Well Connected: Mapping the Brain’s Networks

Brain Scan

At roughly three pounds, a human brain accounts for only about two percent of total body weight, but whether it’s concentrating hard or daydreaming, it consumes roughly 20 percent of the body’s basal energy. Just how the brain converts that energy to the staggering variability of thought, personality, emotion, interest, and perception that humans exhibit has been one of science’s obstinate mysteries.

It’s known, however, that much of the brain’s work occurs in orchestrating the intricate networks that connect its processing units. In fact, “Individual variability in brain connections underlies the diversity of our thinking, perception and motor skills . . . the connections are most of what makes us who we are,” says neuroscientist David C. Van Essen, PhD.

Now in its fourth of five years, the Human Connectome Project (HCP) is mapping those connections and revealing the brain’s interconnectedness in all of its detail. The project’s broadest goal, according to Van Essen: “describe the full pattern of long-distance connections between each brain region and every other brain region.”

The HCP is a consortium of more than 100 scientists and staff at 10 institutions in the United States and Europe — including a large contingent of Mallinckrodt researchers and clinicians — that is scanning the brains of 1,200 normal subjects, analyzing the imaging data and publishing it on the Internet for interested parties to use at no cost. As the project proceeds, the team works to improve every aspect of brain imaging.

Van Essen, Alumni Endowed Professor and head of the Department of Anatomy and Neurobiology at Washington University School of Medicine, is joined by Kamil Ugurbil, PhD, director of the Center for Magnetic Resonance Research and the McKnight Presidential Endowed Chair Professor at the University of Minnesota, as co-principal investigators of the $30-million-plus project. Funding comes from 16 National Institutes of Health (NIH) institutes and centers and the McDonnell Center for Systems Neuroscience at Washington University.

Conventional wisdom about brain function held that discrete regions, 150 to 200 in each hemisphere, were dedicated to particular tasks such as math or language. It was believed that they did their jobs and sent their answers on for further processing, Van Essen explains.

The revamped thinking is that networks of those same regions — collections of cortical areas — form and communicate to do the brain’s work. “It’s exquisitely coordinated, with some elements more active at any moment,” says Van Essen. Imagine the globe with its political subdivisions, each with its individual interest. Where the significance occurs and the work gets done is in their various and changing interactions.

So, how to sort out the daunting complexity of the brain’s connecting cables, those tens of billions of axons that link many of its 90 billion nerve cells? Enter imaging technology with several sophisticated, noninvasive approaches: diffusion imaging (dMRI), functional magnetic resonance imaging (fMRI) and magnetoencephelography (MEG).   

Like most magnetic resonance imaging, dMRI reveals the activity of water molecules. Unconstrained, those molecules bounce around randomly in tissue. But confined within the walls of an axon, they assume directionality, somewhat like water in a hose, explains Joshua S. Shimony, PhD, MD, associate professor of radiology. By recording that movement, images of the axons’ paths through the brain’s white matter (so named because of the white myelin that ensheaths the axons) can be generated, and long-distance connections can be mapped.

Among the imperfections of dMRI is its resolution: In this application, the technique resolves down to about 2 millimeters, roughly 8/100 of an inch. A single axon is less than one percent of that size, though they vary dramatically in caliber. So dMRI is capable of seeing only groups of axons, and some information is lost. Fortunately, axons bundle together as they connect nerve cells in one region to those in another, says Shimony.

Also difficult is deciphering the path where bundles of axons cross one another. Did all of the axons go straight, or did some turn left to join a crossing bundle? Van Essen says one bundle looks like another and that the team is working to improve these analyses as it builds the connectome.

Much of the dMRI analysis is being done at England’s Oxford University, by a group recruited for its experience and expertise in the evaluation of neuroimaging data.

Focused on the processing units in the gray matter of the brain where higher functioning occurs, fMRI plots those regions that are turned on and those where activity is dialed back. Busy regions of the brain require more glucose and bloodflow, Shimony says, and fMRI generates what is called the BOLD signal — for Blood Oxygen Level Dependent — to illuminate relative activity.
The technique does not reveal the axons that connect areas of high activity but rather the active areas themselves. Because particular parcels are working together, the long-distance connections are inferred. By combining this data with data from the dMRI images, the connectome map becomes more complete.

Working together, the Washington University and University of Minnesota groups have reduced the time needed to acquire a single brain image from several seconds to just seven-tenths of a second, one of the consortium’s valuable advances to brain studies.

Consortium members at Saint Louis University and institutions in Europe contribute to the effort via their expertise with magnetoencephelography (MEG), a technique that records magnetic fields produced in areas where high levels of electrical activity correspond to increased brain function. The resulting images provide high-speed, detailed data about activity, but spatial resolution — precisely where the activity is occurring — is more difficult to interpret. Overlaying the MEG and MRI data increases the map’s information density.

To obtain data about particular brain functions, much of the fMRI data is collected as subjects perform tasks in the scanner. Deanna Barch, PhD, professor of psychology in Arts & Sciences and professor of radiology, leads the work developing quantifiable tasks that require the collaboration of specific brain regions.

“We’ve developed seven domains,” she says, “more than most previous brain studies.” Included are sensory motor tasks (finger tapping, for example), emotional tasks (recognizing and responding to faces expressing emotion), working- and long-term memory tasks and language tasks, among others. Comparing the brain activity of all subjects to the tasks being performed will help answer the question: “Are there differences in behavior related to differences in patterns of connectivity and/or activity?” asks Barch.

And to get yet another angle on brain activity, investigators record images of the brain in a resting state. About 20 years ago, it was learned that even while relaxing, wildly different and distant parts of the brain busily operate in synchrony and that networks clearly link them. By analyzing resting state data, Van Essen says the team gets “one more way to inspect the elephant in the dark room.”

Barch also has provided behavioral tests that subjects undertake prior to scanning: personality evaluations, cognitive performance exams, gauges of vision, hearing, smell, and taste, grip strength measurement, and many others. All of the data on individual performance are analyzed to relate it to brain connectivity.

The scans and attendant information produce huge amounts of data to be processed, analyzed and published. When complete, the human connectome will comprise as much as 1 petabyte of digital information, equivalent to about 1 million gigabytes. Just to accommodate digital storage, the project has invested $1 million. According to Daniel S. Marcus, PhD, assistant professor of radiology and director of the Neuroinformatics Research Group at Washington University School of Medicine, results on 500 subjects will be available in the spring of 2014.  

Among the many processing tasks, including eliminating noise from the raw data, one of the team’s biggest challenges has been the creation of an atlas, a sort of average of the scans to which any new scan can be compared. The atlas requires spatial registration of the brains’ parcels for every subject. “Everyone’s head is different, and they may be in the scanner in a slightly different position, so the raw data doesn’t align perfectly,” Marcus says in describing the job. When processing is complete, the resulting atlas is a matrix of every point’s connection strength with every other point in the normal brain.

Marcus also oversees publication of HCP findings. De-identified results are available to everyone — no advanced degree required — at the project’s website: humanconnectome.org. There, a visitor will find:

  • ConnectomeDB — a data management platform that allows users to identify groups of subjects (left-handers, for example) and download data for these groups. The HCP informatics team is currently developing data mining tools to allow users to analyze data and test hypotheses online.
  • Connectome Workbench — a downloadable, interactive tool for visualizing maps of connectivity that talks to the database.
  • Connectome in a Box — the huge package of all of the lightly processed data.

Says Marcus, “The Human Connectome Project represents a major advance in sharing brain imaging data in ways that will accelerate the pace of discovery about the human brain in health and disease.”

Barch, who also contributes to subject recruitment and selection, says that more than 500 volunteers have been scanned to date, and the project is on track to finish on schedule.  

Ideally, each volunteer family includes a set of identical or fraternal twins and at least one other sibling, all ages 22 to 35. “This gives us the most powerful design for determining genetic effects on quantitative traits,” she explains. Already, the model has confirmed that even identical twins have measurably different brain connections and that brain circuits are heritable, some more than others.

Subjects travel to St. Louis to spend two intense days in testing and four scanning sessions. “These individuals make a huge contribution to science,” says Barch. Their efforts make it possible to understand the healthy human brain and also “enable future projects that probe the changes in brain circuits that underlie a wide variety of brain disorders affecting humankind, including autism and schizophrenia,” says Van Essen.

As the HCP expands to include younger and older subjects and its data reaches more users and they imagine more applications for it, the impact on understanding and health made by the generous volunteers will continue to grow.