Magnetic resonance imaging (MRI) is the tricorder of our time, according to Mark Griswold, Ph.D., associate professor at Case Western Reserve University, referring to the Star Trek medical device that seemingly performs all diagnostic tests with a mere scan. Although it isn’t as advanced as its Star Trek cousin, emerging MRI devices are breaking the barriers in terms of imaging speed and what may be imaged successfully.
Dr. Griswold, who was one of a number of speakers at the Gordon Research Conference on “In Vivo Magnetic Resonance” in Andover, NH, in July, has developed a method that images data 384 times faster than was possible in the mid 1990s, and one of his colleagues has pushed the boundary to 970 times faster than mid ’90s speed, he says. The approach tends to focus tightly on only pixels that contain relevant information and ignores irrelevant pixels. “In many clinical situations, only about 10 percent of an image has relevant data, but the entire image is processed through the same techniques.”
One approach Dr. Griswold’s lab is investigating uses a library of prior scans to teach the scanner what a typical scan should look like. Afterward, only the portions of an image not recognized on those scans goes to a lossy data-processing channel. The rest of the data is processed by other channels. The goal is to minimize what goes to the lossy channel, thereby dramatically speeding imaging and simultaneously providing higher quality views of the relevant data.
The three broad limitations to speed, he says, are the signal itself, the methodology that encodes data to the image, and power demands. Using conventional encoding, the power needed to achieve a fourfold speed increase would approximate the output of a small nuclear power plant, Dr. Griswold says. To break these accepted speed barriers, he explored compressed sensing, a concept that came from the math world.
The challenge was how to turn this concept into a broadly accepted method to generate clinically relevant data. In the case of angiography, the algorithm was written to image only the aspects of a tissue that changed slowly. Pixels that didn’t change, or that changed quickly, were excluded. “There are tons of processes that applies to.”
Working with Northwestern University, Dr. Griswold is applying his work in fast MRI to arteriovenous malformations to provide a roadmap intended to help physicians locate the malformation. Interventional magnetic resonance is another potential application for fast MRI.
“Ours is sensitive to things that no other modality sees, but there are several areas where characterizing a lesion is still problematic.” For example, in mammograms, traditional diagnostics offer sensitivity of about 90%, while MRI offers sensitivity of 98 to 99%.
Specificity, however, remains problematic. “If you see something with MRI, you oftentimes still must use pathology to characterize it. A clinician has to have some way of quickly getting a sample from the lesion under MRI guidance.” To that end, researchers are developing ways to use MRI to perform minimally invasive therapies including needle biopsies and using radio frequencies to destroy tumors.
Cardiac imaging is another promising application for fast MRI. For example, using MRI, researchers can watch the beating heart, can ascertain—at a range of hundreds of cells per milliliter—whether the myocytes in the myocardium are dead, and, one day, may be able to determine the best course of treatment.
“At the physics level, magnetic resonance can measure almost any property of almost any tissue,” Dr. Griswold says. Whether it should be used, however, is often determined by the time available for a given measurement. So, each time MRI can be sped up, MRI becomes more practical for a wider range of studies.