Magnetic resonance imaging, MRI, is a vitally important process in medical sciences, but at the same time it is a complicated and takes a lot of time. The patient needs to lie still for 15 minutes or an hour in order to go through the scan, no matter if he is in pain or a child. NYU is trying to speed up this process and Facebook team has joined them. NYU has a vision of reducing the total MRI time by 90% with the help of artificial intelligence, AI, imaging tools.
Before moving forward, it is essential to differentiate this vision from the already present AI based tools in medical imaging field. Once MRI scan, X-ray, has completed, object recognition system can be used to watch for defects that are even missed by the doctor. This effort not only aims on studying the completed scan but also on speeding up the completion process.
MRI is a slow process because it needs to take a lot of 2D images or slices, which further need to be piled up to form a 3D image. If brain tumor is being scanned, lots of time is required.
NYU researchers started the FastMRI project in 2015. They investigated whether it is possible to develop the full scan, of same quality as the traditional scan, but from small data. The whole process will accelerate, if MRI scans only collect the vital information, and, then the trained AI models fill the missing points. The patient will feel relaxed, machine will be able to serve more people in the same time, and the process will get cheaper and less complex.
Three years ago, NYU medicine researchers gave a feasibility report, based on early results, regarding this vision. Now, to quicken the things up, they are teaming up with Facebook AI research (FAIR) group.
“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,” Dan Sodickson, director of the Center of Advanced Imaging Innovation and Research, NYU, said.
Facebook is interested in this project, although they have no plans of beginning a medical imaging arm.
“We’re looking for impactful but also scientifically interesting problems,” FAIR’s Larry Zitnick claimed. He further added that the created scan “doesn’t need to be just plausible, but it needs to retain the same flaws.”
Luckily, present MRI machines are pretty docile, they can be easily told where to take the scans, and, if any area needs to be scanned more or less.
The research is still in its initial stages, although being worked on for three years, but at least NYU report has shown its feasibility. On 20 August, NYU and Facebook announced their partnership and they have faith that others will also join them. Larry Zitnick said that they will be working on this project openly.
The steps they will be following together will be to define the problem, collect the data set and publish it, develop standard and metrics to compute their triumph, and so on.
