guest ::
| Home > Publications database > Statistical Assessment and Neuronal Composition of Active Synfire Chains |
| Home > Publications database > Statistical Assessment and Neuronal Composition of Active Synfire Chains |
| Poster (After Call) | FZJ-2015-02203 |
; ; ; ; ;
2015
Abstract: The synfire chain (SFC) model has been suggested [1] as a network model for cortical cell assemblies [2]. It is composed of consecutive groups of neurons connected in a feedforward fashion by a large number of convergent and divergent inputs. This connectivity structure enables stable propagation of packets of synchronous spikes after stimulation of the first group [3]. Electrophysiological recordings from hundreds of neurons in parallel increase the chance to detect active SFCs. A method to display repeated SFC activations by means of an intersection matrix was suggested in [4,5]. Each entry in the matrix contains the degree of overlap of identical neurons being active at two different time bins. If a particular SFC is activated twice, a diagonal structure appears in the matrix, composed of consecutive time bins of high intersection values. Further convolution of the matrix with a diagonal linear filter enhances diagonal structures compared to isolated high intersection values. Several features of the data, such as high neuronal firing rates, undersampling of the system or stochastic participation in assembly activation can make the diagonal structures wiggly or discontinuous, or may lower their contrast from individual pixels, thus affecting the analysis performance. Here we introduce a quantitative statistical evaluation of the presence of SFCs which enables an automatic identification of the neurons participating in an SFC. Diagonal structures in the intersection matrix are enhanced by convolving the intersection matrix with a rectangular filter, which enables to cope with wiggly structures. The statistical significance of the diagonal structures is then assessed by filtered surrogate intersection matrices, generated by random permutation of the entries of the original matrix. This implements the null hypothesis that the repeated sets of synchronously firing neurons do not have temporal structure. The entries ofall surrogate matrices form the distribution of matrix entries under the null hypothesis. By choosing an upper quantile we define a lower threshold to identify statistically significant entries in the filtered intersection matrix. Using a clustering algorithm we group the significant entries into single diagonal structures, interpreted as signatures of repeated SFC activation. Finally, we identify the neurons composing the SFC and their group membership. We calibrated the method using stochastic simulations consisting of repeating consecutive synchronous spike patterns embedded in otherwise independent data. Our calibration shows that in most realistic scenarios, e.g. for realistic firing rates and downsampled networks, largeportions (> 90%) of the embedded SFCs are identified at low false positive and false negative levels. We discuss future improvements and possible applications to electrophysiological data. AcknowledgementsHelmholtz Portfolio Theme Supercomputing and Modeling for the Human Brain (SMHB), Human Brain Project (HBP, EU grant 604102), BrainScaleS (EU Grant 269912), DFG Priority Program SPP 1665 (GR 1753/4-1).References[1] Abeles M. (1991) Corticonics, Cambridge University Press, Cambridge[2] Hebb DO. (1949) The organization of behavior. Wiley & Sons, New York[3] Diesmann M. et al (1999) Nature 402: 529-533 doi: 10.1038/990101[4] Schrader S. et al (2008) J Neurophysiol 100: 2165-2176 doi: 10.1152/jn.01245.2007[5] Gerstein GL. et al (2012) J Neurosci Methods 206: 54-64 doi:10.1016/j.jneumeth.2012.02.003
|
The record appears in these collections: |