Advertisement
New Scans Enhance "Colin's Brain" Baseline Data Set

High-resolution, single-subject template expanded to allow crossover use for both cadaver and live subjects
John Watson

Colin Holmes, medical marketing manager at SGI, and voluntary subject for an enhanced baseline data set of the human brain.

There are probably better ways to spend five days in Montreal. During the week of July 22, 2002, Colin Holmes, medical marketing manager at SGI, voluntarily underwent numerous inversion recovery MRI scans of his brain, was injected with radioactive substances for a positron emission tomography (PET) scan and subjected himself to x-rays for a computed tomography (CT) scan, along with enduring a two-hour series of functional magnetic resonance imaging (fMRI) scans at the McConnell Brain Imaging Center.

A glutton for punishment? Not exactly. Today, experts are busy aligning these new, highly detailed studies of his brain with 27 already existing MRI scans taken in 1993, known in international brain imaging circles as "Colin's Brain." That existing data set, already widely regarded as the highest quality single-subject template in the world and referenced in the book MRI Atlas of the Human Cerebellum (Academic Press, 2000), will, with the alignment of the newly completed scans, become a super-data set destined for even wider use in brain research circles around the world.

Such data sets have many functions, but their key role is to serve as a baseline against which neurologists can compare scans of patients' brains in order to identify areas of abnormal growth or decay, as well as identify regions important to normal brain function. Until now, the "Colin's Brain" data set couldn't be used as a template for cadaver brains, as brain tissues that are frozen or preserved exhibit different image characteristics than living ones - but the just-completed inversion recovery MRI scans will make this data set useful in working with cadaver brains for the first time. This crossover capability was one of the key motivators for Holmes to undergo the daunting new set of scans, which were made at the request of cadaver brain expert Professor Karl Zilles of the Heinrich-Heine-Universität in Düsseldorf, Germany.

It's no coincidence that Holmes works for SGI, as the connections between this ongoing project and SGI are plentiful. In brain-mapping research centers, scientists from many disciplines rely on scalable, high-performance supercomputers. The parallel computation capabilities of the SGI Origin family supercomputer are utilized to align the brains of different individuals. Visualizing the extremely large volume data sets collected in Dr. Zilles' lab (approaching 600 gigabytes per brain) requires the high bandwidth, large texture memory and visual fidelity of the SGI InfiniteReality4 graphics subsystem, part of the SGI Onyx family of graphics supercomputers.


These images from a magnetic resonance imaging (MRI) scan are typical of the high-resolution data forming the "Colin's Brain" data set. (Courtesy of Drs. Colin Holmes, SGI, and Alan Evans, Montreal Neurological Institute, McGill University.)

SGI supports similar work on a global basis, including research at UCLA's Laboratory of Neuro Imaging, where Holmes has tackled similar problems of large, complex data visualizations that arise when comparing groups of individual neuroanatomies. The current alignment project relies on both an SGI Origin 3800 server and an SGI Onyx 3400 visualization system at the Montreal Neurological Institute, part of McGill University.

"Fifteen years ago, under my direction, the McConnell Brain Imaging Center chose SGI to support all computing," explains Center Director Dr. Alan Evans, a professor at the Montreal Neurological Institute. "It was a natural fit, since brain imaging is inherently three-dimensional, and SGI has always had inherently 3-D technologies. The 'Colin's Brain' data set — which has proven to be so helpful to so many researchers and physicians around the world and now promises to gain even wider acceptance due to the new scans — wouldn't have been possible without SGI technologies."

As a youngster growing up in southern Ontario, Holmes didn't exactly set his heart on having his brain become one of the world's key baseline data sets. In fact, it would be virtually impossible for anyone to orchestrate such a phenomenon, even if they wanted to. As Holmes puts it, "I didn't plan for its adoption. I just wanted to have more accurate brain anatomy for my graduate studies, and getting my own brain scanned back in 1993 was a step toward that goal. The data set was first adopted throughout my lab, and, before I knew it, by people who became aware of it around the world."

Today a distinguished scholar who has co-authored more than 50 abstracts and papers in addition to fulfilling his medical marketing role at SGI, Holmes holds a B.Sc in biology from the University of Guelph in Ontario and a Ph.D. in neuroanatomy from McGill University in Montreal. It was at McGill that Holmes first saw the need for a definitive baseline brain data set.

"As I began postdoctoral studies in the early '90s, I made the transition from working on microscopic study of brain slices to MRI scans of living subjects, and I was struck by how much lower-resolution the scans of living people were," he recalls. "And for good reason: high-resolution studies require absolute immobility for hours, a luxury we can never have with live subjects."

It occurred to him that then-current visualization technologies pioneered on advanced visualization equipment could make it possible to align multiple MRI scans to create a single data set approximating the resolution of a single very long scan. Visiting the center's MRI rooms after hours, he endured 27 scans, aligned them, co-authored a paper about his targeted experiment and thought that would be the end of it.

But, before he knew it, a neurologist in Liverpool had successfully used the data set as a baseline against which to compare a brain scan of a patient who had suffered a stroke. By overlaying Holmes' high-quality brain image onto the patient's image, the physician was able to define the anatomy that had been affected in the patient.

Word spread. "Soon people started using it not only for diagnoses but also as a general example of what is possible with MRI technology," explains Holmes. "And then, of course, there was the book."

In 2000, Julien Doyon, then a visiting scientist from Laval University in Quebec and now a psychology professor at the University of Montreal, was studying the cerebellum in collaboration with Jeremy D. Schmahmann, an associate professor of neurology at Harvard Medical School. They adopted "Colin's Brain" to complete an atlas of this finely structured body at the base of the brain. Their co-authored work brought the data set to the attention of yet more brain-mapping researchers around the world.

The July 2002 scans augment the older MRIs in unique ways. Now, in addition to the soft-tissue anatomy of the original data set, the new data set will include hard structures, such as bone (obtained from the CT scan), metabolic function (from the PET scan) and anatomy in action (from the functional MRI scans). The inversion recovery scan, which provides a halfway point between hard- and soft-tissue studies, is the key scan enabling first-ever use of the data set as a baseline for cadaver studies.

After alignment with the existing data sets, the enhanced "Colin's Brain" will be sent to Germany, where it will be used as an intermediary between the worlds of living brain function and Zilles' extensive library of postmortem human brain anatomy. Eventually, it will be the subject of a peer-reviewed paper to be co-authored by Holmes.

Where will it all end? "There are many other single-subject templates out there, and there's nothing special about my brain itself," says the Holmes. "But, the truth is, no other set has the high image quality that mine does. "So," he adds with a small grin, "slowly but surely, I guess my data set is being adopted just about worldwide."

John Watson is PR Programs Manager, Sciences & Energy, at SGI, also known as Silicon Graphics, Inc. He can be reached at editor@scimag.com.

Advertisement
Advertisement