SIEMENS

Like astronauts on a voyage to a perpetually shrouded planet, scientists at major universities across the United States and Europe have embarked on a mission to make sense of one of the most complex regions of the universe — the 100 billion neurons and 150 trillion synapses that comprise the human brain. Funded with $40 million from the U.S. National Institutes of Health, the effort, which is known as the Human Connectome Project (HCP), is designed to discover the relationship between the structure and function of the brain. “The idea is to decipher as much as current imaging technology can about the wiring of the human brain and about how that wiring contributes to our behavior and to the differences in behavior between people,” says David Van Essen, PhD, Edison Professor and Head of the Department of Anatomy and Neurobiology at Washington University School of Medicine in St. Louis.

“The hypothesis driving this project,” says Dr. Bruce R. Rosen, Professor of Radiology at Harvard Medical School and Director of the university’s Athinoula A. Martinos Center for Biomedical Imaging in Boston, Massachusetts, “is that if we understood the relationship between structure and function we would begin to understand diseases such as autism, and to more effectively treat injuries such as those caused by stroke.”

To accomplish this goal — a vision that, if realized, might revolutionize the diagnosis and management of mental illnesses, neurodegenerative diseases, and brain injuries, researchers must do two things: implement technologies capable of mapping the brain’s three-dimensional architecture both functionally and structurally on a near-micron level, and define — taking into account the vast differences between healthy brains — what is normal. Thanks to the development of two experimental 3-Tesla magnetic resonance imaging (MRI) scanners from Siemens, the first of these goals has been realized, and the second is the subject of a major study centered at Washington University in St. Louis together with the Center for Magnetic Resonance Research (CMRR) at the University of Minnesota Medical School.

One of the scanners*, which has a gradient that is up to 2.5 times stronger than that of clinical MRI scanners, was developed for the CMRR and was recently relocated to St. Louis. (Gradients are used for spatial encoding.) The other scanner, which has a target gradient of up-to 7.5 times that of the latest clinical 3-Tesla MRI scanners, was developed on the request of the Martinos Center in Boston in cooperation with the University of California Los Angeles.

“The Human Connectome Project is designed to help us answer fundamental questions about the human brain,” says Prof. Kamil Ugurbil, Director of the CMRR. “Steady progress in MR techniques such as functional brain imaging (fMRI) and diffusion imaging over the last two decades have made the project possible. However, additional improvements in hardware and image acquisition methods are still necessary — and we are addressing them.”

Breakthrough Technology. Researchers have been imaging neuronal pathways in the human brain for years. Key to this have been technologies such as MR Imaging of anisotropic diffusion of water in the brain, and, most recently, resting state fMRI. Resting state functional MRI and high angular resolution diffusion imaging (HARDI) — a special diffusion imaging technique — are favored at Washington University & the University of Minnesota (WashU-Minn) consortium. On the other hand, Diffusion Spectrum Imaging (DSI) — a general form of Diffusion Tensor Imaging that was pioneered in 2005 by Dr. Van J. Wedeen, Professor of Radiology at the Massachusetts General Hospital and Director of Connectomics at the Athinoula A. Martinos Center for Biomedical Imaging, is being pursued at the MGH/UCLA consortium. Here, the idea is to reveal the fabric-like structure of pathways within each MRI voxel (a voxel is a 3D pixel), thus differentiating between intersecting pathways.

The diffusion imaging approach capitalizes on the fact that water molecules wiggle in tissue and that this motion can be measured as their hydrogen nuclei (protons) produce radio signals in response to radio-frequency pulses in combination with rapidly changing static and dynamic magnetic fields. “Considering the fact that these molecules naturally tend to move along axons — the white matter fibers that connect brain cells — the technology essentially produces an image that duplicates the paths of axons,” explains Wedeen. “When you stitch thousands of voxels together,” he adds, “you wind up with what we call a tract — a white matter pathway. But the important thing is that when two pathways intersect, the system must stitch through them correctly, and DSI is what makes that possible. Thanks to this technology, we are finding that 3D white matter grids are ubiquitous in the human brain.”

Achieving the current state of visualization was not easy, however. It required a huge increase in MRI sensitivity, as well as much higher processing speeds to make sense of the flood of spatial data from water molecules. No commercial scanner could come close to what was needed. The Boston team had already come to the conclusion that the key to higher resolution imaging of neural pathways is not necessarily, as had been previously assumed, higher field strength than 3Tesla of the main magnet, but rather, higher field strength of the much smaller magnetic gradient coils that modulate the main magnet’s field. “The gradient (coil) is the part of the MR machine that encodes the spatial characteristics of water molecules. It tells you where things are,” explains Rosen.

The more powerful the gradient coil is, the faster one encodes the diffusion of water molecules, resulting in a sharper picture of angular resolution, which is the key factor in differentiating between intersecting fibers. “It’s like a camera — the faster the shutter speed, the sharper the picture. The result is that even though we are not directly imaging axons, we can, in effect, see them by inferring how water molecules move,” adds Wedeen.

But why, one might ask, do water molecules follow axonal paths? “The answer,” says Lawrence L. Wald, PhD, Director of the MRI Core Facility at the Martinos Center, “is that since axons are long, thin tubes, water molecules naturally tend to wiggle inside and along them, rather than perpendicular to them. What’s more, they do so in distances that are about the same size as the separations between the axons or about 10 to 20 microns during the time period used to measure our MRI signals. So if we are sensitive to those water molecules, we can be sensitive to the directions in which the axons are aligned.”

Most commercial 3 Tesla MRI scanners for in-vivo clinical imaging have a top gradient strength of 40 to 45 milli-Tesla per meter (mT/m); but Wedeen’s studies on primates had indicated that higher gradient strength than this would be advisable to produce sharp diffusion images of an adult human brain’s wiring patterns. With this in mind, the MGH/UCLA and Minneapolis/St. Louis researchers asked the Siemens MR engineering team, which was led by Eva Eberlein, if they could develop a gradient stronger than anything currently available. “What they came up with was two prototypes that could achieve from two times to nearly eight times the gradient strength. “It was an engineering tour-de-force,”** recalls Wald. With a scalable gradient strength of 80 mT/m up to 300 mT/m, the density of electromagnetic energy can increase up to a maximum factor of 56. **

Nevertheless, this presented a challenge. Knowing that higher gradients would allow higher resolution imaging, more MR slices would have to be acquired to cover the same anatomical region. But doubling the resolution would in turn increase the scan time by a factor of four or more. Therefore, a technology* had to be developed to accelerate encoding and thus reduce scan times. Based on earlier ideas developed by CMRR, such as “Multiband Multislice” imaging, the MGH researchers came up with a refinement called “Simultaneous Multislice,” that allowed the acquisition of several MR slices at the same time, yet would keep them separate with minimal sensitivity loss. The technology “accelerates imaging by a factor of three, but when combined with stronger gradients, the result is actually a factor of four in image acquisition acceleration,” says Wald. “All in all, we have cut the average scan time from about one hour to approximately 15 minutes.”**

Building a Brain Database. While the Harvard-UCLA team’s role in the Connectome project has been the development of a scanner that pushes the boundaries of what is possible with current MRI diffusion imaging technology to the limit, the Washington University / University of Minnesota consortium has not only undertaken methodological developments for improved data acquisition, but also addressed the challenges of understanding functional connectivity in a large-scale study. “Functional imaging approaches, particularly those that utilize resting state fMRI — in which the entire brain is imaged during a period in which the subject is not performing an explicit task — are highly complementary to diffusion imaging since they provide information regarding functional connectivity rather than just hard-wiring,” says Washington University’s David Van Essen, who co-leads the consortium with Prof. Kamil Ugurbil.

Motivated by their high-resolution functional imaging work carried out at an even higher field strength — with a 7 Tesla MR system from Siemens — the CMRR team had been pursuing innovations to accelerate scan times since 2008. Now, in the context of the Human Connectome Project, they turned to adapting these techniques, dubbed “Multiband Multislice Imaging,” to the customized HCP 3 Tesla scanner and managed to speed up data acquisition for functional connectivity by factors of up to nine.

Having overcome these limitations, the team’s next goal is to begin scanning 1,200 genetically-related people with this technology. “The idea behind our side of the project,” explains Ugurbil, “is to acquire data with a previously unavailable level of quality in order to generate a database of brain connectivity patterns and develop tools to perform data mining on that database.”

Additionally, the Van Essen-Ugurbil team will scan many of the subjects at ultrahigh magnetic fields (7 Tesla* ), an approach pioneered by the University of Minnesota group. ”Seven Tesla will be far superior for resting state fMRI and anatomical imaging and is also expected to perform well for diffusion imaging,” says Ugurbil.

Indeed, the project’s emphasis on closely-related people sets the stage for eventually creating and tapping entirely new databases in an emerging field called “imaging genomics.” By exploring the possible connections between imaging information and genetic information, the researchers hope to uncover some of the mechanisms behind mental illnesses. “There are already many published studies along these lines. For instance, some have shown that brain circuits are abnormal — that is, have reduced functional connectivity — in autism,” says Van Essen. “But we are trying to push this research to a different level. Just as the Genome Project opened up a world of bio-informatics, we are looking to open a new world of neuro-informatics that will capitalize on the vast amount of information that is being generated by imaging modalities.”

A first step in that direction has already been taken. In a project that combines data acquisition, data analysis, informatics, and visualization, Van Essen and Ugurbil have developed an interactive composite dataset of the left and right cerebral hemispheres obtained in a pilot study of healthy adults. The images it produces are based on resting state functional MRI, which, in the context of the database, show an average of which brain regions are actually communicating with other specific areas, when one area, represented by a black dot, is probed. Red and yellow sections are strongly related to the seed location. By merely clicking a location in the dataset, a user is interrogating 30 gigabytes worth of data. “This is the first time such a tool has been developed,” says Van Essen. “As our database expands, it will become extremely powerful. But it’s just a preview of what will come out the Human Connectome Project.”