Magnetic resonance imaging scanners produce images of the body using strong magnetic fields and radio waves to scan several images of the same area. By comparing these images, the scanner reveals even the most subtle abnormalities, such as young tumors.
Considering the nature of MRI scanners, it stands to reason that math might improve the time it takes to get and compare these images.
Electrical engineers and computer scientists at MIT's Research Laboratory of Electronics thought so, and they are publishing an algorithm they have devised that speeds scanning time threefold, reducing the amount of time someone would have to lie still in a scanner from as many as 45 minutes down to 15 minutes. (Their work will appear in the journal Magnetic Resonance in Medicine.)
The key is in choosing which information from each scan to keep, so that the machine does not have to start from scratch with every ensuing scan. In other words, information obtained in one scan can be used to inform subsequent images, providing a basic outline that shortens the time it takes to perform the next one.
If that sounds obvious, consider the complexities of determining which parts of a scan to keep. Translating too much information from one scan to the next could mean losing the ability to detect unique tissue features only revealed by stark contrasts. "You don't want to presuppose too much," Elfar Adalsteinsson, an associate professor of electrical engineering, said in a school statement.
So the algorithm calculates which new information it needs in every pixel to construct the image, and which information it can use from previous scans, such as the edges of tissue types.
While the team says some quality is lost using its method, no algorithm has worked so well at this speed. And presumably, it will only improve.