University of Texas supercomputer speeds real-time MRI analysis
Researchers from the Texas Advanced Computing Center, the University of Texas Health Science Center and Philips Healthcare have developed a new, automated platform capable of real-time analyses of magnetic resonance imaging (MRI) scans in minutes, rather than hours or even days.
By leveraging the Stampede supercomputer at the University of Texas-Austin’s TACC, imaging capabilities of a Philips MRI scanner, as well as the TACC-developed Agave application programming interface, researchers were able to demonstrate the system's effectiveness in using a T1 mapping process, which converts raw data into useful imagery.
The full circuit—from MRI scan to Linux-based supercomputer and back—took about five minutes to complete and was accomplished without any additional inputs or interventions, says William Allen, technical lead for the effort and research associate in TACC’s Life Sciences Computing Group.
“It’s really about the speed and flexibility. The whole point of this is to analyze the data faster,” adds Allen, who notes that Philips Healthcare modified the MRI scanner software to accommodate the “pipeline” to enable fast, accurate image processing. “The platform that we developed gives us the ability to link the scanner to a remote supercomputing resource.”
Funded by the National Science Foundation, Stampede open science computing resource is one of world’s fastest supercomputers and is comprised of a Dell PowerEdge cluster equipped with Intel Xeon Phi coprocessors in an effort to push the envelope of computational capabilities by enabling breakthroughs in advancing computational biology and bioinformatics.
Allen describes the Agave API as a “science-as-a-service” platform designed to capture different kinds of biomedical data in real time and turn them into actionable insights for providers. “It’s the same analysis you would normally do with MRI, except now it’s all automated,” he says. “The way we’ve set it up is we’ve removed all need for human intervention.”
According to Allen, the Agave API ensures that there is seamless communication between the MRI scanner and the Stampede supercomputer. “The real benefit here is the Agave platform, which grabs the data automatically as it comes off the scanner, pushing it and then quickly starting the job, and then pulling the data back once the analysis is complete.”
At the same time, Allen acknowledges that the test cases that the research team has conducted so far are “relatively lightweight,” using about 16 processing cores and up to 20 megabytes of RAM. “We’re at the proof-of-concept stage,” he concludes. “Once we get to more complicated analyses, with automated image segmentation and registration, we’ll use easily up to 200 cores.”
Allen is quick to make the point that the platform with the Agave API “is not limited to MRI and could conceivably be done for any medical device or instrument that gathers some sort of data and pushes it to a computer.”
Researchers presented the platform at last week’s International Conference on Biomedical and Health Informatics in Orlando, Fla., which was co-located with the HIMSS17 conference and exhibition.