UPDATED 23:31 EDT / JUNE 18 2018

EMERGING TECH

MIT researchers create a neural network that can process MRI scans in under a second

Researchers at MIT say they’ve developed an algorithm capable of processing magnetic resonance images in less than a second, in what could be a crucial development for the healthcare industry.

Currently, doctors need to compare two separate MRI scans taken at different times in order to track changes in the human body over time. However, creating these comparisons is both complex and time-consuming, because the process involves carefully lining up the two images in order to make accurate measurements. Doctors need to process all of the locations on the images to a 3-D map, and current computers take many hours to do this.

The problem with MRI scans is the amount of information they contain. Essentially, each image is made up of hundreds of 2-D images stacked atop of each other in order to create a 3-D map known as a volume, which in turn is made up of 3-D pixels known as voxels. When trying to align two MRI scans, computers have to sort through millions of voxels to determine their location in a new, unified image.

Because the process takes several hours to complete, doctors are sometimes forced to wait hundreds of hours if there is a backlog of images that need to be processed. MIT researchers said it isn’t practical to use computers with greater processing power, so instead they created a new convolutional neural network called “VoxelMorph” that they claim is much better-suited for the task.

MIT’s researchers trained VoxelMorph using more a set of 7,000 publicly available MRI scans of the human brain. Neural networks such as VoxelMorph work by pushing data into the front end and then passing it through multiple nodes that feed into other nodes, speeding up the comparison process. Meanwhile, the network also learns about different common groups of voxels and their anatomical shapes.

“Using a CNN and modified computation layer called a spatial transformer, the method captures similarities of voxels in one MRI scan with voxels in the other scan,” MIT News reported. “In doing so, the algorithm learns information about groups of voxels — such as anatomical shapes common to both scans — which it uses to calculate optimized parameters that can be applied to any scan pair.”

Following this training, MIT’s researchers fed 250 new scans into VoxelMorph in order to test its effectiveness. The algorithm finished processing these scans within just two minutes, compared with the numerous hours it would have taken a conventional MRI analysis program to complete. This test was carried out using a regular central processing unit, but when the researchers used a graphics processing unit instead, the process was completed in under a second.

MIT’s team said VoxelMorph has obvious applications in healthcare and could even change the way doctors perform some kinds of surgery. For example, it could be possible to create a new and up-to-date scan during an operation and use those images for real-time analysis.

Image: MIT

A message from John Furrier, co-founder of SiliconANGLE:

Your vote of support is important to us and it helps us keep the content FREE.

One click below supports our mission to provide free, deep, and relevant content.  

Join our community on YouTube

Join the community that includes more than 15,000 #CubeAlumni experts, including Amazon.com CEO Andy Jassy, Dell Technologies founder and CEO Michael Dell, Intel CEO Pat Gelsinger, and many more luminaries and experts.

“TheCUBE is an important partner to the industry. You guys really are a part of our events and we really appreciate you coming and I know people appreciate the content you create as well” – Andy Jassy

THANK YOU