Encoding of tactile stimulus location by somatosensory thalamocortical ensembles

Spatiotemporal characteristics of neuronal sensory integration in the barrel cortex of the rat

In primary sensory cortices, neuronal responses to a stimulus presented as part of a rapid sequence often differ from responses to an isolated stimulus. It has been reported that sequential stimulation of two whiskers produces facilitatory modulations of barrel cortex neuronal responses. These results are at odds with the well-known suppressive interaction that has been usually described. Herein, we have examined the dependency of response modulation on the spatiotemporal pattern of stimulation by varying the spatial arrangement of the deflected vibrissae, the temporal frequency of stimulation, and the time interval between whisker deflections. Extracellular recordings were made from primary somatosensory cortex of anesthetized rats. Two contralateral whiskers were stimulated at 0.5 and 8 Hz at intervals ranging from 0 to +/-30 ms. Response interactions were assessed during stimulation of the principal and adjacent whiskers, first from the same row and second from the same arc. When tested at 0.5 Hz, 59% of single units showed a statistically significant suppressive interaction, whereas response facilitation was found in only 6% of cells. In contrast, at 8 Hz, a significant supralinear summation was observed in 19% of the cells, particularly for stimulations along an arc rather than along a row. Multi-unit recordings showed similar results. These observations indicate that most of the interactions in the barrel cortex during two-whisker stimulation are suppressive. However, facilitation can be revealed when stimuli are applied at a physiological frequency and could be the basis for internal representations of the spatiotemporal pattern of the stimulus.

PSTH-based classification of sensory stimuli using ensembles of single neurons

The problem of understanding how ensembles of neurons code for somatosensory information has been defined as a classification problem: given the response of a population of neurons to a set of stimuli, which stimulus generated the response on a single-trial basis? Multivariate statistical techniques such as linear discriminant analysis (LDA) and artificial neural networks (ANNs), and different types of preprocessing stages, such as principal and independent component analysis, have been used to solve this classification problem, with surprisingly small performance differences. Therefore, the goal of this project was to design a new method to maximize computational efficiency rather than classification performance. We developed a peri-stimulus time histogram (PSTH)-based method, which consists of creating a set of templates based on the average neural responses to stimuli and classifying each single trial by assigning it to the stimulus with the 'closest' template in the Euclidean distance sense. The PSTH-based method is computationally more efficient than methods as simple as linear discriminant analysis, performs significantly better than discriminant analyses (linear, quadratic or Mahalanobis) when small binsizes are used (1 ms) and as well as LDA with any other binsize, is optimal among other minimum-distance classifiers and can be optimally applied on raw neural data without a previous stage of dimension reduction. We conclude that the PSTH-based method is an efficient alternative to more sophisticated methods such as LDA and ANNs to study how ensemble of neurons code for discrete sensory stimuli, especially when datasets with many variables are used and when the time resolution of the neural code is one of the factors of interest.

Modeling and Decoding Motor Cortical Activity Using a Switching Kalman Filter

We present a switching Kalman filter model for the real-time inference of hand kinematics from a population of motor cortical neurons. Firing rates are modeled as a Gaussian mixture where the mean of each Gaussian component is a linear function of hand kinematics. A "hidden state" models the probability of each mixture component and evolves over time in a Markov chain. The model generalizes previous encoding and decoding methods, addresses the non-Gaussian nature of firing rates, and can cope with crudely sorted neural data common in on-line prosthetic applications.

Wavelet-based processing of neuronal spike trains prior to discriminant analysis

Investigations of neural coding in many brain systems have focused on the role of spike rate and timing as two means of encoding information within a spike train. Recently, statistical pattern recognition methods, such as linear discriminant analysis (LDA), have emerged as a standard approach for examining neural codes. These methods work well when data sets are over-determined (i.e., there are more observations than predictor variables). But this is not always the case in many experimental data sets. One way to reduce the number of predictor variables is to preprocess data prior to classification. Here, a wavelet-based method is described for preprocessing spike trains. The method is based on the discriminant pursuit (DP) algorithm of Buckheit and Donoho [Proc. SPIE 2569 (1995) 540-51]. DP extracts a reduced set of features that are well localized in the time and frequency domains and that can be subsequently analyzed with statistical classifiers. DP is illustrated using neuronal spike trains recorded in the motor cortex of an awake, behaving rat [Laubach et al. Nature 405 (2000) 567-71]. In addition, simulated spike trains that differed only in the timing of spikes are used to show that DP outperforms another method for preprocessing spike trains, principal component analysis (PCA) [Richmond and Optican J. Neurophysiol. 57 (1987) 147-61].

Long-lasting novelty-induced neuronal reverberation during slow-wave sleep in multiple forebrain areas

The discovery of experience-dependent brain reactivation during both slow-wave (SW) and rapid eye-movement (REM) sleep led to the notion that the consolidation of recently acquired memory traces requires neural replay during sleep. To date, however, several observations continue to undermine this hypothesis. To address some of these objections, we investigated the effects of a transient novel experience on the long-term evolution of ongoing neuronal activity in the rat forebrain. We observed that spatiotemporal patterns of neuronal ensemble activity originally produced by the tactile exploration of novel objects recurred for up to 48 h in the cerebral cortex, hippocampus, putamen, and thalamus. This novelty-induced recurrence was characterized by low but significant correlations values. Nearly identical results were found for neuronal activity sampled when animals were moving between objects without touching them. In contrast, negligible recurrence was observed for neuronal patterns obtained when animals explored a familiar environment. While the reverberation of past patterns of neuronal activity was strongest during SW sleep, waking was correlated with a decrease of neuronal reverberation. REM sleep showed more variable results across animals. In contrast with data from hippocampal place cells, we found no evidence of time compression or expansion of neuronal reverberation in any of the sampled forebrain areas. Our results indicate that persistent experience-dependent neuronal reverberation is a general property of multiple forebrain structures. It does not consist of an exact replay of previous activity, but instead it defines a mild and consistent bias towards salient neural ensemble firing patterns. These results are compatible with a slow and progressive process of memory consolidation, reflecting novelty-related neuronal ensemble relationships that seem to be context- rather than stimulus-specific. Based on our current and previous results, we propose that the two major phases of sleep play distinct and complementary roles in memory consolidation: pretranscriptional recall during SW sleep and transcriptional storage during REM sleep.

Rats exposed to novel objects during periods of wakefulness generate neural activity that is correlated with patterns observed in subsequent sleep episodes

The human primary somatosensory cortex response contains components related to stimulus frequency and perception in a frequency discrimination task

Somatosensory stimulation of primary somatosensory cortex (SI) using frequency discrimination offers a direct, well-defined and accessible way of studying cortical decisions at the locus of early input processing. Animal studies have identified and classified the neuronal responses in SI but they have not yet resolved whether during prolonged stimulation the collective SI response just passively reflects the input or actively participates in the comparison and decision processes. This question was investigated using tomographic analysis of single trial magnetoencephalographic data. Four right-handed males participated in a frequency discrimination task to detect changes in the frequency of an electrical stimulus applied to the right-hand digits 2+3+4. The subjects received approximately 600 pairs of stimuli with Stim1 always at 21 Hz, while Stim2 was either 21 Hz (50%) or varied from 22 to 29 Hz in steps of 1 Hz. Both stimuli were 1 s duration, separated by a 1 s interval of no stimulation. The left-SI was the most consistently activated area and showed the first activation peak at 35-48 ms after Stim1 onset and sustained activity during both stimulus periods. During the Stim2 period, we found that the left-SI activation started to differ significantly between two groups of trials (21 versus 26-29 Hz) within the first 100 ms and this difference was sustained and enhanced thereafter (approximately 600 ms). When only correct responses from the above two groups were used, the difference was even higher at later latencies (approximately 650 ms). For one subject who had enough trials of same perception to different input frequencies, e.g. responded 21 Hz to Stim2 at 21 Hz (correct) and 26-29 Hz (error), we found the sustained difference only before 650 ms. Our results suggest that SI is involved with the analysis of an input frequency and related to perception and decision at different latencies.

Neuronal interactions are higher in the cortex than thalamus in the somatosensory pathway

Previous studies have shown significant correlated discharges (noise correlation) and synergistic information coding among adjacent cortical neurons. In order to investigate whether such interactions are present at an earlier stage of sensory processing, we compared noise correlation and synergistic information transmission in the ventral posterolateral nucleus (VPLn) of thalamus and primary somatosensory cortex (SI) of anesthetized rats. A hind paw was stimulated electrically and responses of several neighboring neurons were recorded simultaneously with a tetrode. Analyses indicated that noise correlation in the SI was about four times higher than in the VPLn, and, interestingly, it was significantly reduced following sensory stimulation in both regions. Spike count distributions of individual VPLn units contained higher amounts of information about the delivery of external stimulation compared with those of SI units. When simultaneously recorded units were considered together, transmission of information was more interactive (synergistic or redundant) among SI than VPLn units. On average, information transmission was independent in the VPLn, but synergistic in the SI. The difference in synergistic information coding was largely attributable to different levels of noise correlation and their modulation by external sensory stimulation. These results indicate that neuronal interactions are relatively low at the thalamic level, but much enhanced at the cortical level along the somatosensory pathway. The enhanced neuronal interactions in the cortex may reflect the role of cortex in extracting higher features of sensory stimuli.

Coding of auditory temporal and pitch information by hippocampal individual cells and cell assemblies in the rat

This study reports how hippocampal individual cells and cell assemblies cooperate for neural coding of pitch and temporal information in memory processes for auditory stimuli. Each rat performed two tasks, one requiring discrimination of auditory pitch (high or low) and the other requiring discrimination of their duration (long or short). Some CA1 and CA3 complex-spike neurons showed task-related differential activity between the high and low tones in only the pitch-discrimination task. However, without exception, neurons which showed task-related differential activity between the long and short tones in the duration-discrimination task were always task-related neurons in the pitch-discrimination task. These results suggest that temporal information (long or short), in contrast to pitch information (high or low), cannot be coded independently by specific neurons. The results also indicate that the two different behavioral tasks cannot be fully differentiated by the task-related single neurons alone and suggest a model of cell-assembly coding of the tasks. Cross-correlation analysis among activities of simultaneously recorded multiple neurons supported the suggested cell-assembly model.Considering those results, this study concludes that dual coding by hippocampal single neurons and cell assemblies is working in memory processing of pitch and temporal information of auditory stimuli. The single neurons encode both auditory pitches and their temporal lengths and the cell assemblies encode types of tasks (contexts or situations) in which the pitch and the temporal information are processed.

Role of thalamocortical sensory suppression during arousal: focusing sensory inputs in neocortex

The thalamus serves as a gate that regulates the flow of sensory inputs to the neocortex, and this gate is controlled by neuromodulators from the brainstem reticular formation that are released during arousal. We found recently that sensory-evoked responses are suppressed in the neocortex during arousal. This sensory suppression results from the activity-dependent depression of the thalamocortical connection caused by increased tonic firing of thalamocortical cells during arousal. In the present study, the functional consequences of thalamocortical suppression during arousal were investigated using the vibrissae system of rodents. The results show that thalamocortical suppression is associated with a strong reduction in the spread of sensory inputs through the cortex, thus reducing the size of sensory representations. In addition, when the responses of single cells to principal and adjacent whiskers are compared, the response to the adjacent whiskers was found to be strongly suppressed, much more so than that of principal whiskers. Consequently, the receptive fields of cortical neurons become more focused to the principal whisker. The results indicate that thalamocortical suppression during arousal serves to focus sensory inputs to their appropriate representations in neocortex, which may be computationally helpful for the spatial processing of sensory information.

Thalamocortical [correction of Thalamcortical] optimization of tactile processing according to behavioral state

We propose a conceptual model that describes the operation of the main thalamocortical loop of the rat somatosensory system. According to this model, the asynchronous convergence of ascending and descending projections dynamically alters the physiological properties of thalamic neurons in the ventral posterior medial (VPM) nucleus as rats shift between three behavioral states. Two of these states are characterized by distinct modes of rhythmic whisker movements. We posit that these simultaneous shifts in exploratory behavioral strategy and in the physiological properties of VPM neurons allow rats to either (i) optimize the detection of stimuli that are novel or difficult to sense or (ii) process complex patterns of multi-whisker stimulation.

Cortical sensory suppression during arousal is due to the activity-dependent depression of thalamocortical synapses

The thalamus serves as a gate that regulates the flow of sensory inputs to the neocortex, and this gate is controlled by neuromodulators from the brainstem reticular formation that are released during arousal. Here we show in rats that sensory-evoked responses were suppressed in the neocortex by activating the brainstem reticular formation and during natural arousal. Sensory suppression occurred at the thalamocortical connection and was a consequence of the activity-dependent depression of thalamocortical synapses caused by increased thalamocortical tonic firing during arousal. Thalamocortical suppression may serve as a mechanism to focus sensory inputs to their appropriate representations in neocortex, which is helpful for the spatial processing of sensory information.

Cortical representation of auditory space: information-bearing features of spike patterns

Previous studies have demonstrated that the spike patterns of cortical neurons vary systematically as a function of sound-source location such that the response of a single neuron can signal the location of a sound source throughout 360 degrees of azimuth. The present study examined specific features of spike patterns that might transmit information related to sound-source location. Analysis was based on responses of well-isolated single units recorded from cortical area A2 in alpha-chloralose-anesthetized cats. Stimuli were 80-ms noise bursts presented from loudspeakers in the horizontal plane; source azimuths ranged through 360 degrees in 20 degrees steps. Spike patterns were averaged across samples of eight trials. A competitive artificial neural network (ANN) identified sound-source locations by recognizing spike patterns; the ANN was trained using the learning vector quantization learning rule. The information about stimulus location that was transmitted by spike patterns was computed from joint stimulus-response probability matrices. Spike patterns were manipulated in various ways to isolate particular features. Full-spike patterns, which contained all spike-count information and spike timing with 100-micros precision, transmitted the most stimulus-related information. Transmitted information was sensitive to disruption of spike timing on a scale of more than approximately 4 ms and was reduced by an average of approximately 35% when spike-timing information was obliterated entirely. In a condition in which all but the first spike in each pattern were eliminated, transmitted information decreased by an average of only approximately 11%. In many cases, that condition showed essentially no loss of transmitted information. Three unidimensional features were extracted from spike patterns. Of those features, spike latency transmitted approximately 60% more information than that transmitted either by spike count or by a measure of latency dispersion. Information transmission by spike patterns recorded on single trials was substantially reduced compared with the information transmitted by averages of eight trials. In a comparison of averaged and nonaveraged responses, however, the information transmitted by latencies was reduced by only approximately 29%, whereas information transmitted by spike counts was reduced by 79%. Spike counts clearly are sensitive to sound-source location and could transmit information about sound-source locations. Nevertheless, the present results demonstrate that the timing of the first poststimulus spike carries a substantial amount, probably the majority, of the location-related information present in spike patterns. The results indicate that any complete model of the cortical representation of auditory space must incorporate the temporal characteristics of neuronal response patterns.

Fine-scale organization of SI (area 3b) in the squirrel monkey revealed with intrinsic optical imaging

Optical imaging of intrinsic cortical activity was used to study the somatotopic map and the representation of pressure, flutter, and vibration in area 3b of the squirrel monkey (Saimiri sciureus) cortex under pentothal or isoflurane anesthesia. The representation of the fingerpads in primary somatosensory cortex was investigated by stimulating the glabrous skin of distal fingerpads (D1-D5) with Teflon probes (3-mm diam) attached through an armature to force feedback-controlled torque motors. Under pentothal anesthesia, intrinsic signal maps in area 3b obtained in response to stimulation (trapezoidal indentation) of individual fingerpads showed focal activations. These activations (ranging from 0.5 to 1.0 mm) were discrete and exhibited minimal overlap between adjacent fingerpad representations. Consistent with previously published maps, a somatotopic representation of the fingerpads was observed with an orderly medial to lateral progression from the D5 to D1 fingerpads. Under isoflurane anesthesia, general topography was still maintained, but the representation of fingerpads on adjacent fingers had higher degrees of overlap than with pentothal anesthesia. Multi- and single-unit recordings in the activation zones confirmed the somatotopic maps. To examine preferential inputs from slowly adapting type I (SA) and rapidly adapting type I (RA) and type II (PC) mechanoreceptors, we applied stimuli consisting of sinusoidal indentations that produce sensations of pressure (1 Hz), flutter (30 Hz), and vibration (200 Hz). Under pentothal anesthesia, activation patterns to these different stimuli were focal and coincided on the cortex. Under isoflurane, activation zones from pressure, flutter, and vibratory stimuli differed in size and shape and often contained multiple foci, although overall topography was maintained. Subtraction and vector maps revealed cortical areas (approximate 250-microm diam) that were preferentially activated by the sensations of pressure, flutter, and vibration. Multi- and single-unit recordings aided in the interpretation of the imaging maps. In conclusion, the cortical signals observed with intrinsic signal optical imaging delineated a somatotopic organization of area 3b and revealed different topographical cortical activation patterns for pressure, flutter, and vibratory stimuli. These patterns were dependent on anesthesia type. Possible relationships of these anesthesia effects to somatosensory cortical plasticity are discussed.

Population coding of stimulus location in rat somatosensory cortex

This study explores the nature of population coding in sensory cortex by applying information theoretic analyses to neuron pairs recorded simultaneously from rat barrel cortex. We quantified the roles of individual spikes and spike patterns in encoding whisker stimulus location. 82%-85% of the total information was contained in the timing of individual spikes: first spike time was particularly crucial. Spike patterns within neurons accounted for the remaining 15%-18%. Neuron pairs located in the same barrel column coded redundantly, whereas pairs in neighboring barrel columns coded independently. The barrel cortical population code for stimulus location appears to be the time of single neurons' first poststimulus spikes-a fast, robust coding mechanism that does not rely on "synergy" in crossneuronal spike patterns.

Figuring space by time

Sensory information is encoded both in space and in time. Spatial encoding is based on the identity of activated receptors, while temporal encoding is based on the timing of activation. In order to generate accurate internal representations of the external world, the brain must decode both types of encoded information, even when processing stationary stimuli. We review here evidence in support of a parallel processing scheme for spatially and temporally encoded information in the tactile system and discuss the advantages and limitations of sensory-derived temporal coding in both the tactile and visual systems. Based on a large body of data, we propose a dynamic theory for vision, which avoids the impediments of previous dynamic theories.

Tactile discrimination of edge shape: limits on spatial resolution imposed by parameters of the peripheral neural population

When the flat faces of a coin are grasped between thumb and index finger, a "curved edge" is felt. Analogous curved edges were generated by our stimuli, which comprised the flat face of segments of annuli applied passively to immobilized fingers. Humans could scale the curvature of the annulus and could discriminate changes in curvature of approximately 20 m(-1). The responses of single slowly adapting type I afferents (SAIs) recorded in anesthetized monkeys could be quantified by the product of two factors: their sensitivity and a spatial profile dependent only on the radius of the annulus. This allowed us to reconstruct realistic SAI population responses that included noise, variation in fiber sensitivity, and varying innervation patterns. The critical question was how relatively small populations ( approximately 70 active fibers) can encode edge curvature with such precision. A template-matching approach was used to establish the accuracy of edge representation in the population. The known large interfiber variability in sensitivity had no effect on curvature resolution. Neural resolution was superior to human performance until large levels of central noise were present showing that, unlike simple detection, spatial processing is limited centrally. In contrast to the behavior of mean response codes, neural resolution improved with increasing covariance in noise. Surprisingly, resolution for any single population varied considerably with small changes in the position of the stimulus relative to the SAI matrix. Overall innervation density was not as critical as the spacing of receptive fields at right angles to the edge.

Techniques for long-term multisite neuronal ensemble recordings in behaving animals

Advances in our understanding of neural systems will go hand in hand with improvements in the experimental techniques used to study these systems. This article describes a series of methodological developments aimed at enhancing the power of the methods needed to record simultaneously from populations of neurons over broad regions of the brain in awake, behaving animals. First, our laboratory has made many advances in electrode design, including movable bundle and array electrodes and smaller electrode assemblies. Second, to perform longer and more complex multielectrode implantation surgeries in primates, we have modified our surgical procedures by employing comprehensive physiological monitoring akin to human neuroanesthesia. We have also developed surgical implantation techniques aimed at minimizing brain tissue damage and facilitating penetration of the cortical surface. Third, we have integrated new technologies into our neural ensemble, stimulus and behavioral recording experiments to provide more detailed measurements of experimental variables. Finally, new data analytical techniques are being used in the laboratory to analyze increasingly large quantities of data.

Temporal frequency of whisker movement. I. Representations in brain stem and thalamus

How does processing of information change the internal representations used in subsequent stages of sensory pathways? To approach this question, we studied the representations of whisker movements in the lemniscal and paralemniscal pathways of the rat vibrissal system. We recently suggested that these two pathways encode movement frequency in different ways. We proposed that paralemniscal thalamocortical circuits, functioning as phase-locked loops (PLLs), translate temporally coded information into a rate code. Here we focus on the two major trigeminal nuclei of the brain stem, nucleus principalis and subnucleus interpolaris, and on their thalamic targets, the ventral posteromedial nucleus (VPM) and the medial division of the posterior nucleus (POm). This is the first study in which these brain stem and thalamic nuclei were explored together in the same animals and using the same stimuli. We studied both single- and multi-unit activity. We moved the whiskers both mechanically and by air puffs; here we present air-puff-induced movements because they are more similar to natural movements than movements induced by mechanical stimulations. We describe the basic properties of the responses in these brain stem and thalamic nuclei. The responses in both brain stem nuclei were similar; responses to air puffs were mostly tonic and followed the trajectory of whisker movement. The responses in the two thalamic nuclei were similar during low-frequency stimulations or during the first pulses of high-frequency stimulations, exhibiting more phasic responses than those of brain stem neurons. However, with frequencies >2 Hz, VPM and POm responses differed, generating different representations of the stimulus frequency. In the VPM, response amplitudes (instantaneous firing rates) and spike counts (total number of spikes per stimulus cycle) decreased as a function of the frequency. In the POm, latencies increased and spike count decreased as a function of the frequency. Having described the basic response properties in the four nuclei, we then focus on a specific test of our PLL hypothesis for coding in the paralemniscal pathway. We used short-duration air puffs, much shorter than whisker movements during natural whisking. The activity in this situation was consistent with the prediction we made on the basis of the PLL hypothesis.

The role of spike timing in the coding of stimulus location in rat somatosensory cortex

Although the timing of single spikes is known to code for time-varying features of a sensory stimulus, it remains unclear whether time is also exploited in the neuronal coding of the spatial structure of the environment, where nontemporal stimulus features are fundamental. This report demonstrates that, in the whisker representation of rat cortex, precise spike timing of single neurons increases the information transmitted about stimulus location by 44%, compared to that transmitted only by the total number of spikes. Crucial to this code is the timing of the first spike after whisker movement. Complex, single neuron spike patterns play a smaller, synergistic role. Timing permits very few spikes to transmit high quantities of information about a behaviorally significant, spatial stimulus.

Cortical coordination dynamics and cognition

New imaging techniques in cognitive neuroscience have produced a deluge of information correlating cognitive and neural phenomena. Yet our understanding of the inter-relationship between brain and mind remains hampered by the lack of a theoretical language for expressing cognitive functions in neural terms. We propose an approach to understanding operational laws in cognition based on principles of coordination dynamics that are derived from a simple and experimentally verified theoretical model. When applied to the dynamical properties of cortical areas and their coordination, these principles support a mechanism of adaptive inter-area pattern constraint that we postulate underlies cognitive operations generally.

Real-time prediction of hand trajectory by ensembles of cortical neurons in primates

Signals derived from the rat motor cortex can be used for controlling one-dimensional movements of a robot arm. It remains unknown, however, whether real-time processing of cortical signals can be employed to reproduce, in a robotic device, the kind of complex arm movements used by primates to reach objects in space. Here we recorded the simultaneous activity of large populations of neurons, distributed in the premotor, primary motor and posterior parietal cortical areas, as non-human primates performed two distinct motor tasks. Accurate real-time predictions of one- and three-dimensional arm movement trajectories were obtained by applying both linear and nonlinear algorithms to cortical neuronal ensemble activity recorded from each animal. In addition, cortically derived signals were successfully used for real-time control of robotic devices, both locally and through the Internet. These results suggest that long-term control of complex prosthetic robot arm movements can be achieved by simple real-time transformations of neuronal population signals derived from multiple cortical areas in primates.