Elsevier

Neurocomputing

Volume 70, Issues 10–12, June 2007, Pages 2112-2116
Neurocomputing

Spatially organized spike correlation in cat visual cortex

https://doi.org/10.1016/j.neucom.2006.10.141Get rights and content

Abstract

Inspired by optical recordings from visual cortex which show maps of orientation selectivity, and the finding that very similar patterns of population activity occur when the neurons fire spontaneously [T. Kenet, D. Bibitchkov, M. Tsodyks, A. Grinvald, A. Arieli, Spontaneously emerging cortical representations of visual attributes, Nature 425 (2003) 954–956], we approach the question of how the concept of cortical maps may be related to the concept of temporal coding. To this end we analyzed parallel spike recordings performed using a 10×10 electrode grid covering an area of 3.6mm×3.6mm of cat visual cortex for occurrence of spike correlation.

We calculated all possible pairwise correlations between multi-unit activities (MUA) by cross-correlation and extracted significantly correlated pairs using a boot-strap procedure. The MUAs involved in correlated pairs were typically involved in more than a single correlated pair. Using methods of graph theory we found that the whole set of correlated MUAs decomposes into a small number of groups of MUAs that have a high degree of the overlap of mutually correlated pairs.

Mapping these groups back onto the spatial arrangement of the recording electrodes revealed that these also correspond to spatially segregated clusters. The spatial scale of this correlation map is in agreement with the scale of orientation tuning maps found by optical imaging.

Introduction

Most cells in the primary visual cortex are selective for orientation and respond best, i.e. by increasing their rates, to their optimal stimulus. Neurons located in different columns but with same stimulus preference coherently show high activity during the same, optimal stimulus, as indirectly shown by optical imaging. Moreover, it has been shown that the firing rate of a spontaneously active single neuron strongly depends on the instantaneous pattern of the ongoing population activity in a larger cortical area [10]. Very similar patterns of population activity were observed, both when the neuron fired spontaneously and when driven by its optimal stimulus [5]. On the other hand, the temporal coding hypothesis suggests that coordinated spiking activity on a fine temporal scale should occur in the nervous system [6] and is suggested to be used for information coding. Accordingly, modern theories of the primary visual pathway contain spatial and temporal aspects accounting for the observed cell behavior in order to explain visual information processing.

We therefore approach the question of how the concept of temporal coding of single neurons are related to the overall functional architecture. More specifically, we ask if correlated neurons are arranged in a specific spatial organization that may be related to maps of orientation tuning. Parallel spike recordings based on using a 10×10 electrode grid (Utah electrode array, Bionic Technologies, Inc., Salt Lake City, UT, USA) covering an area of 3.6mm×3.6mm of cat visual cortex [11] allow us to address the question of the relation of correlated neuronal activity to the distance and the spatial arrangement of the recording sites. Data were recorded from area 17 of anesthetized cat during spontaneous activity, and under full-flash treatment with two different stimulus intensities. We first analyze the simultaneous multi-unit recordings for pairwise correlations using cross-correlation analysis and evaluate their significance using boot-strap techniques. In a next step, we identify groups of correlated pairs that are highly mutually intra-correlated. Finally, mapping these groups back onto the electrode positions and thereby to cortical space, allows us to relate their spatial arrangement to the spatial scales found for orientation tuning maps.

Section snippets

Detection of correlated spiking activity

We analyzed the spiking activities that were recorded from the grid of electrodes (Fig. 1) for pairwise correlations, during spontaneous activity (SP; no stimulus) and during full-flash stimulation with two different intensities. We segmented epochs of high intensity (HI) and low intensity (LI) into two separate data sets and analyzed them separately. For simplicity we restricted ourselves to the evaluation of the multi-unit activities (MUA), and requested a minimal firing rate of 1Hz for

Discussion

This study investigated the spatial organization of correlated activity in cat visual cortex. Multi-unit activities were recorded by a grid spanning an area of 3.6×3.6mm2 allowing us to investigate the relation of correlated MUA pairs to spatial structure. On average 5% of all possible pairs were significantly correlated. To capture the correlation structure, we constructed a graph with each MUA being a node and assigning edges to significantly correlated MUAs. This led to the finding that the

Acknowledgments

Partial funding by the BMBF (BCCN Berlin, grant 01GQ0413), the Stifterverband für die Deutsche Wissenschaft, and the Volkswagen Foundation. This work was carried out while Sonja Grün was based at the Freie Universität in Berlin, Germany.

PD Dr. Sonja Grün was born in 1960 in Germany, where she obtained her Diploma in Physics (Eberhard-Karls University Tübingen). She did her Ph.D. work in the field of computational neuroscience at the Ruhr-University Bochum, Germany, and at the Weizmann Institute of Science, Rehovot, Israel, and obtained her Ph.D. in physics (Ruhr-University Bochum). After her post-doctoral work at the Hebrew University in Jerusalem, Israel, she worked as a senior fellow at the Max-Planck Institute for Brain

References (12)

There are more references available in the full text version of this article.

Cited by (25)

  • Modeling and analyzing higher-order correlations in non-Poissonian spike trains

    2012, Journal of Neuroscience Methods
    Citation Excerpt :

    Fig. 4, right column, shows such an example. The population consists of four cell assemblies where neuronal spiking is correlated within each group, but not across groups (e.g. Berger et al., 2007). More precisely, four populations of lognormal processes (N = 30, 20, 15 and 35 neurons) have been independently generated with fixed firing rate (λ = 1.5, 2.5, 12 and 0.5Hz), coefficient of variation (CV = 1.5, 2.5, 2 and 1.2), pairwise coincidence probability (cp = 0.03, 0.08, 0.12 and 0.005) and mean order of synchronized activity (ζ = 2, 6, 7 and 3), resulting in a population that is heterogeneous in several respects.

  • A new method for evaluating the complexity of animal behavioral patterns based on the notion of Kolmogorov complexity, with ants' hunting behavior as an example

    2012, Neurocomputing
    Citation Excerpt :

    Within populations behavioral variability serves as a basis for behavioral, cognitive and social types of specialization which facilitate tuning of integrative reactions of an integral animal community to unpredictable influences in its changeable environment [16]. At the neurological level evaluation of variability and complexity of digitizing “behavior” of neurons within assemblies should provide a better understanding of the perceptive and cognitive processing in the brain [2,6,23]. The concept of complexity of animal behavior is still mainly intuitive.

  • An information integration model of the primary visual cortex under grating stimulations

    2011, Biochemical and Biophysical Research Communications
    Citation Excerpt :

    Berger et al. analyzed spiking activities in cats’ V1 under stimulus of different luminance using methods of graph theory. They found that the whole set of correlated multi-unit activity (MUA) is decomposed into a small number of groups of MUAs which have a high degree of overlap of mutually correlated pairs and the spatial scale of this correlation map is in agreement with the scale of orientation tuning maps set up by imaging [13]. Gray et al. calculated the correlation between recorded neurons and found the oscillatory responses in synchrony for cells, which indicates that synchronized oscillation is one of the ways the neurons interacted with each other through horizontal connections [14].

  • Statistical technique for analysing functional connectivity of multiple spike trains

    2011, Journal of Neuroscience Methods
    Citation Excerpt :

    For example, a procedure of using some appropriate surrogate data looks like a promising improvement of pair-wise correlation based method (see, Louis et al., 2010a,b). This method of surrogate data has been successfully applied to analyse data of multi-electrode recordings from the visual cortex of a cat (Berger et al., 2007, 2010) and, in particular, to decide on participation of a neuron in synchronous population activity. There are several recent techniques which can analyse multiple spike trains at once.

  • Complexity distribution as a measure for assembly size and temporal precision

    2010, Neural Networks
    Citation Excerpt :

    When faced with this problem, one option in the analysis of massively parallel spike trains is to limit oneself to pairwise analysis. In that way functional structures can indeed be extracted from the population (Berger, Borgelt, Louis, Morrison, & Grün, 2010; Berger, Warren, Normann, Arieli, & Grün, 2007; Eldawlatly, Jin, & Oweiss, 2009). It has been suggested in log-linear modeling studies, that taking into account only pairwise interactions is sufficient to faithfully characterize neural population activity (Schneidman, Berry, Segev, & Bialek, 2006; Shlens et al., 2006).

View all citing articles on Scopus

PD Dr. Sonja Grün was born in 1960 in Germany, where she obtained her Diploma in Physics (Eberhard-Karls University Tübingen). She did her Ph.D. work in the field of computational neuroscience at the Ruhr-University Bochum, Germany, and at the Weizmann Institute of Science, Rehovot, Israel, and obtained her Ph.D. in physics (Ruhr-University Bochum). After her post-doctoral work at the Hebrew University in Jerusalem, Israel, she worked as a senior fellow at the Max-Planck Institute for Brain Research in Frankfurt/M, Germany. From 2002 to 2006 she was an Assistant Professor for Neuroinformatics/Theoretical Neuroscience at the Free University in Berlin, Germany and was a founding member of the Bernstein Center for Computational Neuroscience in Berlin. Since 9/2006 she is the head of a research unit at the RIKEN Brain Science Institute in Wako, Japan. Her main interests are in statistical neuroscience, which includes modeling of stochastic processes and the development of data analysis techniques for multiple parallel neuronal time series.

Denise Berger was born in 1980 in Germany. In 2005, she received her M.Sc. in bioinformatics from the Free University in Berlin, Germany. Currently, she is working on her Ph.D. thesis in the field of Computational Neuroscience at the Free University in Berlin, and is a member of the Bernstein Center for Computational Neuroscience in Berlin (BCCN). Her current interests are in statistical neuroscience and the analysis of massively parallel electrophysiological data.

View full text