Spatio Temporal Patterns of Neural Activity

The nervous system includes several subsystems with peculiar features that add up to the complexity. In addition to the input (sensory channels) and output (motor and humoral channels) subsystems, there is an extensive network of interconnected ganglia and nuclei that are mainly located in the brain, the most prominent of which is the cerebral cortex (for review on the quantitative analysis, see Braitenberg and Schuz, 1991). The cerebral cortex (the cortex) is formed by two hemispheres and represents the major part of the nervous system in humans, although the vital centres are located elsewhere in the so-called primordial brain. The cerebral cortex, in particular, focused the attention of neuroscientists since the early times of the scientific investigation, as discussed in the prologomenon of this chapter. It includes the overwhelming majority of all brain tissues and contains several tens billion neurons. These numbers are already impressive but the connectivity of the cortex is even more impressive.

The slow integration time of nervous cells, operating in the milliseconds range, somewhat a million times slower than presently available supercomputers and the huge number of connections established by a single neuron (in the order of tens of thousands) has suggested that information in the nervous system might be transmitted by simultaneous discharge of a large set of neurons. As much as 99% of all input neural fibres to the cerebral cortex originated from another cortical area (mostly within the same hemisphere). The reflexive nature of corticocortical connections is such that using a metaphor we could say that the cerebral cortex essentially talks to itself. Multiple dimensions of sensory and behaviorally relevant stimuli are processed by thousands of neurons distributed over the many areas of the cortex. The hypothesis that neurons process information along time both individually and jointly following precise time relationships pervaded the Neurosciences since the nervous system was conceptualised as dynamic networks of interacting neurons (McCulloch and Pitts, 1943).

The activity of each cell is necessarily related to the combined activity in the neurons that are afferent to it. Due to the presence of reciprocal connections between cortical areas, re-entrant activity through chains of neurons is likely to occur in all brains. Developmental and/or learning processes are likely to potentiate or weaken certain pathways through the network by affecting the number or efficacy of synaptic interactions between the neurons. Despite the plasticity of these phenomena it is rational to suppose that whenever the same information is presented in the network the same pattern of activity is evoked in a circuit of functionally interconnected neurons, referred to as a cell assembly.

In cell assemblies interconnected in this way some ordered sequences of interspike intervals will recur. Such recurring, ordered, and precise (in the order of few milliseconds) interspike interval relationships are referred to as spatio-temporal patterns of discharges or preferred firing sequences. (Fig. 3).

The term spatio-temporal pattern of discharges encompasses both their precision in time, and the fact that they can occur across different spike trains, even recorded from separate electrodes. For this to be true, temporal firing patterns must occur to a significant level above chance.

Several evidence exist of spatio-temporal firing patterns in behaving animals, from rats to primates (Villa et al., 1999; Shmiel et al., 2005), where preferred firing sequences can be associated to specific types of stimuli or behaviours. Furthermore, recent studies on active propagation of action potentials in the dendrites have provided biophysical models supporting the existence of precise neuronal timing and its modulatory role in determining the strengthening/weakening of the synaptic coupling between pre- and post-synaptic neurons, spike timing dependent plasticity (STDP) (Bell et al., 1997). The synaptic response increased if the pre-synaptic spike preceded the post-synaptic spike, but in the reverse order the synaptic response decreased. The time window for synaptic plasticity to occur is in the order of tens of milliseconds, and a difference in spike timing of only few milliseconds near coincidence may switch plasticity from potentiation to depression.

These findings provide an important support to the general view that a stronger synaptic influence is exerted by multiple converging neurons firing in coincidence, thus making synchrony of firing ideally suited to highlighting responses and to expressing relations among neurons with high temporal precision. An influential and remarkable model based on the assumptions of high temporal precision in brain processing is the synfire chain hypothesis (Abeles, 1982, 1991). This model suggests how precise timing can be sustained in the central nervous system by means of feed-forward chains of convergent/divergent links and re-entry loops between interacting neurons forming a cell assembly (Fig. 4).

A fundamental prediction of synfire chains is that simultaneous recording of activity of cells belonging to the same cell assembly involved repeatedly in the same process should be able to reveal repeated occurrences of such spatio-temporal firing patterns like those observed in the experimental studies cited above. Structures like synfire chains may exhibit patterns of activity where a group of neurons excite themselves and maintain elevated firing rates for a long period and allowing the same neuron to participate in many different synfire chains.

If one assumes that a synfire chain is associated to a basic primitive (i.e. an instance of a certain type of information), then dynamic bindings of synfire chains are able to give rise to compositionality of brain processes. Cognitive processes are likely to be compositional, that is, they must start from primitives and be capable of extracting and building up structured representations by means of progressively complex hierarchical processes (Bienenstock, 1995). Thus, the specific and dynamics relations that bind the primitives, and not only the structure of the primitives themselves, are emphasised in the role of forming the composite expressions (Abeles et al., 2004).

Simultaneous recording of spike trains *ll III I_I Ml -II_£L_

Pattern 1 Pattern 2

(b) Detection of statistically significant spatiotemporal firing patterns Pattern 1 Pattern 2

tt 2

'Complexity c=3 1

repetition number r=3 patterns found r\-2

expected count N=0.04

significance of the patterns excess ^pr{2,0.04} 8-10-4<1Cr^

Rasters of spikes aligned on pattern start

i 11

1

III 1

11

1!

*

1 II 1

1

III II

III II

1 II III

11

1

1! Ill

1 1

*! 1 I

1 I 1

1 1 1 II

'Ii , !i in1 i "1 i!i ii1,

A

\ c

\ B

II 1 1 II

1

Fig. 3 Outline of the general procedure followed by pattern detection algorithms. (a) Analysis of a set of simultaneously recorded spike trains. Three cells, labelled A, B, and C, participate to a patterned activity. Three occurrences of two precise patterns are detected. Each occurrence of the first pattern has been labelled by a specific marker in order to help the reader to identify the corresponding spikes. The spikes belonging to the second pattern are indicated by arrows. (b) Estimation of the statistical significance of the detected patterns. Two patterns, n = 2, <A,C,B> and <C,C,C> were found. Each pattern was formed by three neurons, c = 3, and was repeated three times, r = 3, in the analysed record. The expected number of patterns of this complexity and repetition number was N = 0.04. The probability to observe two or more patterns when 0.04 patterns are expected is noted as pr{0.02, 4}. (c) Display of the pattern occurrences as a raster plot aligned on the patterns start. (Adapted from Tetko and Villa, 2001.)

Fig. 4 A schematic synfire chain characterised by the number (w) of neurons in each pool of the chain, and its multiplicity (m), defined as the number of projections from a neuron in the pool n to a neuron in pool n + 1. The synfire chain is said to be incomplete if m < w, like in this example: w = 5 and m = 3. Note that individual neurons can appear in multiple pools of the same or different synfire chains. (Reproduced from Iglesias, 2005. With permission.)

Fig. 4 A schematic synfire chain characterised by the number (w) of neurons in each pool of the chain, and its multiplicity (m), defined as the number of projections from a neuron in the pool n to a neuron in pool n + 1. The synfire chain is said to be incomplete if m < w, like in this example: w = 5 and m = 3. Note that individual neurons can appear in multiple pools of the same or different synfire chains. (Reproduced from Iglesias, 2005. With permission.)

The unlimited complexity offered by the unbound number of possible temporal combinations of preferred firing sequences permits to the semantics of composite objects to reach a level of complexity far enough to represent the highest brain activities that characterise human thoughts. This model is undoubtedly appealing and is a source of extensive investigation in computational neuroscience, experimental psychology, and neurophysiology. The theoretical framework that has been delineated can address effective elements associated to the problem of binding but the initial question of coding, as it was defined in the previous section, is not yet solved. The fact that a custom-made statistical analysis can detect the significant firing patterns that are associated to cognitive processes does not tell us much about the read-out mechanisms that should be embedded in the neural networks for decoding the transmitted information (Hopfield and Brody, 2001). Following the neuroheuristic approach we prefer to hold on provisional goals so that the question of neural coding per se is temporarily removed and replaced by the study of neural activity in the dynamical perspective.

Was this article helpful?

0 0

Post a comment