Interested in reading more?

New machines and approaches are offering neuroscientists unprecedented access to the working human brain By Douglas Steinberg Photo: Neil Michel/Axiom Sylvia WIRED FOR AN IMAGE: Research associate Valerie Clark gets her brain waves recorded by Ron Mangun, a psychology professor at the University of California, Davis. Mind-reading, that staple of science fiction, is inching closer to science fact, thanks to steady progress in the field of brain imaging. In the last few years, neuroimagers have established a new technology while fine-tuning an older one. They are now building more powerful machines and beginning to make headway against their ultimate challenge--determining the order in which brain regions become activated as a person thinks, moves, senses, and learns. Until now, brain scans have been subject to the neural equivalent of physics' Heisenberg uncertainty principle: They could detect the areas or the timing of neural activation, but not both. "It's a really exciting field," says Stephen Koslow, director of the division of basic and clinical neuroscience at the National Institute of Mental Health. While cautioning that "we're not going to get some early definitive answers," he adds, "We all believe there's great potential and are quite happy to fund" grant proposals involving brain-imaging technology. Purchases of brain scanners, which cost more than $1 million apiece, are not covered by federal grants, but the National Institutes of Health's National Center for Research Resources has funded up to $400,000 worth of improvements to existing machines. The two main technologies used in brain-imaging research are positron-emission tomography (PET) and functional magnetic resonance imaging (fMRI) with blood-oxygen-level- dependent (BOLD) contrast. PET machines first appeared in 1974. There are now about 360 worldwide, most of which are used clinically, according to Ronald Nutt, technology director of CTI Inc. in Knoxville, Tenn., which produces such units. Seiji Ogawa, a biophysicist at AT&T Bell Laboratories in Murray Hill, N.J., discovered fMRI-BOLD in 1988 (S. Ogawa et al., Magnetic Resonance in Medicine, 14:68-78, 1990). The current prevalence of this technology is impossible to gauge, but Diar Shipman, a marketing manager at General Electric Medical Systems-Americas in Waukesha, Wis., says that about 800 of the MRI machines installed by his firm in the last three years had BOLD capability. A PIONEER'S VIEW: Joy Hirsch, a pioneer in the use of fMRI technology at Memorial Sloan-Kettering Cancer Center, sits among the images of her work. With its longer history, PET boasts the richer scientific literature, but fMRI has prompted much of the recent buzz in the imaging field. Nurtured by researchers at a handful of institutions in the early 1990s, it has become a common tool only since 1995. "Just about every major medical center is trying to come up to speed," says Joy Hirsch, an fMRI pioneer at Memorial Sloan-Kettering Cancer Center in New York. A major reason for the spread of fMRI is that it can be run on the magnetic resonance imagers already owned by hospitals. But additional hardware and software are needed, and the expertise to run these systems is not yet widely available. In addition to neuroscientists, physicists and computer professionals are being tapped as more federal dollars pour into fMRI research. Opportunities in the smaller PET field are expected to grow as a result of recent federal legislation. The Food and Drug Administration (FDA) Modernization Act of 1997 incorporates provisions (sections 121 and 122) prompting FDA to "come up with less burdensome [PET] regulations," says Jane Axelrad, associate director for policy at FDA's Center for Drug Evaluation and Research. Both fMRI and PET rely on the fact that more blood is routed to brain regions where neurons are active. The signals triggering increased blood flow are unknown, and reasons for the increase are hotly debated. It is still unclear, for example, whether active neurons need more blood-delivered oxygen (M. Barinaga, Science, 276:196-8, 1997). Researchers using PET typically inject water or butanol containing oxygen-15 into the bloodstream. In the blood, the isotope ejects positrons that collide with their anti-particles, electrons, creating high-energy photons. These travel to radiation detectors placed around the head, and the resulting images, color-coded by computers, indicate areas of increased blood flow. A more convoluted process generates the fMRI signal. In areas of neural activity, blood levels of oxyhemoglobin increase since active neurons use little if any of the extra oxygen available to them from heavier blood flow. A corresponding drop in deoxyhemoglobin levels, however, weakens an effect that a strong magnetic field has on the "spin" of hydrogen protons in and around the blood vessel. A magnetic resonance imager detects a higher resonance signal that results from the weakening of that effect. The outcome of fMRI "is not an image that relates in any way--at least, in any direct way--to a physiological parameter" like blood flow, says Peter Fox, director of the Research Imaging Center at the University of Texas Health Science Center at San Antonio. Other neuroimagers note the artifacts that can afflict fMRI studies. For example, some brain areas, such as those at air-tissue interfaces, don't generate signals. FUNDING AVAILABLE: Stephen Koslow of the National Institute of Mental Health says the agency is eager to fund research using brain-imaging technology. Nevertheless, Fox, a veteran PET researcher, concedes that "the more that we can move work onto fMRI, the better. It's more widely available, and it doesn't require radiation." He himself uses fMRI to identify "task-induced" changes in the brain. Unlike Fox, a number of former PET researchers have switched their allegiances entirely to fMRI. Though scientists in academia and industry agree that PET is the best imaging tool for examining metabolism and drug binding in the brain, some of its drawbacks are unavoidable. The use of short-lived radioisotopes is expensive (an on- site cyclotron is needed), and experiments are cramped because subjects can be injected with only limited amounts of radioactive tracer. IMAGES OF CHANGE: Peter Fox, director of the Research Imaging Center at the University of Texas Health Science Center at San Antonio, uses fMRI to identify task-induced changes in the brain. PET is less sensitive than fMRI, meaning that its signal-to-noise ratio is lower--there's more static--and that its images suffer from poorer spatial resolution--they're blurrier. Newer PET scanners and techniques exhibit improved sensitivity, and a machine making its debut this summer may largely erase present imaging deficiencies. Developed by CTI and Siemens Medical Systems Inc. of Iselin, N.J., the High Resolution Research Tomograph (HRRT) should be shipped to a European laboratory this June, according to CTI's Nutt. This noncommercial system will have 119,808 detectors, far more than the 18,000 found in PET machines now in common use. Each detector will harbor a crystal of lutetium orthooxysilicate, a fast-acting, high-output scintillator. Current scanners have 4 mm resolution, but HRRT should be able to distinguish active brain areas that are 2 mm apart. Compared with current machines, HRRT should be four times more sensitive to radioactivity, and its signal-to-noise ratio should double. To achieve super-scanning, fMRI researchers have been steadily boosting the strength of the magnetic fields to which they expose their human subjects. Most fMRI systems generate a 1.5-tesla (T) field, but Kamil Ugurbil, a radiology professor at the University of Minnesota in Minneapolis, notes that many 3-T systems and a handful of 4-T setups are already in operation. The sensitivity afforded by a 4-T system allowed a team, including Ugurbil, to see thin columns in the visual cortex that respond to one eye or the other (R.S. Menon et al., Journal of Neurophysiology, 77:2780-7, 1997). High magnetic fields offer another advantage. In studies using 1.5-T fields, subjects perform a task repeatedly so that researchers can detect a clear signal by averaging the faint signals from each trial. Yet subjects' performance may not remain consistent across trials, and the brain may respond differently after tasks are repeated. "We have demonstrated that you can see single events at high magnetic fields," says Ugurbil. "Subjects do one execution of a task, and you have a temporally resolved functional-imaging datum as a result of that [4-T trial]." Ugurbil is now installing a 7-T magnet at his research center. HIGHER POWER: Radiology professor Pierre-Marie Robitaille will coordinate a new program at Ohio State that will use a scanner that boasts an 8-tesla magnetic field. An 8-T scanner recently went online at Ohio State University College of Medicine. In October, Pierre-Marie Robitaille, the radiology professor who coordinates that facility, spent 10 minutes inside the 77,000-pound, liquid-helium-filled magnet when its field strength was 7.35 T. "I was fine," he recalls. "Once I was in the magnet, I couldn't sense that I was in it at all." In animal studies, high magnetic fields have not yet been found to pose a safety problem. According to Robitaille, the 8-T scanner contains the world's highest-field magnet for human scanning. Despite improvements in gadgetry, both fMRI and PET suffer from an intractable drawback. They depend on an increase in cerebral blood flow that starts within two seconds of stimulation and takes seven seconds to reach a peak, notes Mark Cohen, an associate professor in the brain mapping division of the University of California, Los Angeles, School of Medicine. Communication between neurons, however, occurs over tens of milliseconds, not seconds. Like long-exposure photographs, fMRI and PET consequently can't establish the order of the events that they depict. To sidestep this problem, neuroimagers have evolved strategies that involve existing technologies and combine the results through complex calculations. Using PET, George Mangun, a psychology professor at the University of California, Davis, and his colleagues localized brain activity associated with spatial attention. Then they clocked that activity at 80 to 130 milliseconds by recording event- related potentials, which are averaged from electroencephalograms (EEGs) taken from the scalps of subjects (H.J. Heinze et al., Nature, 372:543-6, 1994). Mangun is now trying the same strategy with fMRI instead of PET. THUMBS UP: A research volunteer is set to undergo an fMRI brain scan at the Center for Neuroscience, University of California, Davis. EEGs promptly detect electrical currents in the brain, while magnetoencephalographs (MEGs) promptly register the magnetic effects of these currents. Recently, Anders Dale, a Harvard Medical School radiology professor, and his colleagues showed words to subjects and, using fMRI for localization and MEG for timing, found that brain activity in the visual cortex began 100 milliseconds after word presentation. The activity spread within 50 milli- seconds to frontal and temporal lobes of the cortex. "By combining the two methods, we get the movie, if you will, of activation," says Dale. Combining methods, however, does not strike Keith Thulborn, a radiology professor at the University of Pittsburgh Medical Center, as "an intrinsically elegant approach." Preferring to "get both spatial and temporal information out of a single signal," Thulborn has adapted fMRI so that it detects magnetic changes associated with sodium ions, not hydrogen protons. These ions generate signals almost immediately upon moving across the membranes of active neurons. Thulborn has gotten sodium imaging to work when subjects see an image repeatedly flashed for 10 minutes; he is now trying to show it works with brief stimuli. Last winter, his lab also succeeded in generating phosphocreatine images of the brain. To secure better temporal resolution, Gabriele Gratton and Monica Fabiani, assistant professors in the psychology department at the University of Missouri in Columbia, have dispensed with fMRI and PET altogether. Their $40,000 system shines near-infrared light on a subject's head. After the light has passed through the brain, it is detected by 16 optic fibers placed against the skull. Neural activity up to 1.5 centimeters below the surface of the cortex scatters the light, generating an immediate signal, according to Gratton. The spatial results yielded by his technique agree with those provided by fMRI, while its temporal findings are consistent with event-related potentials, the EEG-derived method (G. Gratton et al., NeuroImage, 6:169-80, 1997). "The major challenge for these techniques is to go beyond the idea that the brain has individual structures that do individual things," says Memorial Sloan-Kettering's Hirsch, who collaborated with Gratton. "We must look at the integration of those individual little functional departments in terms of coordinated assemblies that actually do complex tasks. That is the very exciting challenge of the next few years." Douglas Steinberg is a science writer based in New York.

Brain Imaging Assumes Greater Power, Precision

Apr 12, 1998

Interested in reading more?

Become a Member of