Which method best allows researchers to examine brain anatomy?

Cited by (0)

Brain Structure

Brain anatomy can be examined by voxel-based morphometry (VBM), which is a neuroimaging analysis technique that allows investigation of focal differences in brain structure. VBM registers every brain to a template, which gets rid of most of the large differences in brain anatomy among people. Then the brain images are segmented into gray and white matter and cerebrospinal fluid. Finally, the local concentration of gray or white matter is compared between groups of subjects at every voxel. So far, only two studies approached the systematic examination of white and gray matter differences between children with DD and age- and gender-matched control children (Rotzer et al., 2008; Rykhlevskaia, Uddin, Kondos, & Menon, 2009). Results point to deficits in gray matter in the intraparietal sulcus and adjacent regions, including the superior parietal lobe (see Figure 2). Similarly, reduced gray matter has been reported in the parietal lobe in adolescents with very low birth weight and difficulties learning mathematics (Isaacs, Edmonds, Lucas, & Gadian, 2001). Because the parietal lobe in general, and particularly the intraparietal sulcus, are presumed to represent core regions implicated in the development of numerical skills, reduced gray matter volumes in these areas might represent the morphometric correlate of atypical number development. Therefore, gray matter volume in parietal areas seems to be a crucial indicator for the accurate development of numerical skills.

The intraparietal sulcus is thought to be the neural substrate most closely associated with DD. However, recent results also argue for a more distributed neuronal pattern. For example, reported gray matter deficits in children with DD suggest widespread but subtle malformations in the brain. Reduced gray matter volume has also been reported in the anterior cingulate cortex, the left inferior frontal gyrus, and the dorsolateral prefrontal cortex (Rotzer et al., 2008). Based on the fact that an important component in the development of arithmetic skills is the growth of memory for numerical information, the authors argue that deficits in these frontal areas may be of major importance in the development of dyscalculia. Similarly, VBM analysis has revealed that individual differences in the arithmetic scores of typically achieving children are significantly and positively correlated with the gray matter volume in the left intraparietal sulcus (Li, Hu, Wang, Weng, & Chen, 2013).

The ventral visual stream seems to be affected as well (Rykhlevskaia et al., 2009). In particular, children with DD showed deficits in the fusiform gyrus, parahippocampal gyrus, and right anterior temporal cortex which might hinder the development of semantic memory representations important for rapid numerical fact retrieval (Rykhlevskaia et al., 2009).

Regarding white matter integrity, the results of existing studies report reduced white matter volume in children with DD, but regions are not consistent (Rotzer et al., 2008; Rykhlevskaia et al., 2009). Whereas Rotzer et al. (2008) reported white matter deficits in the left frontal lobe and in the right parahippocampal gyrus in dyscalculic children, Rykhlevskaia et al. (2009) found reduced white matter volume in the right temporo-parietal region and the splenium of the corpus callosum. Deficits in these regions may have a negative influence on fact retrieval and spatial memory processing, as well as on visuospatial processes during the acquisition of some types of mathematical problem solving (Rotzer et al., 2008; Rykhlevskaia et al., 2009). Taken together, DD is characterized by reduced gray and white matter volume in areas that are thought to play a key role for the domain-specific development of numerical representations and abilities, as well as in regions important for the development of domain-general abilities.

III. History

Comparative brain anatomy studies of great apes were underway early in the 20th century, including studies of orangutan and chimpanzee (Campbell, 1905), orangutan (Mauss, 1911), chimpanzee (Brodmann, 1912), and gorilla (Bolk, 1910). These included examination not only of surface topography, but also detailed microscopic examination of underlying tissue and documentation of homologies between ape and human brain areas (Campbell, 1905). Research on the effects of ablations and lesions on learning, memory, and emotionality continued for many years (Blum, 1948; Rosvold et al., 1961; see Markowitsch, 1988, for a review). Especially significant was a series of studies by Bailey and colleagues comparing and mapping human and chimpanzee neocortex (Bailey, 1948; Bailey et al., 1950). While these studies did not focus on aging, the results are useful in understanding the functional neurobiology of great ape and human brains. It is unlikely that further ablation or lesion studies will be done on great apes. Most invasive research involving great apes has ceased, partly due to studies such as those cited below that indicate how similar the great apes are to humans.

Subcortical Structures

The majority of brain anatomy studies using MRI confirm the basal ganglia involvement and bring some evidence for the predictive power of decreased caudate volume in childhood for tic severity in adolescence (Bloch, Leckman, Zhu, & Peterson, 2005; Peterson et al., 1993, 2003). However, the up-to-date published studies bring substantial controversy regarding directionality of basal ganglia changes. Early reports on smaller basal ganglia in TS children were challenged by findings confirming larger putamen volume in the absence of thalamus, caudate, or pallidum changes (Roessner et al., 2011). Similarly, a study using a high-precision surface-based diffeomorphic technique in neuroleptic-naive TS adults failed to show any significant volume or shape differences in the basal ganglia or thalamus (Wang et al., 2007). Conversely, in a large study sample (n = 149), the authors observed larger thalamus volumes in TS without significant correlation with tic severity (Miller et al., 2010).

Computational anatomy on whole-brain MRI data is a validated approach to the unbiased investigation of brain structural correlates of brain disorders (Ashburner et al., 2003). A number of published studies on TS using a voxel-based morphometry or cortical thickness assessment report differences from normal subjects within the basal ganglia, limbic system, and pre-frontal cortex (PFC), corroborating ROI volumetry results (Ludolph et al., 2006; Muller-Vahl et al., 2009). The majority of ROI-based and whole-brain computational anatomy studies assessed basal ganglia anatomy using T1-weighted imaging data, which turned out to be methodologically challenging (Babalola et al., 2009; Wonderlick et al., 2009). The difficulties in assuring accurate and robust computer-based delineation of the basal ganglia are mainly due to the presence of a high amount of iron (Hallgren & Sourander, 1958), which results in poor and variable contrast on T1-weighted magnetic resonance images (Callaghan et al., 2014; Patenaude, Smith, Kennedy, & Jenkinson, 2011).

3.12.2.1 Bigger Animals Have Bigger Brains, and Bigger Brains Have More and More Cortex

Comparative studies of mammalian brain anatomy have been largely limited to analysis of volumetric data on large brain divisions of different species published by a small number of labs (Stephan et al., 1981; Frahm et al., 1982), often based on measurements of only one brain of each species. These have established that brain size is related to body size by a power law of exponent inferior to 1.0 (Martin, 1981; Fox and Wilczynski, 1986), such that brain size increases with body size, but at a slower pace (Figure 1a). The cerebral cortex increases hypermetrically in volume in relation to the remaining brain structures (Figure 1b), such that, within each order, the relative size of the cerebral cortex increases with brain size (Figure 1c): larger brains are more and more dominated by cortex (Frahm et al., 1982).

In comparison, larger brains have isometrically larger cerebella, which accompany almost linearly the size of the cerebral cortex (Figure 1d), and retain a stable relative size with increasing brain size: larger brains have cerebella of the same relative volume (Figure 1e).

3.28.2.3 Dopamine Deficits

Aging affects not only brain anatomy but also brain physiology, including the function of neurotransmitter systems, such as serotonin, acetylcholine, and dopamine (Strong, 1998). Associated with volume decrements in PFC are decreases in dopamine (DA) concentration and transporter availability (Volkow et al., 2000). Additionally, dopamine D2 receptor density declines at a rate of 8% per decade, beginning in the 40s. There is abundant evidence that DA systems play an important role not only in motor operations but also in higher-order cognitive processes. DA function can be measured in vivo using PET (Bäckman and Farde, 2005).

There is strong evidence of age-related losses in pre- and postsynaptic DA markers, which may reflect decreases in the number of neurons, the number of synapses per neuron, and/or the expression of receptor proteins in each neuron. D1 and D2 receptor binding declines from early adulthood at a rate of 4–10% per decade, and this decline is correlated with the decline of dopamine transporter, possibly reflecting a common causal mechanism. DA loss with aging has been observed in frontal, temporal, and occipital cortices as well as in hippocampus and thalamus (Kaasinen et al., 2000; Inoue et al., 2001). The magnitude of extrastriatal DA decline mirrors that observed within the striatum itself. Given the cognitive role of frontostriatal loops, age-related striatal DA deficits could also account for age-related cognitive deficits associated with PFC dysfunction. Moreover, age-related deficits in DA binding have been observed in PFC, as well as in posterior cortical and hippocampal regions. Evidence is mixed as to whether these declines are linear (see Reeves et al., 2002) or exponential (Bannon and Whitty, 1997; Rinne et al., 1998; Ghilardi et al., 2000) across adulthood.

The relationship between age-related changes in DA and age-related cognitive differences has been examined in only a small number of studies. Despite the paucity of data, findings are remarkably consistent. Age deficits in striatal DA have been associated with reduction in episodic memory (Bäckman et al., 2000; Erixon-Lindroth et al., 2005), executive function (Volkow et al., 1998; Mozley et al., 2001; Erixon-Lindroth et al., 2005), and motor performance (Wang et al., 1998; Mozley et al., 2001). Furthermore, several studies have also found that striatal DA markers serve as a significant predictor of cognitive performance, after controlling for the effects of age (Bäckman et al., 2000; Volkow et al., 1998), as well as that age-related cognitive deficits are mediated by reductions in striatal DA functioning (Erixon-Lindroth et al., 2005).

Large-Scale Brain Systems

Using MRI-measured brain anatomy and functional connectivity from 1000 healthy adults, Yeo et al. (2011) recently observed the remarkable replicability of the same seven patterns of cortical networks within the human brain. These same networks are identified in adults, adolescents, children, and even in infants as assessed through resting-state neuroimaging technologies (Menon, 2013). Many brain regions within these networks are involved in both nonlanguage related and language processes. Our “bullet point” is that specific “hubs” or “nodes” within these cortical regions have reciprocal connections with different modules within the cerebellum and nodes within basal ganglia (Buckner, personal communication, April 3, 2014; Habas et al., 2009; Middleton & Strick, 1996). (Many subsystems and networks can be identified; we focus upon this model for the purpose of illustration, since this model has been the subject of a preponderance of investigations.)

These large-scale cortical systems include frontoparietal networks (FPNs) commonly engaged during effortful cognitive task performance requiring information or rules to be held in mind for purposeful, goal-directed behavioral guidance. FPNs consist of the dorsolateral prefrontal cortex, the anterior cingulate cortex, the anterior prefrontal cortex, the lateral cerebellum, the anterior insula, the caudate nucleus, and the inferior parietal lobe. The left hemisphere FPN is responsible for internally guided behavior; the right hemisphere FPN is activated by external influences when situations are unfamiliar and require the development of new problem-solving strategies. The dorsal and ventral attention networks are involved in goal-directed executive control processes and salience evaluations respectively, which are necessary operations for the control of spatial attention and the orientation of attention to a specific area of interest. The ventral attention network (VAN) includes the temporoparietal junction, the supramarginal gyrus, the frontal operculum, and the anterior insula. The focus of the VAN is primarily upon allocentric space, or knowing about objects that lie beyond immediate reach, including information about what those objects are used for. The dorsal attention network (DAN) is anchored in the intraparietal sulcus and the frontal eye fields. The DAN includes a focus upon egocentric space to generate sensory-motor information about functions such as reaching, grasping, the “data” that are important for knowing about how to use objects. The occipital lobe, the lateral temporal region, and the superior parietal lobule, making up the visual network (VN), interact with the DAN and VAN to sustain attention and to suppress attention to extraneous, irrelevant variables. Therefore, these are critical components of the brain’s “action control” system, and as we will demonstrate throughout this chapter, certain hubs within these regions are involved in linguistic functions. The limbic network interacts with these systems to generate motivational and reward influences. The sensory motor network (SMN) consists of the primary motor cortex, the primary and secondary sensory cortices, the supplementary motor cortex, the ventral premotor cortex, the putamen, the thalamus, and the cerebellum. These regions are involved in language, and in certain motor abnormalities that are also observed in developmental disorders. In addition, a default mode network (DMN) whose activity is high until active, goal-directed cognitive processing is required is anchored in two regions, the anterior medial prefrontal cortex and the posterior cingulate cortex as well as two additional systems, the dorsomedial prefrontal system and the medial temporal lobe memory system. The DMN is less active during the performance of cognitive tasks in normal control subjects, but it is considerably more active in various psychopathologies (Sandrone, 2012). (This chapter is a selective review of brain networks specifically involved in language; for comprehensive reviews and illustrations, see www.humanconnectomeproject.org.)

Simply put, what this “distributed network” perspective implies is that it is no longer appropriate to think of language as a single or unitary entity, given the complexity of the neural systems involved. The brain regions involved in many of these networks participate in multiple brain functions, including linguistic processes. Language cannot be viewed as a dedicated domain that resides separate from any other aspect of cognition, regardless of how cognition is operationalized. This emerging network-based appreciation of the complexity of brain–behavior relationships clearly goes well beyond the “canonical” or generally accepted, traditional model of receptive and expressive language functions as well as other aspects of thinking and behavior that are often considered a specific functional domain. Large-scale networks and even smaller, more modularized subnetworks are believed to have strong degrees of specialization for certain behavioral functions and operations (Castellanos & Proal, 2012; Cortese et al., 2012). On the one hand, this might superficially seem to replace a strict localizationist perspective (i.e., one brain region = one cognitive function) with a similar framework that substitutes a single brain region for a group of regions. However, this initial conclusion is misleading. These specific regions, hubs, or nodes within large-scale brain networks interact with the learning, inhibitory, and modulatory influences provided by vertically organized, subcortical systems (Koziol, 2014; Koziol, Budding, & Chidekel, 2013). These interactions allow for the critical understanding of all distributed cognitive processes. This includes the intrinsic relationship between language and other distributed functional brain-behavioral systems, all of which are dependent upon corticocortical and cortical–subcortical interactions. Language and communication disorders can be better appreciated and understood within this integrated framework.

The simplified seven-network parcellation proposed by Yeo et al. (2011) features numerous brain regions that are recruited by linguistic tasks. For example, language tasks recruit the anterior PFC, the caudate nucleus, and the parietal lobe of the left-lateralized FPN, and the bilateral DLPFC; within the VAN, the temporal-parietal junction, upper regions of the superior temporal sulcus (STS), the frontal operculum of the left hemisphere, and right-lateralized supramarginal gyrus are all known to be involved in language functioning; the intraparietal sulcus of the DAN and the lateral temporal and superior parietal lobule of the VN are activated in language tasks as well as the medial temporal lobe system of the DMN. Nearly all hub regions of the SMN are recruited in language functioning (Hillert, 2014; Lieberman, 2000; Mariën, Engelborghs, Fabbro, & DeDeyn, 2001; Murdoch & Whelan, 2009; Ullman, 2004; this listing represents an introductory sampling, not intended to be interpreted as complete or exhaustive).

Therefore, most components of cortical hub regions are multifunctional with subregions that contribute to various functional processes. These hubs are not totally dedicated to any specific behavioral process. The neurologic substrates of language are perhaps best characterized by neural multifunctionality (Cahana-Amitay & Albert, 2014). This means that the evolving brain slowly incorporated functions that initially were nonlinguistic to support a language system that required dynamic, ongoing interactions between neural networks that subserve praxic, affective, “thinking,” and behavioral functions. Aspects of these networks became specialized for semantic classification, language comprehension, lexical word retrieval, and discourse, conversational processing. The most critical questions concern how these networks and nodes share their resources and how they communicate and coordinate their functions to provide the integration necessary to complete language tasks. The specific recruitment of a brain region in language is determined by patterns of connectivity; in other words, there is significance in the nature or kind of interaction with other regions. This distributed network perspective is critical in attempting to understand all features of language functioning; in this regard, language can no longer be considered apart from other aspects of thinking or behavior. Language needs to be understood within the context of shared neuroanatomic resources and correlates for language and nonlanguage functions; these include attention, the operations of multiple memory systems, audition and vision, motor functions, and even affect/emotion. Although all of these processes cannot be included in this selective review, the basic framework has been established and a few subtle features are illustrated by two simple neuropsychological testing-type examples. In our opinion, these critical points are seldom even considered in attempts to understand linguistic processing.

First, on a study of the California Verbal Learning Test, second edition (Delis, Kramer, Kaplan, & Ober, 2000), in a word list acquisition task, the left hemisphere was clearly activated as observed by fMRI imaging. However, the immediate memory trial, novel, or unfamiliar words, additionally activated the right hemisphere anterior hippocampus. Across all trials, including a recognition condition, the overall best “verbal learners” were those subjects who recruited the highest levels of right hemisphere DLPFC and right anterior hippocampal activity (Johnson, Saykin, Flashman, McCallister, & Sparling, 2001). In another study that assessed the analysis of metaphors (such as the expression, “I’m having a rough day” or a “hard time”) fMRI studies identified bilateral brain activation of those regions necessary for “texture” processing, but not “visual” processing, or areas for processing other sensory modalities (Sathian et al., 2011). These (and other) studies clearly reveal that a “fixed hemispheric assignment” is erroneous to conclude for any brain function, that a distributed brain network perspective is absolutely necessary for understanding linguistic processing, and that language simply cannot be understood as an isolated domain residing only in the left cerebral hemisphere, apart from other motor and nonmotor functions.

Structural Neuroimaging

Most of the information regarding brain anatomy in adolescents comes from magnetic resonance imaging (MRI) studies. MRI is particularly well suited for in vivo pediatric studies, because unlike conventional X-rays and computerized axial tomography it does not use ionizing radiation. This allows not only the initial scanning of children and adolescent but also of repeated scans over time.

The spatial resolution of MRI scans is constantly improving and greater resolution can be purchased with the currency of time in the scanner, but in current studies the usual size of the smallest volume with a single MRI signal value (i.e., voxel) is approximately 1 ml. This is worth noting with respect to interpretation of the MRI findings, because within a given 1 ml voxel there may be millions of neurons and trillions of synaptic connections. For example, one unit of the mouse cerebral cortex contains about 30% axons, 30% dendrites, 14% cell bodies, 12% dendritic spines, and 9% glia. However, given that a single MR voxel of cortical gray-matter in the human brain contains a mixture of cell types, it is impossible, at this point, to attribute the observed effects to any single cellular compartment.

White matter volumes increase throughout childhood and adolescence, although rates of increase vary by region and age. This increase in volume is thought to be due to ongoing myelination, the wrapping of an insulating material around axons by oligodendrocytes, which serves to increase greatly the speed of neuronal signal transmission and facilitates the integration of distributed neural networks.

An MR technology called diffusion tensor imaging (DTI), which enables assessment of directionality of fiber tracts, is being used to characterize further white matter development during childhood and adolescence. Pediatric DTI studies have begun to explore relationships between white matter development and cognitive capacities.

Gray matter volumes follow a distinctly different developmental path following an inverted ⋃ shaped trajectory. Gray matter volume trajectories vary by region but peak during adolescence for frontal, temporal, and parietal lobes. Particularly late to reach adult levels of cortical thickness are areas in the prefrontal cortex and posterior superior temporal region. Although specifics of the relationships between the anatomic changes and cognitive or emotional changes have not been well elucidated, discussions regarding the implications of the relatively late anatomical maturation of the prefrontal cortex (a key component of circuitry involved in judgment, decision making, and impulse control) have prominently entered social, educational, and judicial realms.

An important, but unresolved, question is the extent to which these gray matter volume changes reflect intracortical myelination versus dendritic and axonal arborization/pruning (the number and extent of neuronal branches). Although arborization/pruning and myelination processes both occur, understanding their relative contributions to the gray matter changes has implications for the conceptualization of brain development in health and illness, the design of future studies, and ultimately as a guide to interventions.

Publisher Summary

The tensor model assumes that the brain anatomy consists of a fiber population with a uniform fiber angle within a pixel. No matter what the fiber architecture is, tensor fitting forces the results to be fitted into one diffusion ellipsoid, in which there is only one long axis that is supposed to represent the fiber orientation. This is true when there is only one fiber population in a pixel. If there are two fiber populations, different fiber architectures may lead to the same fitting results, and the longest axis may not represent the orientation of any fibers. Two different types of neuroanatomy may lead to the same diffusion tensor imaging (DTI) result (information degeneration). In this case, measurements are under-sampled, and the estimation of fiber architecture from diffusion measurement becomes ill-posed. Depending on the angle and population ratios of fiber populations, diffusion may even look isotropic. The real issue is that many regions in our brain seem to have fiber architectures. In such cases, the tensor model simply fails. Another practical issue is that DTI provides six parameters/pixels. With six times more information per pixel, the tensor model may contain too much data to analyze.

3.3 Brain Development in Adolescence

Adolescence is characterized by changes to brain anatomy, with indications that some changes are specifically related to puberty; most evidence comes from neuroimaging of typical adolescents, using structural MRI. There are changes in the relative proportion of gray matter (containing cell bodies) and white matter (fiber tracts), with decreases in the amount of gray matter and increases in the amount of white matter (Lenroot & Giedd, 2006); these changes occur approximately 2 years earlier in girls than in boys, consistent with the sex difference in puberty (Lenroot et al., 2007). The developmental changes in cortical surface are illustrated in Figure 2.

In general, the changes are associated with increases in sex-specific pubertal hormones, with effects most robust when age is experimentally or statistically controlled (Herting et al., 2014; reviewed in Peper, Hulshoff Pol, Crone, & van Honk, 2011). More powerful evidence comes from longitudinal studies, which have shown that, for both boys and girls, pubertal development (assessed with Tanner stage or indexed by testosterone and estradiol) is related to volumetric changes in subcortical brain regions (e.g., amygdala, hippocampus, caudate); the direction of effect and influence of age were seen to depend on sex and brain region (Goddings et al., 2014; Herting et al., 2014).

Although these findings confirm that pubertal processes are involved in brain development, many questions remain. First, little is known about the mechanisms responsible for the changes, particularly whether they reflect brain reorganization brought about by pubertal hormones (see below) and the extent to which they depend on the social environment. Second, it is unclear what specific neural processes are changing. It is assumed that changes represent cell death and synaptic pruning, but there is little direct evidence to show this. Third, the implications of anatomical changes for brain function and behavior in adolescence have hardly been studied, that is, little is known about how these changes in brain structure affect the psychological changes of adolescence (Casey, Tottenham, Liston, & Durston, 2005; Giedd, 2004). For example, changes in the brain might be directly responsible for the psychological changes or both sets of changes might simply co-occur as the result of a third factor (such as chronological age or pubertal hormones).

Social theories

The theories outlined so far have focused on brain anatomy and physiology, psychologic processes within the individual brain and computation algorithms. But there are several reasons for suggesting that consciousness has an important social dimension. First, we have seen that the Latin root of “consciousness” referred, originally, to knowledge shared with another. Second, the sharing of knowledge with oneself, in awareness, and the sharing of knowledge with others, in social exchanges, may be interconnected: there is a theoretic argument and empiric evidence that awareness of self and awareness of others are acquired in parallel (Strawson, 1974; Parker et al., 1994). Third, language is a vital contributor to human awareness, and language, clearly, is a social phenomenon. Proponents of social theories sometimes claim that the social dimension of consciousness explains the bafflement we tend to feel when we try to explain how the brain can generate experience: on these views experience is as much a social construction as a biologic and psychologic phenomenon (Rose, 1998; Singer, 1998).

Humphrey (1978) provided a lucid example of theories which propose a social function for awareness. He suggested that the purpose of consciousness is to allow social animals to model each other’s behavior on the basis of their insight into its psychologic motivation. In other words, our knowledge of our own mental states supplies us with insight into the mental states underlying the actions of others; the ability to predict these actions is a major determinant of our biologic success. More recently, such knowledge has been described in terms of the possession of theory of mind: some social theories broadly associate this with consciousness. The identification of mirror neurons – cells that are activated by performing actions oneself and by watching others perform the same actions – provides one potential mechanism for the rapid identification of the mental states of others.

There is no doubt that a comprehensive theory of consciousness needs to take account of its social dimension. But most commentators agree that this is the wrong level of explanation for the simpler forms of consciousness, providing an avenue by which to understand varieties of self-awareness or “higher-order consciousness” rather than addressing the more basic phenomenon of perceptual awareness.