Glimpses of the Living Brain
Encased in a bony vault and wrapped in layers of tough membrane, the brain has long been considered the most inaccessible of human organs. For centuries, this spongy mass that controls an astonishing array of functions was off limits to medical examination except in the most dire circumstances: grievous injury, surgery undertaken as a last, desperate measure, or postmortem dissection. In all these instances, of course, the brain was very far from its usual level of functioning, and so it was never clear how much these studies could contribute to an understanding of the “normal” brain. Surgeries, autopsies, and neurological examinations of people who had survived brain injury or stroke offered rich details, but the full view of an awake, behaving human brain has continued to tantalize and elude researchers until very recently. In only the past 20 years, options for medical imaging of the brain have increased tremendously, using everything from magnetic waves to charged subatomic particles to computer algorithms as a means of allowing clinicians and researchers to “see” the living brain at work.
THE BRIGHT IMAGES OF PET
Of the many imaging techniques in use today, each has its limits and its strengths. Such factors as cost, labor intensiveness, and fineness of detail in the image help to determine the best uses for each form of imaging, whether in research or in clinical practice. One technique that shows promise in both areas, but that has made a name for itself thus far primarily in research, is positron emission tomography, or PET.
A PET scan measures the distribution in the body of a radioactively “labeled” substance that the patient has received shortly before the scan. Adding a radioactive label to compounds such as glucose (the main source of energy for the brain) permits researchers to the monitor metabolism—roughly, the rate at which energy is being used, or the rate of activity going on—at particular sites. Brain cells take up the radioactively tagged glucose for use; the glucose is metabolized, and the transiently radioactive atoms remain inside the cells, giving off positively charged particles. These “positrons” quickly collide with nearby electrons, then give rise to gamma radiation, which can be detected outside the body and mapped by a computer.
The resulting image of the brain, which is often enhanced with color to make the different values easier to see, quite clearly distinguishes those parts of the brain that are using more glucose from those that are using less. Thus PET can show where nerve cells are more active and where they are less so, not only in cases of disease or disorder but during a particular task or thought or emotion. As an even more flexible measurement of function in the normal brain, PET can detect changes in local blood flow. Small blood vessels respond very rapidly to the needs of nerve cells, axons, and dendrites, so that a full scan's worth of information can be gathered in 40 seconds, as compared with the 45 minutes required to measure glucose metabolism.
Although the PET scan may be familar to most readers as a series of still pictures, the technique is in fact highly dynamic and well suited to giving an ongoing picture of moment-by-moment changes in the working brain. In this aspect it borrows from several earlier techniques that were developed to
show ongoing processes in the body—for example, autoradiography, which used a radioactive tracer to make visible the circulation of blood or the metabolism of energy sources in the brain.
Researchers must also have some independent means of siting the information from a PET scan, perhaps with a few landmarks for comparison, if the moving, changeable images traced by the computer are to have any accuracy with reference to a living three-dimensional brain. Therefore, before a PET scan, the patient or the research subject may be fitted with a mask and given an x-ray. The skull x-ray is useful in two ways: it reveals the overall anatomy of the individual 's brain, which will allow for more precise orienting of the PET images, and it can be entered in a computer and “molded” to fit a standardized version of the brain, so that the information from a particular scan can be applied toward a more general understanding of brain anatomy. The resolution of the x-ray is good enough—down to a few millimeters —that researchers can confidently identify a small focus of activity within a subarea (such as the primary visual cortex).
Pinpointing the site of activity is not the only challenge posed by a PET scan. There is also the need to distinguish between slight but meaningful shifts in activity and random irregularities, or background “noise.” To take the primary visual cortex as an example, a bright light flashed in the subject's eyes evokes an unmistakable peak of activity visible in a PET image; but a less drastic stimulus, such as a word or two appearing on a screen, will act on nerve cells in the brain much less vigorously, and the corresponding PET image will be harder to read. Moreover, the response may be dispersed among several areas, perhaps those having to do with learning and memory, language, and emotion, as well as the areas containing the visual receptors that must take in the stimulus to begin with. To find subtle responses that might otherwise be overlooked, PET researchers borrow the statistical technique of aggregating the data from a small group of subjects and averaging the responses over the group as a whole. Many of the random signals thus cancel out one another, and the responses that are meaningful show up more clearly.
Armed with the technology to observe local patterns of ac-
tivity in the brain as they are taking place and a statistical approach for uncovering some of the more subtle signals, PET researchers began in the 1980s to look for research questions worthy of their new technological capabilities. The laboratory of Marcus Raichle, at Washington University School of Medicine, is one that chose to examine how the brain handles language—a fundamentally human feature.
As with any inviting area of research, the first task was to define specifically what would be studied. From the new perspective of the PET scanner, many appealing interrelated aspects of human language abilities called for investigation. If any progress were to be made in unraveling the larger riddles, it could only come from focusing on a series of smaller questions, one or two at a time. So, rather than having subjects perform a complex, multilevel task such as reciting the days of the week—which would call for the active participation of at least half a dozen separate brain areas, even if carried out “unthinkingly”—Raichle and his colleagues decided to start at the other end of the scale. They would begin with as simple a stimulus as possible and build up to more complex stimuli. In fact, the first step was to offer a stimulus that had nothing to do with language but was simply a shape on a screen. This would provide a baseline, showing what the PET scan would look like when the subject was merely sitting at rest under experimental conditions, watching. The levels of activity measured by PET at this stage could be subtracted from all later stages as having nothing to do with the processing of language.
The next stage was the simplest possible presentation of language: asking the subject just to look while words appeared on a screen. The visual cortex is active for this task, as are several other regions, and the pattern of activity that appears is much more extensive than that elicited by showing a plain nonverbal shape. The newly active regions are indeed known to handle language; yet, they clearly cannot comprise all of the language areas, for when words are presented for the subject to hear, rather than to view, the parts of the brain that respond are completely different. Still other areas become active if the task is expanded so that the subject does not merely see or hear the word but must produce its meaning.
Raichle's group took a “building-block” approach to this
riddle, choosing to separate out the purely visual aspects of the brain's response to the appearance of a word on a screen. The researchers presented subjects first with arbitrary symbols having the characteristic shapes and angles of letters; next with real letters grouped into units about the length of average words but unpronounceable; then, with strings of letters that looked like normal words and that could be pronounced but happened not to be words in the English language; and finally, with authentic English words. A typical sequence might look like this:
The goal was to single out very specifically, according to the subjects ' responses rather than any preconceived schema, the elements of the brain that responded to written language (as opposed to those that analyze visual patterns, for example). The findings were dramatic: an area known as the medial extrastriate cortex on the left side of the brain came smartly into action as the tasks reached the third and fourth levels (those of pronounceable nonwords and of real words). This was consistent with findings, from a number of earlier studies, that brain injuries at this site interfered with the ability to read words, although they often did not affect speaking or even writing.
Such a pattern of activity, however robust, presumably must be learned rather than innate. One would hardly suppose that the brain is genetically programmed to respond to the sight of graphic symbols that look like words, in English or in any other language. What this means for the researcher is that it may be possible, in long-term studies, to observe the first stirrings of such a response in infants, small children, those just learning the rules of reading, and so on. Ultimately, the goal is to explain just how the ability to comprehend visual word form develops in this exceedingly plastic, versatile territory of the human brain. The benefits to be derived from understanding this process could be very great for the millions of Americans who struggle with dyslexia and other learning disorders.
In addition to exploring the visual aspects of reading, Raichle's
group also designed a building-block series of tasks dealing with the sounds of words (see Plate 1 ). First, subjects were simply shown pairs of words; then they were shown pairs of words and asked to judge whether they rhymed. The subjects also had to read aloud words that appeared on the screen at the quick pace of one per second.
As the tasks gained in complexity, additional areas of the brain became active. And still further areas became active for tasks that required subjects to turn their attention to the meaning of the words: not only a few sites in the left frontal lobe, in or near the so-called association areas that process many modes of information, but also —and this was a surprise—several sites in the cerebellum. This finding was unexpected because the cerebellum is best known for many functions that are far removed from language: the coordination of movement, fine manual skills, repetitive physical tasks, and so on.
Strikingly, the areas that are active do not remain at their original high levels but actually change and rearrange themselves as the subject becomes more familiar with the task. In terms of behavior, in the course of learning a particular task, there is clearly a point at which the subject begins to make fewer errors and perform the task more fluently. PET scans reveal that internally, too, the brain' s functions follow a dynamic course, marking what appears to be the same threshold of familiarity. The left frontal sites and the cerebellum drop out of action as the task is mastered.
These observations begin to suggest new ideas about how the brain at first tackles and then masters a new task—with implications, perhaps, for methods of teaching in a wide variety of subjects that call for well-learned performance.
MEDICAL IMAGES OF EMOTION
Positron emission tomography is not limited to images of learning and motor function. Because it permits close observation of how the brain is using energy, for whatever function, PET actually allows researchers to map what may be called the anatomy of an emotion, in terms of activity patterns in the brain. A topic well suited to this approach is the study of
panic disorder, because the behavioral and physiological aspects of the disorder have been well characterized and provide consistent criteria for study. In addition, with the consent of intrepid volunteers who suffer from this disorder, a panic attack can actually be brought on at will in the laboratory, under controlled conditions, by an infusion of sodium lactate. The advantages of laboratory observation are several: the scanning equipment can be set up in advance and timed so as to obtain the maximum information from the panicking subjects in the shortest possible time; and many extraneous factors that might otherwise appear to be associated with a panic attack can, in this neutral setting, be ruled out.
For the Washington University group, eager to see what PET could reveal about panic disorder, the first question was: How does “normal ” anxiety (that which would be felt by most individuals, to varying degrees, in a stressful situation) compare with the overwhelming anxiety that is experienced for no discernible reason in a panic attack? The researchers elicited “normal” anxiety in subjects not affected by panic disorder by telling them that they were about to receive an electric shock. (In fact, the measurements were first taken and then a small electric charge was delivered afterward, simply to keep the procedure credible for the subject—otherwise there would be no anxiety and nothing to measure.) The PET scans of normal subjects who went along with this procedure showed patterns of activity in the brain that were much like those of an attack of panic disorder. It is as if the brain produces the state felt as “panic” by one main route regardless of the cause, be it external or internal.
Why then do feelings and physical symptoms of panic mount to the level of a near-crippling disorder in some individuals? PET scans offer some clues about the physical basis of a predisposition toward panic attacks. In a resting, nonpanic state, the brains of panic-disorder sufferers consistently show an asymmetry of blood flow and oxygen use in the two halves of the brain, with the left hemisphere measuring lower on both counts. The site of this asymmetry is known as the parahippocampal gyrus; with its close connections to the hippocampus, it figures largely in the processing of emotional states (see the discussion of the limbic system, the “emotional brain,” in Chapter 2 ).
The particular form of the asymmetry, too, which gives the effect of increased metabolism in the right hemisphere over the left, is intriguing in itself, because a number of other studies have suggested that the right hemisphere is strongly responsible for mental arousal, attention, anxiety, and physiological readiness to respond. Further studies using PET imaging and behavioral measures as well as other approaches should improve our understanding of the physical basis for anxiety, and also of specialization in the two hemispheres of the brain and the physiological circuits underlying emotion.
Other mental disorders are also open to investigation with PET, but they may call for a different experimental approach, because they offer no known “control state.” In other words, there may not be a way to reliably induce the characteristic emotion in the laboratory, as sodium lactate can reliably bring on a panic attack in a subject who suffers from panic disorder. Raichle and his co-workers have begun to look at depression with PET; they obtained PET scans from a number of people who had been diagnosed with chronic depression and from a psychiatrically screened, nondepressed population. In comparing the two groups, the researchers found that depressed subjects exhibited less activity than normal subjects in a deep region of the brain known as the caudate nucleus; in a small area of the left frontal cortex and in the limbic system in general, however, activity was increased. The findings were quite specific and corresponded well with other evidence that has turned up in the past few years about physiological factors in depression.
More of a surprise, however, and suggestive of new avenues of investigation was the next experiment. The researchers brought “normal” nondepressed subjects into the laboratory and asked them to think about something sad. This time the resulting PET scans resembled those of the depressed patients: they showed the same site of increased activity in the left frontal cortex. If the same local pattern of reduced activity in the brain holds true both for long-term clinical depression and for transient feelings of sadness, this observation may point toward new ways of exploring both the disabling ailment of depression and the physiological basis for healthy feelings of sadness.
THE X-RAY TODAY
Not only PET but virtually every form of medical imaging can be understood as an attempt to relate an object in space (whose size might range from the entire brain down to a single cell) to events occurring in time (from the human life span down to fractions of a second). The workhorses of medical imaging for most of this century have done a good job on one or the other of these dimensions, addressing either space or time with great precision. X-ray technology, for example, produces highly readable images with fine resolution of detail down to 0.1 millimeter, but these images are static and show only existing structures rather than ongoing processes. On the other hand, electroencephalography, or EEG, which detects small shifts in electric potential at the surface of the skull, measures overall activity of the brain in real time but indicates only roughly where in the brain the information is coming from. (Nevertheless, EEG has produced “images,” in the form of characteristic patterns of waves, that are useful for study of an aroused or relaxed mental state, of certain phases of sleep, and of a few disorders such as epilepsy and brain tumors.)
The structure of the brain is so intricate, with its folds and fissures, overlapping connections, and underlying compartments, that a clear view of the spatial organization of these 1,400 or so cubic centimeters is crucial for understanding the many ways that disease can affect the brain. For a long time, conventional x-rays fell short on this criterion, because to produce an image on a sheet of film, they necessarily reduced a three-dimensional structure to two dimensions. But the newer version of x-ray imaging gets around this problem with the help of a computer.
Computed tomography uses many x-ray transmissions, sometimes more than a thousand, to produce an image. Each transmission passes an x-ray through the brain at a slightly different angle; integrating all this information is the job of the computer. Because each exposure gives a “tomogram,” a rotating view centered on a single axis through an anatomical structure, the three-dimensionality of the body can be pieced together by summing up the different views. Thus, clear, sharp images can be built up by the computer from many tomograms, with unmistakable distinctions among the brain's gray matter or cortex,
its white matter (made up largely of myelin sheaths enclosing the nerve branches), and its fluid-filled ventricles.
The practical limitations of computed tomography are built into the technology itself, and it appears unlikely that they will be overcome in the near future. For one thing, the resolution of detail in areas smaller than about 0.5 millimeter is precluded by the focal spot of the x-ray tube, which can never be reduced entirely to a point, and by the x-ray dose itself, which must be kept to a healthy minimum. Another limitation is that x-ray tomograms can show only the comparative densities of various structures; thus they may be very helpful in disease states that affect the overall size or shape of body tissues, but they offer little information on maladies that leave the boundaries of a particular structure unchanged. Within these limits, computed tomography will continue to be used widely for imaging the brain, particularly for possible diagnoses of brain hemorrhage, stroke, or disorders involving the cerebral ventricles.
CREATING PICTURES FROM MAGNETIC WAVES AND SOUND WAVES
A trio of imaging techniques are based on the highly specific ways in which different cells of the body respond to magnetic waves. The best known of the three techniques is magnetic resonance imaging (MRI), in which the patient is actually positioned inside a large magnet (see Plate 3 – Plate 5 ). The magnetic field acting on the patient is just powerful enough to bring the magnetic poles of the hydrogen nuclei in the body into alignment. When a single strong pulse of radio waves is fired, the nuclei are knocked into disarray but then realign under the influence of the magnetic field. As they do so, the various types of cells emit a distinct radio signal of their own, which a computer transduces into a visual image.
Different kinds of tissue are distinguished by MRI not on the basis of their density, as with x-ray imaging, but according to the levels of hydrogen they contain. Since hydrogen is found most often as a component of water in the body, the image is largely of structures with different water contents. Tissues with little water content appear bright. Thus fatty tissues show up as bright areas, whereas fluid-filled spaces appear quite
dark. MRI produces highly readable images in any plane desired, which is especially helpful for an irregularly shaped mass such as the brain. The contrast among various tissues is often stronger than with x-ray images, although the resolution of detail is not as fine, perhaps about 2 millimeters on the average. Because gray matter has many fluid-containing cell bodies, whereas white matter has more fatty tissue, MRI shows a clear distinction between gray and white matter. The technique is particularly suitable for diagnosing and monitoring the course of such a disease as multiple sclerosis, which affects the fatty myelin sheath enclosing the nerves. MRI is also good for defining the precise location and extent of tumors.
Similar to MRI, and using much of the same equipment and procedures, is magnetic resonance spectroscopy (MRS). Here the magnet and radio waves are “tuned” for atoms other than hydrogen. Phosphorous atoms, for example, form part of phosphate molecules, which are involved in energy metabolism. Phosphorous MRS can thus be used to measure changes resulting from diseases of muscle cells that involve energy metabolism. Under- or overactivity of some areas of the brain could ultimately be used to warn of a disorder detectable by MRS. After making a diagnosis, the clinician could continue to use MRS to keep an eye on the course of a disease and the effect of drug therapy. MRS has also been used as a research tool, yielding new information on Alzheimer's disease, schizophrenia, autism, stroke, and the normal development and aging of the brain.
One of the latest applications of magnetic signals is in magnetic source imaging, which takes advantage of fact that every discharge of electricity is accompanied by a magnetic field, albeit often a weak one. This technique locates the sources of the very weak fields that accompany the electrical firing of nerve cells in the brain. The extremely sensitive detectors of magnetic source imaging are arranged around the patient's head to track varying levels of activity in many locations of the brain. At present, magnetic source imaging is most effectively used to locate the point of origin of epileptic seizures. Precision is the chief concern here; often, further seizures can be prevented altogether by excision of the tiny area of the brain identified as the focal point. Surrounding areas are able to compensate for any loss of function.
One other imaging technique, ultrasound, has a rather spe-
cific application for the brain. A noninvasive technique, like all those discussed here, ultrasound offers the additional advantage that the form of radiation it uses—acoustic waves—is widely agreed to pose virtually no threat of side effects. Ultrasound has become familiar to many women during pregnancy, because it is often used as a means to examine the fetus in the uterus, without adverse effects. The technique works by directing a pulse of acoustic waves at some body structure and then measuring the strength and speed of the waves that bounce off the body and return as echoes. Assuming an average speed for sound waves through body tissues (about 1,540 meters per second), the ultrasound device assembles the various echo times into an image. Organs and connective tissue tend to reflect sound waves quite differently, so that the outline of structures can be seen distinctly. Ultrasound can also produce useful images of blood flow and of structures that are moving, with a fairly high resolution of about 0.5 to 1 millimeter.
Sound waves penetrate very poorly through bone, and this fact might appear to make ultrasound imaging of the brain impossible. However, in one special case—the newborn baby, in whom the bones of the skull are quite thin and have not yet fused together—ultrasound can be invaluable, offering a chance to examine the brain without surgery.
Each imaging technique presented in this chapter has its own set of advantages and drawbacks that determine its best applications. For instance, the comparative approach described earlier for panic disorder and depression and their functional equivalents in everyday experience is a way to use the young technology of PET to full advantage. Even more powerful as an aid to research is the combination of PET with other forms of imaging. Although no technique currently available can do justice to the whole picture, each one offers a unique glimpse into the intricacies of the living brain.
Chapter 3 is based on presentations by Marcus Raichle.