Neuroimaging

2010: A Brain Odyssey
The last decade saw an explosion of neuroimaging innovation and research; now it’s time to put it to use

By Dan Krotz

Sidebars:
Imaging the mind’s image of the body
This is your brain on fear

A 50-year-old recovering alcoholic hasn’t touched a drink in three months. Suddenly, a tantalizing image of a martini hovers in front of him. After several seconds, the drink disappears, and he finds himself staring into the off-white interior of a PET scanner. Later, Dr. Nora Volkow of New York City’s Brookhaven National Laboratory examines his scan for abnormal metabolic activity. She doesn’t have to look long; his orbital frontal cortex had barely budged. Six months ago, the martini image had sent this area into overdrive. Addiction, it seems, is somehow mediated by the orbital frontal cortex [Fig. 1].

At Harvard University, a young woman closes her eyes and visualizes the letter C. She’s asked a few questions. Are there any curved lines? Any enclosed spaces? Later, Stephen M. Kosslyn, Ph.D., a professor of psychology, puzzles over several functional MRI scans taken during the test. Remarkably, the same area devoted to vision and perception, the primary visual cortex, is activated during mental imagery. This prompts the question: How much of what we see is reality, and how much is a product of brain activity?

Elsewhere, researchers are imaging schizophrenics while they memorize words and geniuses while they solve equations. They’re determining how the brain adapts to disease and how it reacts to a baby’s cry. Slowly, one cognitive quirk at a time, they’re seeing the neural fingerprints of what makes us human.

And they are just beginning. Kosslyn compares today’s neuroimaging technology to computers in the early ‘70s. Back then, everyone in an office hovered around one machine with a stack of punch cards; today, there’s a personal computer on every desk. In the future, imaging platforms will be smaller and more widespread, he said. And neuroimaging won’t occur in a vacuum. It will be plugged into the growing body of genetic and psychosocial research, enabling scientists to connect genes and brain function with how we think and feel, Kosslyn said. So what does the future hold, and how will we get there?

It’s 2005. A psychiatrist is deciding how best to treat a clinically depressed patient [Fig. 2]. Drug treatment is an option, but which type of drug? Each targets a different part of the brain. Before deciding, he refers the patient for a PET scan, which reveals that portions of the patient’s basal ganglia aren’t functioning properly. The patient is in luck, as a newly released drug targets that particular dysfunction.

So predicts Diana J. Vincent, Ph.D., an associate professor of radiology and neurological surgery at the Medical University of South Carolina’s Center for Advanced Imaging Research. Her rationale is simple: If you know how a healthy brain works, you can detect anomalies and possibly fix them. The same can be said for educating children or learning a second language late in life. The more we know about how the brain wraps itself around new tasks, the better we can guide those processes.

Getting there, or making sense of the millions of blips, hot spots, and electrical charges storming through the brain will require prototypical neuroimaging labs like Vincent’s. It’s new—only five years old—it’s multidisciplinary, and it’s multinational. There’s a Russian neurophysiologist, a Korean psychiatrist, and a small task force of mathematicians, computer gurus, neurosurgeons, and physicists. The lab was established to merge these disciplines and push functional imaging as far as it can go, Vincent said. They’re off to a strong start, having cataloged activation patterns associated with depression, speech, anxiety, and alcoholism.

Mapping The Mind

Jump to 2008. A 30-year-old man who has suffered a seizure is wheeled into the emergency room. He’s stable but confused; he’s never had a seizure before.

The radiologist on call conducts an MR scan. She suspects the right frontal lobe, but everything appears normal.

Stumped, she goes online and enters her password into a secure server. She compares the patient’s scan with a composite brain image culled from several thousand healthy people. Compared with the morphed image, her patient’s scan looks slightly abnormal. Delving deeper, she types in a brief description of the patient: 30-year-old male, left-handed, two years of college. A matching scan pops up, another composite image of several people with similar characteristics. She immediately sees that her patient’s cortex is 0.5 mm too thick in one area. Perhaps this is the cause of the seizure.

It’s a probabilistic atlas of the human brain, and it’s the goal of the International Consortium for Brain Mapping funded by the National Institutes of Health. The wide-ranging group of research labs is acquiring high-resolution structural and, in some cases, functional MR images of 7000 people in seven countries. They’re also collecting behavioral and demographic data: what those people eat, their education, parents’ background, medical history. The result is a vast database of both the human brain and the human experience that can be manipulated to provide a composite brain snapshot of a person of any given age, ethnicity, and experiential background.

“We call it the human phenome project,” said Dr. John C. Mazziotta, director of the University

of California, Los Angeles Brain Mapping Center, referring to the fact that both genes and environment shape physical attributes.

Mazziotta, lead investigator of the nearly completed 10-year project, said clinicians may someday compare ambiguous brain scans with an appropriately matched scan collated from their database—statistically speaking, the most normal of normals. Need a benchmark scan to help interpret an MRI of a bilingual 45-year-old Asian woman? Just type in the variables and receive a probabilistic-based image that accounts for the normal variance in the population.

While Mazziotta is busy melding thousands of MR scans into the quintessential brain, David Van Essen, Ph.D., a self-described cortical cartographer and department head of anatomy and neurobiology at Washington University, is busy unrolling, flattening, and inflating the cerebral cortex [Fig. 3]. The rationale behind his computer-aided investigation is the fact that the cerebral cortex is actually flat, roughly the size of a medium pizza. And the way in which it’s scrunched into the skull varies from person to person.

“If you take two beach balls and crinkle them up, there will be similarities in how they’re folded, but also important differences. Nature, with the cortex, is the same way,” Van Essen said.

Unfortunately, nature’s vagaries can confound image interpretation. Two people undergoing the same functional scan might evince what appears to be different activation patterns simply because their cortexes are folded differently, not because different areas are activated. The best way to address these quirks, according to Van Essen, is to gently inflate the cortex, smoothing its features like a baby’s brain, or flatten it into his much-favored pizza shape.

It’s a simple solution that’s come a long way since he first drew pictures of monkey cortexes 20 years ago. Today, thanks to a $1.4 million grant from the National Institute of Mental Health, Van Essen grafts fMRI data onto structural MRI images, which are then rendered into surface-based maps of the brain. In this manner, person-to-person cortical variances are ironed out, and the precise location, size, and shape of functional areas can be teased from the folds of the brain, he said.

The next step is to use this technique to pinpoint cortical differences between mentally ill and healthy

people, Van Essen said. Another possibility is to explore why some people do better than others at certain tasks. Just as there’s interpersonal variance in how the cortex is folded, there can be a two-fold difference in the size of functional areas from person to person. These differences can be cataloged and possibly correlated with different talents and skills.

“Ultimately, we can explore what makes us human, and what makes us unique,” he said.

Like cartography, brain mapping involves charting the lay of the land and determining which areas are devoted to hearing, emotions, and memory, for instance. This analogy works to some extent, but it’s too simplistic. Joy Hirsch, Ph.D., a professor of neuroscience at Memorial Sloan-Kettering Cancer Center, is among a new group of researchers who are discovering that functional areas aren’t areas at all, but rather intricate networks. She believes that in five years, systems of remotely connected brain areas will be viewed as the fundamental units governing cognition.

Hirsch is qualified to know. Her lab is the O’Hare Airport of functional imaging, churning out roughly 300 fMRIs each year for presurgical and research applications. Along the way, she’s learned that many cognitive tasks require several brain regions to work as one system. She’s also identified one region that may play a role in several seemingly unconnected task-related systems. Consider three common tasks: naming an object, determining whether two things are the same or different, and multiplying four by 12. All require a unique network comprising at least three areas, and all share activity in the left medial frontal gyrus, Hirsch said.

Ultimately, understanding such cognitive systems will better enable physicians to repair neural breakdowns. If a patient can’t move his arm, for example, don’t suspect just the motor strip; the problem could lie in several preplanning areas or in the thalamus, Hirsch said.

Bringing It All Together

It’s a few years later, maybe 2010. A functional scan of a stroke patient reveals damage to Broca’s area of the brain. The physician plugs this information into a software-driven model of brain function to find out how other areas of the brain will respond over the next few months. According to the model, which employs the same supercomputing wizardry meteorologists use to predict hurricanes, several associative areas will shoulder the workload normally handled by Broca’s area. With this information, physicians can jumpstart the rewiring process, shepherding the patient through a series of mental exercises designed to stimulate new neural connections.

Just where are we in understanding the brain? Naturalists spent the better part of the 19th century cataloging life, painting pictures of ferns and finches and giving them Latin names—a dry line of work called taxonomy. But only after this exhaustive fieldwork was completed could they ponder what it all means. Darwin spent five years traveling the globe on the H.M.S. Beagle and returned home to write The Origin of Species. Likewise, neurologists have been imaging the brain in action for 20 years. They have the fieldwork data, they have a good idea of what many areas of the brain do, but they don’t yet know what it all means.

“We’re a little beyond taxonomy. Now it’s time to take this raw data and understand the rules that govern how networks operate,” said Dr. Gregory V. Simpson, director of the dynamic neuroimaging laboratory at the University of California, San Francisco.

Simpson believes the next big step is learning these neural bylaws, which can then be used to develop predictive models of brain function. The way to get there, he said, is by using every imaging tool available: PET and fMRI, as well as several other techniques not typically associated with clinical radiology. One is electroencephalography (EEG), which measures patterns of electrical activity emanating from neural pulses, and another is magnetoencephalography (MEG), which measures the faint magnetic field encircling the brain, also a product of neurons’ electrical activity.

Because these two modalities directly detect the brain’s electrophysiological chatter, they offer split-second temporal resolution, an important advantage given that neural activity takes place on the millisecond scale. By comparison, PET and fMRI measure the lumbering byproducts of neural activity such as increased metabolism and blood flow. MEG and EEG detect not only precisely when a neuron fires, but how it fires. Is it a series of wavering oscillations, or is it a long, sustained pulse?

“We’re detecting the very nature of neural activity,” Simpson said.

Take a basic thought process, such as reading a digital watch. We spend about half a second on this task, or 500 µsec in brain time. In that split-second, PET and fMRI can detect which areas are involved, but unlike MEG and EEG, they can’t detect whether activation in area A precedes area B, and whether there is any feedback between the two.

Unfortunately, both EEG and MEG collect surface measurements of phenomena that occur deep in the brain. An algorithm translates the surface data into brain activation patterns, but it’s an imperfect calculation, like pinpointing precisely where a pebble falls in a lake by measuring the ripples as they wash ashore.

The trick, then, is to fuse the split-second temporal data from MEG and EEG with MRI’s excellent spatial resolution [Fig. 4]. Neurologists at Huntington Memorial Hospital’s Epilepsy and Brain Mapping Program have used this coregistration technique to localize the neural origins of seizures to areas the size of a postage stamp at a time resolution equal to one-twentieth the blink of an eye, said Dr. William W. Sutherling, the program’s director. But there’s still a long way to go.

“Combining MRI and MEG is in its infancy,” said Richard M. Leahy, Ph.D., director of the University of Southern California’s Neuroimaging Research Group.

The next step, he said, is to improve MEG’s spatial resolution by developing algorithms that better correlate the magnetic field on the surface of the head with what’s taking place in the brain.

In the meantime, researchers will rely on what’s out there. Marcel Just, Ph.D., codirector of the Center for Cognitive Brain Imaging at Carnegie Mellon University, has developed a computer model of sentence comprehension based on fMRI data as well as decades of cognitive research. The program is broken down into hundreds of task-related modules that correspond to areas of the brain. As it runs through a sentence, each module mirrors what its neural counterpart does. Type in the simple sentence, “The pen is on the table,” and the program parses it out much like the brain: left-to-right, each word building on the other, until it incrementally creates a representation of the sentence’s meaning.

“It’s a computational model that gives us enormous powers to explain cognition,” Just said.

And although pushing a six-word sentence through a computer is small potatoes compared with predicting the aftermath of stroke or developing the most efficient way to learn a second language, it’s a start.

Sidebars:
Imaging the mind’s image of the body
This is your brain on fear