1.8 REDES INALÁMBRICAS MESH EN IEEE 802.11-2012
1.8.4 SUBCAPA MAC
In Chapter three, I suggested that a possible reason for the lack of robust BOLD signal increases in primary somatosensory cortex (SI) during simple somatosensory stimulation may have been that the stimuli used were not optimally tuned to excite SI. ‘Tuning’ is a principle common to all sensory systems: for example, the rods and cones of the human eye are ‘tuned’ to detect electromagnetic energy of a particular wavelength. Similarly, Merkel receptors in the human hand are ‘tuned’ to detect the application of forces to the skin surface. From a signal-processing perspective, these sensory receptors act as filters: only stimuli that pass through the filter are processed centrally by the nervous system. This concept can be extended to include the relative responsiveness of neurons within the CNS: for example, it is appropriate to classify V4 neurons as ‘tuned’ to detect light of particular wavelengths (Zeki, 1980). Thus, if a neuroimaging experiment is designed primarily as a ‘probe’ of the responsiveness of a particular cortical area, the known neurophysiological profile of the area should influence the choice of stimuli.
The experiments described in Chapter three were designed to map the somatotopical layout of the digits o f a single hand. Stimuli were chosen to maximise the likelihood of seeing BOLD signal change without consideration of the relationship between the magnitude of a particular stimulus dimension and activation magnitude. Thus the stimuli were treated as a means to an end: they allowed me to attempt to map the somatopical layout of the body surface within SI, with a view to eventually examining experimental alterations of this topography. The dimensions of the stimuli used to elicit BOLD signal change within SI (e.g. vibrotactile frequency, airpuff intensity) were regarded as subordinate to their ability to activate SI. However, as the results of Chapter 3 demonstrated, this initial challenge proved more complex than first imagined. Few studies have investigated the responses of human somatosensory cortex to a systematically varied input function in fMRI (although see Kampe et al., 2000 for human data and Gyngell et al., 1996 for animal data). I therefore chose to examine in more detail the stimulus-response characteristics of the somatosensory system to simple repetitive airpuff stimulation.
5.1.1 Neural Coding and Sensory ‘Tuning’
The lack of either robust BOLD signal changes or recognisable patterns of somatopy in SI may have been caused by the inability of previous stimuli to maximally excite SI. Sensory receptors are not digital filters that simply switch ‘on’ in response to a preferred form of stimulation - while they may be tuned to a particular quality of stimulation, they will typically respond to similar stimuli, albeit with lesser efficacy (Fig.5.1).
Firing Rate (units/s)
S h a rp T u n in g
_ _ _ B r o a d T u n in g
Stimulus Dimension (units)
F ig u re 5 .1 . S c h e m a tic r e p re s e n ta tio n o f tu n in g o n s e n s o r y s y s te m s . D if f e r e n t c e lls d is p la y d if f e r e n t d e g r e e s o f s e le c tiv ity . T h e g ra p h s h o w s tw o id e a lis e d p lo ts o f r e s p o n s e s fro m n e u r o n s o v e r d if f e r e n t v a lu e s o f a s tim u lu s . W h ile b o th c u r v e s s h o w a m a x im a l re s p o n s e fo r a p a r tic u la r v a lu e , th e ir r e s p o n s e s a re g r a d u a te d s u c h th a t v a lu e s a r o u n d th e p r e fe r r e d s tim u lu s v a lu e w ill a ls o e lic it a n in c r e a s e in firin g . D e p e n d in g o n th e p r o p e r tie s o f th e n e u r o n , th e re m a y b e a s m a ll r a n g e o f v a lu e s a ro u n d th e o p tim a l v a lu e th a t th e n e u r o n w ill fire to {sh arp tuning), o r th e c e ll m a y r e s p o n d to a b r o a d r a n g e o f v a lu e s {b r o a d tuning).
So, while receptors and cortical neurons may respond in a categorical fashion when presented with different classes of stimuli (e.g V4 cells show minimal responses to movement, and strong responses to colour differences), they will also invariably display graduated responsiveness to sub-dimensions of these stimulus classes (i.e. V4 cells show selectivity to specific wavelengths of light). Since neurons typically signal stimulus preference via an increase in firing rate (which is linked to neuronal metabolism), the greatest BOLD signal change should be elicited in SI when presenting stimuli that the cells are maximally tuned to detect. Logically, then, efficient mapping stimuli should be those that produce the highest firing rates in SI. However, in SI at least, it may be possible to rank the ‘best’ stimuli according to two, possibly orthogonal criteria: whether the stimuli produce the greatest signal change, or
whether the stimuli provide the best delineation of the underlying somatopy. This represents a further challenge for non-invasive investigations of map topography.
5.1.2 Convergent and Divergent Connectivity in the Somatosensory System
The situation outlined above can be explained by the physiological and anatomical processes underlying map structure in somatosensory cortex (Dykes and Ruest, 1986). Before incoming afferent information reaches the primary somatosensory cortex in humans, it must synapse in the spinal cord, brainstem nuclei, and the somatosensory thalamus. As discussed in Chapter One, there is no strict Tabelled-line’ code for afferent somatosensory information - at each synapse information from adjacent ascending fibres can be combined, and the extensive divergence of ascending somatosensory projections results in a ‘funnelling’ effect, becoming most pronounced in the thalamus. For example, the terminal arborization of a single lemniscal fibre can branch in close proximity with up to two hundred thalamic neurons in monkeys (Jones, 1983). These aggregations of cells are known as lamellae. In the monkey, it is estimated that adjacent thalamic cells can project to cortical cells separated by up to 1.5mm. This divergence was posited as a possible mechanism to explain the topographic map expansions in SI seen after peripheral injury (Wall, 1977; Merzenich et al., 1984), and was initially thought to represent a ‘hard limit’ for the extent of map changes. However, while adjacent thalamic cells project to areas separated by only 1.5mm (in monkey cortex), the distance covered by thalamic cells belonging to the same
lamella can project to points that are separated by a number of millimetres.
Thalamocortical projections to SI do not therefore form an exact isomorph of the peripheral receptor sheet - overlaps exist (Rausell and Jones, 1995).
The functional significance of this divergence can be seen after central injury: it has been estimated that upward of 35% of the ventroposterior lateral nucleus, including a substantial portion of the cells projecting to a single digit representation, can be destroyed before any change in the digit’s representation in area 3b can be detected (Jones et al., 1997). In other words, the pattern of divergence makes it unlikely that limited central lesions will produce ‘silent zones’ within the cortex that are not responsive to peripheral stimulation. A similar pattern of divergence is present in single
thalamocortical axons: projections to layer IV of macaque primary somatosensory cortex display similarly extensive arborizations (axons with arbours extending up to 2.5/3mm in cortex have been described; Garraghty and Sur, 1990).
5.1.3 SI Maps: Subthreshold Influences
However, in addition to the diffuse pattern of connectivity suggested by analysis of projection neurons to SI, there is a growing body of evidence to suggest that intracortical circuitry may also contribute to this arrangement (Lund et al., 1993). Under physiologically ‘normal’ conditions approximately 20% of racoon SI cells in which excitatory postsynaptic potentials (EPSPs) can be elicited by stimulation of a single digit also display EPSPs to stimulation of adjacent digits (Smits et al., 1991). Experiments involving pharmacological manipulation of GABAergic transmission in SI support this view: Dykes and Ruest (1986) found that iontophoretic infusion of bicuculline (a G ABA antagonist) caused expansion of SI receptive fields in the cat. Alloway and Burton (1991) found similar results in primates. Thus the classical receptive fields of cortical neurons in primary sensory areas as mapped under anaesthesia are merely one possible configuration of a dynamic, context-specific map.
These findings do not mean that there are no constraints on the representations o f peripheral representations in SI. Some studies have suggested that hard anatomical boundaries exist that act to limit neuronal representations from changing their size greatly under normal physiological conditions (e.g. Hickmott and Merzenich, 1998). However, these mechanisms still have the potential to obscure ‘natural’ map boundaries in SI when studied using neuroimaging techniques. For example, stimuli of a particular form may cause responses outside the ‘classical’ receptive field to be expressed. In the owl monkey, neurons within MT/V5 respond to complex stimuli across a region that may be up to 50/100 times the size of the classical receptive field (Allman et al., 1985). While V5 is a ‘higher’ sensory area, it contains an orderly retinotopic map of visual space, and thus any attempts to form a map of visual space within V5 would be suboptimal if stimuli of the kind used by Allman were employed. It is therefore important to carefully choose stimuli for mapping experiments.
5.1.4 Stimulus Rate and Neuroimaging Studies
It is possible to characterise tactile stimuli according to a number of different dimensions (e.g. intensity/predictability/location/roughness), and thus qualify the relationship between BOLD signal in SI and variations along each of these dimensions. I chose to examine the frequency- or rate-dependence of SI signal to airpuff stimulation. Although ‘rate’ and ‘frequency’ are essentially interchangeable terms, it is important to distinguish between them as it is typical in somatosensory neurophysiological/psychophysical studies to use ‘frequency-dependence’ in the context of experiments that use sinusoidal stimuli applied by vibrotactile devices (e.g. Bowlanowski et al., 1988).
When using naturalistic somatosensory stimuli, it is rarely possible to specify stimulus attributes to the fine degree permitted by median nerve stimulation. A more simplistic but more realistic way to characterise these stimuli is in terms of the rate at which they are delivered. This concept of the ‘delivery rate’ of a stimulus is similar to that used in parametric neuroimaging experiments (e.g. Fox and Raichle, 1985; Grafton et al., 1992; VanMeter et al., 1995; Price et al., 1996). A clear relationship between increasing the rate of finger movement and the maximum BOLD signal has previously been demonstrated (Rao et al., 1996; Schlaug et al., 1996), and in PET a linear increase of rCBF and median nerve stimulation frequency between 0-4Hz was demonstrated by Ibanez and colleagues (1995), with a subsequent plateau effect at frequencies between 4- 20Hz. There has, to date, been only one systematic investigation of the frequency dependence of the BOLD signal measured with fMRI (Kampe et al., 2000). The results of this study bear out the concerns voiced above - stimulating the median nerve, the authors found that increasing stimulus frequency resulted in a linear increase in BOLD signal, but with an accompanying increase in the number of activated voxels. Therefore, increasing the rate of stimuli too much may act to obscure patterns of somatopy in SI.
If the spatial organisation of activity in topographically mapped areas is thought to be a useful metric of information processing (and there is ample evidence to suggest that map structure in SI is not merely an epiphenomenon of cortical development; e.g. Kaas, 1997), the stimuli used in mapping experiments must be carefully chosen. If the stimulus maximally excites neurons within SI, cells outside the ‘classical’ receptive fields may also be
excited. As SI digit representations are organised along a strip of cortex, any recruitment/lateral inhibition of surrounding neurons will result in a ‘smearing’ of the BOLD signal, and a loss o f functional resolution. Similarly, if the stimulus only minimally excites SI, the BOLD signal change may be too small to permit reliable detection. Thus, for mapping purposes, the ‘best’ stimulus is one that excites SI enough to produce detectable changes in SI, yet produces minimal lateral spread to neighbouring representations. Sheth and colleagues (1998) have explored similar issues using optical imaging in rat barrel cortex.