Efficient Neural Coding Research Articles

Efficient neural codes can lead to spurious synchronization

Experimental and computational evidence shows that cognitive function requires an optimal balance between global integrative and local functionally specialized processes (Tononi et al., 1998). This balance can be described in terms of transient short-lived episodes of synchronized activity between different parts of the brain (Friston, 2000; Breakspear, 2002). Synchronization over multiple frequency bands is thought to subserve fundamental operations of cortical computation (Varela et al., 2001; Fries, 2009), and to be one of the mechanisms mediating the large-scale coordination of scattered functionally specialized brain regions. For instance, transient synchronization of neuronal oscillatory activity in the 30–80 Hz range has been proposed to act as an integrative mechanism, binding together spatially distributed neural populations in parallel networks during sensory perception and information processing (Singer, 1995; Miltner et al., 1999; Rodriguez et al., 1999). More generally, synchrony may subserve an integrative function in cognitive functions as diverse as motor planning, working or associative memory, or emotional regulation (Varela, 1995). Over the past 15 years, cognitive neuroscientists have tried to capture and quantify neural synchronies across distant brain regions both during spontaneous brain activity and in association with the execution of a wide range of cognitive tasks, using neuroimaging techniques such as functional resonance imaging, electro- or magneto-encephalography. Theoretical advances in various fields including non-linear dynamical systems theory have allowed the study of various types of synchronization from time series (Pereda et al., 2005), and to address important issues such as determining whether observed couplings do not reflect a mere correlation between activities recorded at two different brain regions but rather a causal relationship (Granger, 1969) whereby a brain region would cause the activity of the other one. However, not all measured synchrony may in fact represent neurophysiologically and cognitively relevant computations: various confounding effects may mislead into identifying functional connectivity, defined as the temporal correlations between spatially remote neurophysiological events, with effective connectivity, i.e., the influence one neuronal system exerts over another (Friston, 1994). For instance, measured synchrony may stem from common thalamo-cortical afferents or neuromodulatory input from ascending neurotransmitter systems, or may be the visible part of indirect effective connectivity. Other technique-specific artifactual sources of synchrony, for instance induced by volume conduction, are also well-known to cognitive neuroscientists (Stam et al., 2007). Here, we address a further (extra-cranial) confounding source: the appearance of simultaneous, yet uncorrelated stimuli. We show how the activity of two groups of binary neurons, whose output code is optimized to represent rare events with short codes, can exhibit a synchronization when such rare events appear, even in the absence of shared information or common computational activities.

Adaptation in Sound Localization Processing Induced by Interaural Time Difference in Amplitude Envelope at High Frequencies

BackgroundWhen a second sound follows a long first sound, its location appears to be perceived away from the first one (the localization/lateralization aftereffect). This aftereffect has often been considered to reflect an efficient neural coding of sound locations in the auditory system. To understand determinants of the localization aftereffect, the current study examined whether it is induced by an interaural temporal difference (ITD) in the amplitude envelope of high frequency transposed tones (over 2 kHz), which is known to function as a sound localization cue.Methodology/Principal FindingsIn Experiment 1, participants were required to adjust the position of a pointer to the perceived location of test stimuli before and after adaptation. Test and adapter stimuli were amplitude modulated (AM) sounds presented at high frequencies and their positional differences were manipulated solely by the envelope ITD. Results showed that the adapter's ITD systematically affected the perceived position of test sounds to the directions expected from the localization/lateralization aftereffect when the adapter was presented at ±600 µs ITD; a corresponding significant effect was not observed for a 0 µs ITD adapter. In Experiment 2, the observed adapter effect was confirmed using a forced-choice task. It was also found that adaptation to the AM sounds at high frequencies did not significantly change the perceived position of pure-tone test stimuli in the low frequency region (128 and 256 Hz).Conclusions/SignificanceThe findings in the current study indicate that ITD in the envelope at high frequencies induces the localization aftereffect. This suggests that ITD in the high frequency region is involved in adaptive plasticity of auditory localization processing.

Warm Body Temperature Facilitates Energy Efficient Cortical Action Potentials

The energy efficiency of neural signal transmission is important not only as a limiting factor in brain architecture, but it also influences the interpretation of functional brain imaging signals. Action potential generation in mammalian, versus invertebrate, axons is remarkably energy efficient. Here we demonstrate that this increase in energy efficiency is due largely to a warmer body temperature. Increases in temperature result in an exponential increase in energy efficiency for single action potentials by increasing the rate of Na+ channel inactivation, resulting in a marked reduction in overlap of the inward Na+, and outward K+, currents and a shortening of action potential duration. This increase in single spike efficiency is, however, counterbalanced by a temperature-dependent decrease in the amplitude and duration of the spike afterhyperpolarization, resulting in a nonlinear increase in the spike firing rate, particularly at temperatures above approximately 35°C. Interestingly, the total energy cost, as measured by the multiplication of total Na+ entry per spike and average firing rate in response to a constant input, reaches a global minimum between 37–42°C. Our results indicate that increases in temperature result in an unexpected increase in energy efficiency, especially near normal body temperature, thus allowing the brain to utilize an energy efficient neural code.

Sparse coding in striate and extrastriate visual cortex

Theoretical studies of mammalian cortex argue that efficient neural codes should be sparse. However, theoretical and experimental studies have used different definitions of the term "sparse" leading to three assumptions about the nature of sparse codes. First, codes that have high lifetime sparseness require few action potentials. Second, lifetime-sparse codes are also population-sparse. Third, neural codes are optimized to maximize lifetime sparseness. Here, we examine these assumptions in detail and test their validity in primate visual cortex. We show that lifetime and population sparseness are not necessarily correlated and that a code may have high lifetime sparseness regardless of how many action potentials it uses. We measure lifetime sparseness during presentation of natural images in three areas of macaque visual cortex, V1, V2, and V4. We find that lifetime sparseness does not increase across the visual hierarchy. This suggests that the neural code is not simply optimized to maximize lifetime sparseness. We also find that firing rates during a challenging visual task are higher than theoretical values based on metabolic limits and that responses in V1, V2, and V4 are well-described by exponential distributions. These findings are consistent with the hypothesis that neurons are optimized to maximize information transmission subject to metabolic constraints on mean firing rate.

Effective connectivity of neural networks in automatic movements in Parkinson's disease

Patients with Parkinson's disease (PD) have difficulty in performing learned movements automatically. The neural mechanism of this deficiency remains unclear. In the current study, we used functional MRI (fMRI) and psychophysiological interaction (PPI) methods to investigate the changes in effective connectivity of the brain networks when movements become automatic in PD patients and age-matched normal controls. We found that during automaticity, the rostral supplementary motor area, cerebellum, and cingulate motor area had increased effective connectivity with brain networks in PD patients. In controls, in addition to these regions, the putamen also had automaticity-related strengthened interactions with brain networks. The dorsal lateral prefrontal cortex had more connectivity at the novel stage than in the automatic stage in normal subjects, but not in PD patients. The comparison of the PPI results between the groups showed that the rostral supplementary motor area, cerebellum, and cingulate motor area had significantly more increased effective connectivity with several regions in normal subjects than in PD. The changes of effective connectivity in some areas negatively correlated with the Unified Parkinson's Disease Rating Scale (UPDRS). Our findings show that some of the factors related to PD patients having difficulty achieving automaticity are less efficient neural coding of movement and failure to shift execution of automatic movements more subcortically. The changes of effective connectivity become more abnormal as the disorder progresses. In addition, in PD, the connections of the attentional networks are altered.

Internally Mediated Developmental Desynchronization of Neocortical Network Activity

During neocortical development, neurons exhibit highly synchronized patterns of spontaneous activity, with correlated bursts of action potential firing dominating network activity. This early activity is eventually replaced by more sparse and decorrelated firing of cortical neurons, which modeling studies predict is a network state that is better suited for efficient neural coding. The precise time course and mechanisms of this crucial transition in cortical network activity have not been characterized in vivo. We used in vivo two-photon calcium imaging in combination with whole-cell recordings in both unanesthetized and anesthetized mice to monitor how spontaneous activity patterns in ensembles of layer 2/3 neurons of barrel cortex mature during postnatal development. We find that, as early as postnatal day 4, activity is highly synchronous within local clusters of neurons. At the end of the second postnatal week, neocortical networks undergo a transition to a much more desynchronized state that lacks a clear spatial structure. Strikingly, deprivation of sensory input from the periphery had no effect on the time course of this transition. Therefore, developmental desynchronization of spontaneous neuronal activity is a fundamental network transition in the neocortex that appears to be intrinsically generated.

Modifications of the interactions in the motor networks when a movement becomes automatic

A crucial feature of the motor system is the ability to control some movements automatically. We have previously shown that all parts of the motor networks reduce their activity with automaticity, and, while this change may indicate increased efficiency in terms of neural processing, it is not clear how motor skill can be maintained after a reduction of neural activity. In the current study, we used functional MRI (fMRI) to investigate influences on the effective connectivity of the brain motor networks when movements become automatic. Subjects practiced a sequential movement until they could execute it automatically, and task-related brain fMRI activation was measured before and after they achieved automaticity. Using the psychophysiological interaction (PPI) method, we found that the cerebellum, cingulate motor area, supplementary motor area, and putamen had significantly greater connectivity, whereas the precuneus had less connectivity in the motor networks at the automatic stage. Our findings demonstrate that the importance of the attention networks decrease when movements become automatic. Moreover, the process of automaticity is accompanied by a strengthened interaction of central motor networks even though the magnitude of the activation is decreased. We speculate that this increase in connectivity reflects more efficient neural coding of movement at the automatic stage.

Efficient motor control: how can less be more?

The results of a scientific study sometimes surprise us, and at other times comfort us. Such reactions reflect the framework of expectations we had before the study was carried out. This framework ideally summarizes our shared prior knowledge and makes new results seem more or less plausible in light of that knowledge. In the present issue of The Journal of Physiology, Wu et al. (2008) report a comforting result. When a new motor task is practiced to the point that subjects can perform it automatically, that is, without degradation of performance when a second task is concurrently carried out, a group of motor areas in the brain become more strongly connected. The same authors had previously shown that when a movement is practiced until it becomes automatic, a major change in brain activation pattern is a reduction of neural activity in task-related areas (Wu et al. 2004). That is, as movement guidance changes from a ‘novice’ mode, requiring attention, to an automatic mode, there is no qualitative change in the pattern of brain activation: the same areas are activated by the task before and after practice. However, these areas become able to control the same movements with a reduced amount of activation. This was (or at least should have been) a surprising result. It might be tempting to interpret reduction of activation as the neural equivalent of subtraction of task components: activity before practice = attention + motor control; reduced activity after practice = motor control alone. However, there is a paradox in this interpretation. If the difference in brain activity reflected only a difference in attention, this would imply that the portion of activity responsible for motor control is the same before and after practice. But then how can motor control after practice be resistant to distraction by a concurrent task? In order for automaticity to have arisen, the neural substrate responsible for controlling the movement must have changed: if the same movements are now resistant to interference by a simultaneous second task, they must be guided by a different motor control strategy, with a presumably different neural code. How could a reduction of activation in motor areas be responsible for the radical change in the nature of motor control associated with automatic execution? The answer provided by Wu et al. in the present study resolves this paradox, and is reassuring at three levels. First, the authors identified a change in functional connectivity among motor areas associated with the development of automaticity. This finding reassures us that the neural substrate responsible for automatic motor control is indeed different from that responsible for novice performance. Second, the change identified is of the type we have come to expect, ever since Donald Hebb's postulate (Hebb, 1949), to be associated with learning: if a movement must be practiced in order to become automatic, then learning must occur, and thus synaptic strengths must change. Third, the combined findings of Wu et al.'s original (Wu et al. 2004) and present studies, are evidence of an increase in neural efficiency: the combination of reduction of activation and strengthening of functional connectivity suggests a more efficient neural code for controlling a given task. Neural efficiency has long been investigated in relationship to human intelligence (Vernon, 1993), but these studies have been fraught with the long list of potential confounding factors (task complexity, use of strategy, past experience) associated with the study of such a complex measure of nervous system function. A task as simple as tapping fingers in a sequence, when combined with analysis not only of brain activation patterns, but also of functional connectivity, offers a more easily controlled system for investigating concepts like neural efficiency. By presenting evidence of increased neural efficiency in the automatic performance of movement sequences, Wu et al. have brought an exciting new level of detail to our understanding of the neural code underlying movement.

Imprecise correlated activity in self-organizing maps of spiking neurons

How neurons communicate with each other to form effective circuits providing support to functional features of the nervous system is currently under debate. While many experts argue the existence of sparse neural codes based either on oscillations, neural assemblies or synchronous fire chains, other studies defend the necessity of a precise inter-neural communication to arrange efficient neural codes. As it has been demonstrated in neurophysiological studies, in the visual pathway between the retina and the visual cortex of mammals, the correlated activity among neurons becomes less precise as a direct consequence of an increase in the variability of synaptic transmission latencies. Although it is difficult to measure the influence of this reduction of correlated firing precision on the self-organization of cortical maps, it does not preclude the emergence of receptive fields and orientation selectivity maps. This is in close agreement with authors who consider that codes for neural communication are sparse. In this article, integrate-and-fire neural networks are simulated to analyze how changes in the precision of correlated firing among neurons affect self-organization. We observe how by keeping these changes within biologically realistic ranges, orientation selectivity maps can emerge and the features of neuronal receptive fields are significantly affected.

Neural Coding of Natural Stimuli: Information at Sub-Millisecond Resolution

Sensory information about the outside world is encoded by neurons in sequences of discrete, identical pulses termed action potentials or spikes. There is persistent controversy about the extent to which the precise timing of these spikes is relevant to the function of the brain. We revisit this issue, using the motion-sensitive neurons of the fly visual system as a test case. Our experimental methods allow us to deliver more nearly natural visual stimuli, comparable to those which flies encounter in free, acrobatic flight. New mathematical methods allow us to draw more reliable conclusions about the information content of neural responses even when the set of possible responses is very large. We find that significant amounts of visual information are represented by details of the spike train at millisecond and sub-millisecond precision, even though the sensory input has a correlation time of ∼55 ms; different patterns of spike timing represent distinct motion trajectories, and the absolute timing of spikes points to particular features of these trajectories with high precision. Finally, the efficiency of our entropy estimator makes it possible to uncover features of neural coding relevant for natural visual stimuli: first, the system's information transmission rate varies with natural fluctuations in light intensity, resulting from varying cloud cover, such that marginal increases in information rate thus occur even when the individual photoreceptors are counting on the order of one million photons per second. Secondly, we see that the system exploits the relatively slow dynamics of the stimulus to remove coding redundancy and so generate a more efficient neural code.

Neural coding of natural stimuli: information at sub-millisecond resolution

Our knowledge of the sensory world is encoded by neurons in sequences of discrete, identical pulses termed action potentials or spikes. There is persistent controversy about the extent to which the precise timing of these spikes is relevant to the function of the brain. We revisit this issue, using the motion – sensitive neurons of the fly visual system as a test case. New experimental methods allow us to deliver more nearly natural visual stimuli, comparable to those which flies encounter in free, acrobatic flight, and new mathematical methods allow us to draw more reliable conclusions about the information content of neural responses even when the set of possible responses is very large. We find that significant amounts of visual information are represented by details of the spike train at millisecond and sub-millisecond precision, even though the sensory input has a correlation time of ~60 ms; different patterns of spike timing represent distinct motion trajectories, and the absolute timing of spikes points to particular features of these trajectories with high precision. Under these naturalistic conditions, the system's information transmission rate still increases with higher photon flux, even though individual photoreceptors are counting more than one million photons per second. Further, exploiting the relatively slow dynamics of the stimulus, the system removes redundancy and so generates a more efficient neural code.

Illusory Bands in Orientation and Spatial Frequency: A Cortical Analog to Mach Bands

Illusory bands at a luminance transition in space (ie an edge) are well known. Here we demonstrate illusory bands of enhanced orientations or spatial frequencies at transitions between higher-contrast and lower-contrast image content along the orientation and spatial-frequency dimensions--the dimensions of cortical spatial coding. We conclude that this illusion is a consequence of cortical-level suppression of units of similar orientations and spatial frequencies and serves to aid texture segmentation while providing efficient neural coding.

Adaptation and temporal decorrelation by single neurons in the primary visual cortex.

Limiting redundancy in the real-world sensory inputs is of obvious benefit for efficient neural coding, but little is known about how this may be accomplished by biophysical neural mechanisms. One possible cellular mechanism is through adaptation to relatively constant inputs. Recent investigations in primary visual (V1) cortical neurons have demonstrated that adaptation to prolonged changes in stimulus contrast is mediated in part through intrinsic ionic currents, a Ca2+-activated K+ current (IKCa) and especially a Na+-activated K+ current (IKNa). The present study was designed to test the hypothesis that the activation of adaptation ionic currents may provide a cellular mechanism for temporal decorrelation in V1. A conductance-based neuron model was simulated, which included an IKCa and an IKNa. We show that the model neuron reproduces the adaptive behavior of V1 neurons in response to high contrast inputs. When the stimulus is stochastic with 1/f 2 or 1/f-type temporal correlations, these autocorrelations are greatly reduced in the output spike train of the model neuron. The IKCa is effective at reducing positive temporal correlations at approximately 100-ms time scale, while a slower adaptation mediated by IKNa is effective in reducing temporal correlations over the range of 1-20 s. Intracellular injection of stochastic currents into layer 2/3 and 4 (pyramidal and stellate) neurons in ferret primary visual cortical slices revealed neuronal responses that exhibited temporal decorrelation in similarity with the model. Enhancing the slow afterhyperpolarization resulted in a strengthening of the decorrelation effect. These results demonstrate the intrinsic membrane properties of neocortical neurons provide a mechanism for decorrelation of sensory inputs.

Coding of Natural Scenes in Primary Visual Cortex

Natural scene coding in ferret visual cortex was investigated using a new technique for multi-site recording of neuronal activity from the cortical surface. Surface recordings accurately reflected radially aligned layer 2/3 activity. At individual sites, evoked activity to natural scenes was weakly correlated with the local image contrast structure falling within the cells' classical receptive field. However, a population code, derived from activity integrated across cortical sites having retinotopically overlapping receptive fields, correlated strongly with the local image contrast structure. Cell responses demonstrated high lifetime sparseness, population sparseness, and high dispersal values, implying efficient neural coding in terms of information processing. These results indicate that while cells at an individual cortical site do not provide a reliable estimate of the local contrast structure in natural scenes, cell activity integrated across distributed cortical sites is closely related to this structure in the form of a sparse and dispersed code.

An energy budget for signaling in the grey matter of the brain.

Anatomic and physiologic data are used to analyze the energy expenditure on different components of excitatory signaling in the grey matter of rodent brain. Action potentials and postsynaptic effects of glutamate are predicted to consume much of the energy (47% and 34%, respectively), with the resting potential consuming a smaller amount (13%), and glutamate recycling using only 3%. Energy usage depends strongly on action potential rate--an increase in activity of 1 action potential/cortical neuron/s will raise oxygen consumption by 145 mL/100 g grey matter/h. The energy expended on signaling is a large fraction of the total energy used by the brain; this favors the use of energy efficient neural codes and wiring patterns. Our estimates of energy usage predict the use of distributed codes, with <or=15% of neurons simultaneously active, to reduce energy consumption and allow greater computing power from a fixed number of neurons. Functional magnetic resonance imaging signals are likely to be dominated by changes in energy usage associated with synaptic currents and action potential propagation.

Characterizing the sparseness of neural codes

It is often suggested that efficient neural codes for natural visual information should be 'sparse'. However, the term 'sparse' has been used in two different ways--firstly to describe codes in which few neurons are active at any time ('population sparseness'), and secondly to describe codes in which each neuron's lifetime response distribution has high kurtosis ('lifetime sparseness'). Although these ideas are related, they are not identical, and the most common measure of lifetime sparseness--the kurtosis of the lifetime response distributions of the neurons--provides no information about population sparseness. We have measured the population sparseness and lifetime kurtosis of several biologically inspired coding schemes. We used three measures of population sparseness (population kurtosis, Treves-Rolls sparseness and 'activity sparseness'), and found them to be in close agreement with one another. However, we also measured the lifetime kurtosis of the cells in each code. We found that lifetime kurtosis is uncorrelated with population sparseness for the codes we used. Lifetime kurtosis is not, therefore, a useful measure of the population sparseness of a code. Moreover, the Gabor-like codes, which are often assumed to have high population sparseness (since they have high lifetime kurtosis), actually turned out to have rather low population sparseness. Surprisingly, principal components filters produced the codes with the highest population sparseness.

Characterizing the sparseness of neural codes

It is often suggested that efficient neural codes for natural visual information should be `sparse'. However, the term `sparse' has been used in two different ways - firstly to describe codes in which few neurons are active at any time (`population sparseness'), and secondly to describe codes in which each neuron's lifetime response distribution has high kurtosis (`lifetime sparseness'). Although these ideas are related, they are not identical, and the most common measure of lifetime sparseness - the kurtosis of the lifetime response distributions of the neurons - provides no information about population sparseness. We have measured the population sparseness and lifetime kurtosis of several biologically inspired coding schemes. We used three measures of population sparseness (population kurtosis, Treves-Rolls sparseness and `activity sparseness'), and found them to be in close agreement with one another. However, we also measured the lifetime kurtosis of the cells in each code. We found that lifetime kurtosis is uncorrelated with population sparseness for the codes we used. Lifetime kurtosis is not, therefore, a useful measure of the population sparseness of a code. Moreover, the Gabor-like codes, which are often assumed to have high population sparseness (since they have high lifetime kurtosis), actually turned out to have rather low population sparseness. Surprisingly, principal components filters produced the codes with the highest population sparseness.

Characterizing the sparseness of neural codes

It is often suggested that efficient neural codes for natural visual information should be `sparse'. However, the term `sparse' has been used in two different ways - firstly to describe codes in which few neurons are active at any time (`population sparseness'), and secondly to describe codes in which each neuron's lifetime response distribution has high kurtosis (`lifetime sparseness'). Although these ideas are related, they are not identical, and the most common measure of lifetime sparseness - the kurtosis of the lifetime response distributions of the neurons - provides no information about population sparseness.We have measured the population sparseness and lifetime kurtosis of several biologically inspired coding schemes. We used three measures of population sparseness (population kurtosis, Treves-Rolls sparseness and `activity sparseness'), and found them to be in close agreement with one another. However, we also measured the lifetime kurtosis of the cells in each code. We found that lifetime kurtosis is uncorrelated with population sparseness for the codes we used.Lifetime kurtosis is not, therefore, a useful measure of the population sparseness of a code. Moreover, the Gabor-like codes, which are often assumed to have high population sparseness (since they have high lifetime kurtosis), actually turned out to have rather low population sparseness. Surprisingly, principal components filters produced the codes with the highest population sparseness.

Effects of the synaptic transmission's dynamics on possible neural codes

To examine the effects of paired pulse facilitation, long-term synaptic modifications as well as spike frequency adaptation on neural signal transmission, a simple model was applied. This way various input–output properties of the model units were described. Particularly, the transmission of the mean and S.D. of the simulated synaptic currents were studied. The results indicate that the transfer of the mean value of the membrane currents cannot be described in terms of synaptic weights. So firing rate can hardly be an efficient neural code, especially for adaptive channels of the central nervous system (CNS). On the contrary, the transfer of S.D. of synaptic currents behaves in accordance with the synaptic weights. So it is supported that S.D. of the synaptic currents is a biological relevant subclass of the variation codes [see Perkel, H., Bullock, T.H., 1968. Neurol coding. Neurosci. Res. Program Bull. 6, 221–344]. It is discussed how this code can be established and how it works.