The visualization of the functional organization of the brain by electromagnetic or optical methods

The search for the "engram is a mapping expedition. It requires "maps that chart both anatomy and function. Anatomy per se only rarely tells us what specific brain structures do. This is evident, by the way, from some neuroanatomical terms that mean only a fruit or a sea monster (e.g. "amygdala, "hippocampus). Surely if early anatomists would have known something about what these structures do, they would have called them by other names. Indeed, some idea on function could occasionally be obtained from tracing the connections between various sites, e.g. a pathway from the olfactory bulb to the piriform cortex does suggest that the latter deals with olfaction, not vision. But the insight into function obtained this way is still limited. Anatomical brain mapping was born in antiquity (Galen 2nd century ad; Thorndike 1923). It underwent a revolution in the nineteenth century with the development of microscopy and histology (Brazier 1988; history fans might also wish to see DeFelipe and Jones 1988). The anatomical cartography of the brain has reached new heights in the second half of the twentieth century. This was made possible by the introduction of sophisticated tracing methods, based on specific molecular probes such as lectins (proteins that bind sugars characteristic of specific cells and cellular compartments), radioactive tracers, and enzymatic reactions (Brodal 1998; Cowan 1998).

Functional brain mapping has a long history as well: the same Galen mentioned above was said to have already noted the effect of brain lesions on behaviour (Thorndike 1923). This approach had gained popularity among brain scientists in the nineteenth century; so was the study of the behavioural consequence of stimulation of loci in the brain by electric currents (Brazier 1988; Finger 2000). Almost a century later, the introduction of the microelectrode has already permitted the analysis of the response of single nerve cells in the brain to specific sensory "stimuli. This has yielded the first detailed functional brain maps (for a "classic example, see Hubel and Wiesel 1977). However, the term 'functional neuroimaging' nowadays does not commonly connote the construction of functional brain maps by meticulous analysis of the response of single neurons, nor the inference of function from anatomical or metabolic lesions, but rather the visualization of the function of large terrains of the brain by electromagnetic or optical methods.

Functional neuroimaging is based on electrophysio-logical methods; or tomographic methods (tomos, a 'cut' in Greek, so called because these methods involve construction of three-dimensional images from planar images, or 'cuts'); or optical methods. The various methods will be explained below. Three points should be noted at the outset, which reflect on the contribution of the functional neuroimaging methods to the analysis of learning and memory in brain. First, these methods differ in their spatial and temporal resolution (Figure 30). Second, they differ in the nature of the signal that they measure; some detect electrical activity directly, others only haemodynamic and metabolic changes secondary to electrical activity. And third, these methods differ in the degree of invasiveness, which means that some could be safely applied to behaving volunteers, others to more daring volunteers, yet others to immobilized laboratory animals only.

1. Electrophysiological imaging. The earliest functional neuroimaging method was electroencephalography (EEG), which measures electrical potential differences among locations on or in the brain as a function of time and place. (The abbreviation 'EEG' is used both for the method and for the records that it generates, 'electroencephalograms'.) The first to report that spontaneous electrical activity can be recorded from the scalp of animals was Caton (1875). The effect of physiological conditions on these patterns of activity was further investigated by his contemporaries (e.g. Danielvsky, see Brazier 1988). But it took several decades before EEG had captivated the world's attention. Berger (1929) pioneered EEG in humans, including on his own son, and discovered a rhythmic oscillation of electric potential with a frequency of about 8-12 Hz, associated with relaxed wakefulness. Adrian and Matthews (1934) replicated Berger's findings and localized the source of the 'Berger rhythm' to the occipital lobe. This was the first use of EEG in functional brain mapping.

Scalp EEG is assumed to reflect mainly the summation of graded postsynaptic potentials originating in the "cerebral cortex. Detection of EEG sources deep in the brain requires insertion of invasive electrodes. The contribution of cortical neurons to the EEG is itself a function of the endogenous states of the neuron and the activity of its input circuits (Lopes da Silva 1991). To make the story even more complex, the electrical signal of scalp EEG is distorted by the intervening tissue and is highly sensitive to the location of the recording electrode. Not surprisingly, the task of the investigator trying to identify and understand the source and physiological function of scalp EEG has been likened to

Brain

Column

Layer

w Dendrite

Synapse

Fig. 30 Spatial resolution and temporal characteristics of various functional neuroimaging methods.'Time' refers both to the temporal resolution (which limit for each method corresponds to the most left-handed point on the time-scale) and to the duration of the time window during which experiments could be conducted. Optical imaging is the imaging of intrinsic signals, or voltage-sensitive dyes, calcium-sensitive dyes, and other optical probes; these methods differ in their spatial and temporal resolution but are combined in the diagram for the sake of simplicity. EEG, electroencephalography; MEG, magnetoencephalography; fMRI, functional magnetic resonance imaging; PET, positron emission tomography. PET time window is closed because it involves radionucleides that cannot be administered repeatedly over a long time period. Currently, optical imaging is the most invasive of these methods yet provides the highest temporal and spatial resolution. EEG is the least invasive. So far, because of reasons indicated in the text, PET and fMRI have contributed most to the search for the *engram, and fMRI is considered of the highest potential in the field of human memory research.

Fig. 30 Spatial resolution and temporal characteristics of various functional neuroimaging methods.'Time' refers both to the temporal resolution (which limit for each method corresponds to the most left-handed point on the time-scale) and to the duration of the time window during which experiments could be conducted. Optical imaging is the imaging of intrinsic signals, or voltage-sensitive dyes, calcium-sensitive dyes, and other optical probes; these methods differ in their spatial and temporal resolution but are combined in the diagram for the sake of simplicity. EEG, electroencephalography; MEG, magnetoencephalography; fMRI, functional magnetic resonance imaging; PET, positron emission tomography. PET time window is closed because it involves radionucleides that cannot be administered repeatedly over a long time period. Currently, optical imaging is the most invasive of these methods yet provides the highest temporal and spatial resolution. EEG is the least invasive. So far, because of reasons indicated in the text, PET and fMRI have contributed most to the search for the *engram, and fMRI is considered of the highest potential in the field of human memory research.

that of an expert attempting to diagnose the problems of a computer by holding a voltmeter up to it (Bodis-Woll-ner 1987). Keeping this in mind, still, a lot can be done. The ongoing activity recorded from the scalp of a healthy "subject is at a frequency range of 1-30 Hz, with a few dominant state-dependent frequencies: a (the above 'Berger rhythm'); P (12-30Hz); 5 (0.5-4Hz); and 0 (4-7Hz). EEG can also be used to detect 'evoked potentials' (EP) or 'event-related potentials' (ERPs, e.g. Picton et al. 1995; the term ERP should be preferred to EP, as the latter is also used in cellular electrophysiology to denote the stimulus response of single neurons and "receptor cells). ERPs are time-locked to a sensory stimulus or a cognitive event. For example, "cued expectancy of an imperative stimulus is preceded by a low frequency negative wave ('contingent negative variation', CNV, Walter et al. 1964); a rare stimulus in a sequence of frequent stimuli evokes a characteristic positive component after >300 ms ('P300', Sutton et al. 1965, "surprise).

EEG is characterized by an excellent temporal resolution, in the millisecond range. Scalp EEG is completely noninvasive, relatively inexpensive, and can be used on freely moving subjects, including in "real-life situations. These are some of the pros. But there are, as usual, cons. The spatial resolution is poor, in the centimetre range, and the correlation of the signals with an identified source, as noted above, is problematic. A sister method, magnetoencephalography (MEG), offers some advantages, including a smaller sensitivity of the magnetic signal to distortion by the skull and simpler localization of the source signal (Wikswo et al. 1993). At the time of writing, MEG requires a cumbersome and expensive set-up, and the subject must remain immobilized throughout the measurement. In recent years the use of many electrodes (>100), the improvement in computational power, and the accumulating supporting data from other disciplines such as cellular physiology, has boosted EEG as a functional mapping method (Gevins et al. 1999). Still, so far the major proven strength of EEG in research is in detecting and separating, by time and type, broad electrophysiological phe-nomenological correlates of fast cognitive processes in humans, e.g. in "attention, "acquisition, and "retrieval of memory (e.g. Kutas 1988; Johnson 1995; Paller et al. 1995; Friedman et al. 1996; Miltner et al. 1999). Areas in which EEG has been proven particularly useful are the analysis of "working memory, and the acquisition, retrieval, and the dependence of long-term retrieval on acquisition in "declarative tasks in humans; in all these cases the task typically involves verbal material.

2. Positron emission tomography (PET). This method can be used to measure certain aspects of cerebral metabolism, such as local blood flow, glucose utilization ("nutrient), or "receptor occupancy. PET is based on the use of isotopes that decay by the emission of positrons (positive electrons). The emitted positron collides with an electron to produce two y-rays, which travel 180° apart. Coincident detection of the two y-rays permits the position of their source to be determined. In practice, a biologically relevant compound, labelled with the appropriate isotope, is injected intravenously or inhaled. The emitted radioactivity is monitored by an external ring scanner and used to compute a map of the distribution of the compound in the tissue. The isotopes used have a short half-life, in the range of minutes, so that large doses of radioactivity can be administered within acceptable safety limits. The short half-life necessitates access to an accelerator to produce the tracers on site. For example, compounds containing 15O, typically H215O, are used to measure task-related differences in regional cerebral blood flow. The underlying assumption is that such haemodynamic changes reflect alteration in neural activity; this notion can be traced back at least to James (1890) and Roy and Sherrington (1890; for a related pre-scientific idea, see Descartes 1649).Amino acids labelled with 11C are used to measure protein synthesis, 18F-labelled drugs to map receptor occupancy, and 2-deoxy-2-18F-d-glucose (18F-2DG) to map glucose metabolism. The 2DG method was originally used for post-mortem imaging (Sokoloff et al. 1977). 2DG is an analogue of glucose that is taken up by the cells, but cannot be metabolized like glucose, and is therefore trapped in an amount proportional to the level of glucose consumption. The assumption is that a high metabolic rate is indicative of increased neuronal activity. In the original method, 14C-2DG was administered before the task, the animal killed afterwards, and the active regions detected by quantitative autoradiography of brain slices. In PET, of course, the subject can live happily afterwards.

PET was developed in the mid-1970s, and since then has contributed tremendously to neurocognitive mapping (Raichle 1983; Cherry and Phelps 1996). Selected examples of the localization of memory systems in the mammalian brain by PET are provided in Buckner and Tulving 1995; Buckner et al. 1995; Fletcher et al. 1997; Maguire et al. 1997, 1998; Smith and Jonides 1997. For example, PET has enabled remarkable insight into the brain systems that subserve spatial memory and its use in navigation in humans, in situations that simulate "real-life (Maguire et al. 1997, 1998; "hippocampus). It has also provided a tremendous boost to the analysis of brain substrates of "retrieval (e.g. Buckner et al. 1995). PET has a respectable spatial resolution (millimetres), but a poor temporal one (minutes). It is moderately invasive because of the use of radionucleides.

3. Functional magnetic resonance imaging (fMRI). MRI detects and measures the characteristic interaction of certain atomic nuclei with radiofrequency waves in a strong magnetic field. (It is also known as 'nuclear magnetic resonance', NMR, only that the use of 'nuclear' was assumed to frighten the public, and dropped except in chemistry.) MRI can detect hydrogen in water, and as the large water content of organisms differs from one tissue to another, the method can be used to map internal organs. It can also detect and delineate lesions, for example, in the brain of "amnesics (e.g. Corkin et al. 1997). Early attempts to extend the use of MRI to map functional changes in blood flow (and hence develop fMRI) were based on the introduction into the blood an exogenous contrast agent (Rosen et al. 1989). But then a better method has been devised. It was based on the observation that the magnetic properties of oxyhaemoglobin and deoxyhaemoglobin differ (Pauling and Coryell 1936), and that paramagnetic deoxyhaemoglobin in venous blood can serve as a contrast agent for MRI. MRI methodology was thus adapted to obtain in vivo images of brain microvascula-ture with image contrast that reflects blood oxygen level (blood oxygenation level-dependent contrast, abbreviated 'BOLD contrast' or 'BOLD signal'; Ogawa et al. 1990,1998). The underlying tenet, as in some usages of PET, is that the haemodynamic changes reflect regional brain activity.

The temporal resolution of fMRI detectors is in the subsecond range, but the kinetics of the haemodynamic signal is slower, and therefore, practically, the temporal resolution is in the range of a few seconds. The spatial resolution at the time of writing is in the millimetre range (e.g. Logothetis et al. 1999; Ugrubil et al. 1999). Advanced, high field fMRI methods, based on the detection of early local changes in oxygen consumption rather than the delayed, more diffused alterations in blood flow, could potentially improve the resolution (Kim et al. 2000; on this issue see also Vanzetta and Grinvald 1999; Logothetis 2000). In behavioural experiments, to increase the signal-to-noise ratio, paradigms that use fMRI (or PET) commonly collect multiple images over extended time periods that contain successive trials of the same task ('block task paradigms'). Such protocols do not supply information about responses that are time-locked to single trials and about trial-to-trial change—as opposed, for example, to EEG studies of ERPs (see above). In recent years, protocols have been developed that allow analysis of trial-to-trial responses in fMRI as well (Buckner et al. 1996). This 'event-related fMRI' is particularly useful in investigating brain response in situations in which different stimuli are mixed over time, and where the response is expected to vary from one stimulus to another.

fMRI is considered noninvasive, although the long-term safety for humans of the high magnetic fields that are already used on the "monkey must still be determined. The advantages of the high resolution and the assumed noninvasiveness clearly outweigh the fact that fMRI does not measure neuronal activity directly. At the same time, this situation emphasizes the need to understand in fine details the source, specificity, and physiological significance of the biological signal (Vanzetta and Grinvald 1999; Kim et al. 2000; Logothetis et al. 2001). The contribution of fMRI to the field of memory research is clearly on the ascending limb, and the pace of publication is admittedly almost alarming (the question how can a single human being read all this literature certainly deserves a special functional neuroimaging study). Results cited in many entries in this book are either based on or supported by fMRI (e.g. "acquisition, "cerebral cortex, "hippocampus, "retrieval, "skill; for selected examples see Buchel et al. 1998; Fernandez et al. 1998; Kelley et al. 1998; LaBar et al. 1998; Poldrack et al. 1998; Dolan and Fletcher 1999; Wagner et al. 1999; Cabeza et al. 2001; Figure 31).

4. Optical imaging. Optical functional neuroimaging methods rely on the detection of intrinsic activity-dependent optical changes in neural tissue, or on the use of voltage-sensitive dyes, ion sensitive dyes and other extrinsic optical probes and tracers. Alterations in the intrinsic optical properties of nerve fibres were first reported in the late 1940s (Hill and Keynes 1949). Staining with voltage-sensitive dyes to measure neuronal activity was described by Tasaki et al. (1968). The voltage-sensitive-dye imaging methodology was subsequently improved and put to use by a number of groups (e.g. Cohen et al. 1974, Grinvald et al. 1981).

ma

HAMMER

H'- * "

y

0 0

Post a comment