and then, in principle, that frequency information is transformed into spatial information, almost like a prism,” Sodickson continued. ‘So if you take a prism and you separate the colors, then to the left everything will be blue, to the right everything will be red. That’s the transformation we’re doing … we take all the different frequencies and we sort them out. And when you do that – boom – comes out, your trusted image. ”
But instead of waiting for k-space to fill, fastMRI only needs 25 percent of the data that traditional MRI machines need to generate the same images (below). To be clear, this neural network does not analyze existing MRI images at faster rates, it actively generates them from the raw data itself and they are effectively identical to traditional scans.
Facebook AI
Facebook has recruited six radiologists to examine two sets of MRI sequences from a patient’s knee, one from a traditional MRI, the other using fastMRI. “The study found that there were no significant differences in the radiologists’ evaluations,” per a Facebook post on Tuesday. “Five of the six radiologists could not properly determine which images were generated using AI.” Some give that sixth radiologist an increase.
“We wanted to start with a large dataset so we did not end up with too many,” Nafissa Yakubova, a researcher at Facebook AI, told Engadget. ‘That we had, I think, thousands of MRI cases out of the knee,’ as well as a repository of MRI brain scans, each containing as many as 800 still images, to use in the training of the fastMRI model.
Not only will this system help reduce the stress of people who may be sluggish when they spend an hour in a box-sized cylinder that converts their hydrogen atoms into spicy radio transmitters, but it can also enable hospitals to carry more patients. serve.
“Not every institution, every hospital, every country has an abundance of MRI machines, so a lot of the time you have people waiting to scan,” Sodickson said. “I want to reduce that burden.”
What’s more, the system works with existing MRI machines – there is no need to do anything behind it, because this is all just software, it can be installed like a DLC. “Because it is one that is open, any manufacturer can currently access and use it for further testing,” Yakubova said. Of course, device manufacturers will still need to obtain FDA certification before implementing it.
