New Technology with Clinical Applications

Scientists at Pittsburgh Supercomputing Center, Carnegie Mellon University and the University of Pittsburgh Medical Center have created a powerful new technology for viewing the brain at work. Using high-speed networks to link an MRI scanner with a supercomputer, they've made it possible to convert scan data almost instantaneously into an animated 3-D image showing what parts of the brain "light up" during mental activity.

"Using the CRAY T3E and high-speed networking," says PSC neural scientist Nigel Goddard, "processing that used to take more than a day takes about 10 seconds, and we're working to get it under a second."

Download larger version (230K).

This image shows what regions in a subject's brain were involved in a memory task. This kind of study leads to improved understanding of "working memory", which affects clinical treatment of schizophrenia and amnesias.

Typically, techniques that provide pictures of the functioning brain involve a substantial delay, a day or more, between gathering data and the availability of high-quality 3-D images. The Pittsburgh team has cut this to seconds.

"We expect that this technology will set the stage to use brain-mapping as a clinical tool in diagnosis and treatment of brain pathology," says Dr. Jonathan Cohen, who codirects the Laboratory for Clinical Cognitive Neuroscience, a joint venture of the University of Pittsburgh and Carnegie Mellon. Real-time capability will aid neurosurgeons in precision surgical planning, and it can be used to test and diagnose cognitive dysfunctions such as schizophrenia, amnesia and epilepsy. With high-speed networking, doctors at locations distant from the MRI scanner can actively consult in patient testing.

Mapping the Brain

For several years, Cohen and his colleagues have used a technique known as functional MRI (fMRI) to do "brain-mapping" experiments that investigate and map the brain regions involved in a particular kind of memory activity known as "working memory." Data from an MRI scanner shows what sites in a subject's brain are active during mental activity. These experiments generate huge amounts of information quickly, and initially it took days to process the data into a high-resolution 3-D image.

Download larger version (305K).

This 3-D image of a subject's brain shows that the primary visual cortex becomes activated while the subject, who is inside an MRI scanner, looks at the whirling pattern.

To eliminate this bottleneck, the researchers turned to PSC. Carnegie Mellon statistician William Eddy and UPMC physicist Doug Noll worked in collaboration with Goddard and PSC research programmer Greg Hood to exploit the CRAY T3E, a highly parallel system that divides the computing among many processors. In November 1996, the researchers reduced processing time so that a realistic 3-D image of the brain could be viewed live, while the subject was in the scanner, with a delay between mental activity and image availability of about six minutes. The team has now cut this delay to seconds, and they are working to get it under a second, which will allow an improvement in image quality.

In demonstrations of this real-time brain-mapping capability, a test subject - one of the researchers - lies inside an MRI scanner at the University of Pittsburgh Medical Center and performs a simple mental task. The MRI scanner records data from her brain and transmits it via high-speed network to PSC's CRAY T3E, which converts the raw fMRI data into 3-D images, compensates for head movement and identifies active areas of the brain. From the T3E, the data travels to a remote location via high-speed network, where observers see the subject's brain as a translucent animation showing what regions "light up" as she does the mental task.



Movies of the Brain at Work



University of Pittsburgh
Medical Center

The MRI scanner records data from a region of the brain and transmits it via high-speed network to PSC's CRAY T3E, where a series of computations converts the raw MRI data into a 3-D animated format. First, mathematical data from the scanner is transformed to a representation of physical space. Next, the data is registered to compensate for head movement. Finally, the data is statistically analyzed to determine regions of activity. The 3-D representation is transmitted via high-speed network and "volume rendered" for screen display on the visualization platform.

Download a single graphic version of this schematic (50K).


Remote Location
High
Bandwidth
Network
High
Bandwidth
Network
Reconstruction
Fourier
Image
Head
Movement
Correction
Determine
Activation
Regions

Researchers: Jonathan Cohen, University of Pittsburgh and Carnegie Mellon University.
Nigel Goddard, Pittsburgh Supercomputing Center.
Bill Eddy, Carnegie Mellon University.
Doug Noll, University of Pittsburgh.
Hardware: CRAY T3E
Software: User developed code
Keywords: brain, cognitive neuroscience, functional MRI (Magnetic Resonance Imaging), imaging technology, brain-mapping, working memory, neurons, AIR (Automatic Image Registration) software, brain disturbances.

Related Material on the Web:
NGI Events - Netamorphosis
NGI Events - Netamorphosis poster 4
Blue98 - Human Brain Mapping
Functional MRI Activities at PSC
Live Functional MRI demonstration at Supercomputing 96
Laboratory for Clinical Cognitive Neuroscience
Projects in Scientific Computing, PSC's annual research report.

References, Acknowledgements & Credits


© Pittsburgh Supercomputing Center (PSC)
Revised: September 2, 1997

URL: http://www.psc.edu/science/Goddard/goddard.html