@INPROCEEDINGS{Yegenoglu2018,
author = {Yegenoglu, Alper and Quaglio, Pietro and Torre, Emiliano and Endres, Dominik and Grün, Sonja},
title = {{S}patio-{T}emporal {S}pike {P}attern {R}ecognition in {M}assively {P}arallel {S}pike {T}rains},
reportid = {FZJ-2018-02514},
year = {2018},
abstract = {Introduction: Cortical neurons form a highly interwoven network. Cell assemblies (Hebb, 1949), i.e., interacting groups of neurons, were suggested as the building blocks of information processing in the brain (Singer et al., 1997; Harris, 2005). It is observed that i) spike time coordination at millisecond precision shapes synaptic efficacy (Bi and Poo, 1998), ii) neurons emit a spike more reliably upon synchronous than asynchronous input (Abeles, 1982), and iii) synchronous input may result from pre-synaptic spikes emitted at different times but arriving simultaneously at the post-synaptic site. Modern electrophysiological techniques allow to record hundreds of neurons simultaneously and thereby increasing the chances to observe neurons involved in assemblies expressed by spatio-temporal spike patterns (STPs). Method: We developed a statistical method to detect STPs in massively parallel spikedata (MPST), i.e., on the order of 100 or more neurons. The method is able to deal with the combinatorial explosion of the number of patterns occurring in such high-dimensional data by employing a combination of frequent item set mining (Torre et al.,2013) and a stability analysis algorithm (Kuznetsov, 2007), exploiting the fact that the mathematical foundation of frequent item set mining is equivalent to formal concept analysis (Ganter and Wille, 1999). Our proposed method can statistically assess the patterns and reduce considerably the multiple testing problem by use of Monte-Carlo approaches. As a result the method extracts STPs that occur significantly in excess as compared to STPs that occur by chance. Results: We evaluate our method on ground truth MPST data generated by stochastic simulations. The performance of the method (in terms of false positive and false negative detections) is affected by a variety of parameters, such as the number of STP occurrences, the number of neurons involved in each pattern, and the firing rates of the neurons. Various features of experimental data are considered in analysis, such as non-stationary firing rate in time or inhomogeneity across neurons, and inter-spike interval regularity. We demonstrate the robustness of our method in respect to these parameters by simulating different scenarios which replicate such features. Our results show that the method is suited for the analysis of STPs in massively parallel spike trains thereby offering the possibility to relate such patterns to behavior and show their computational relevance.},
month = {Feb},
date = {2017-02-08},
booktitle = {HBP Student Conference, Austria (Vienna), 8 Feb 2017 - 10 Feb 2017},
cin = {INM-6 / IAS-6 / JARA-BRAIN},
cid = {I:(DE-Juel1)INM-6-20090406 / I:(DE-Juel1)IAS-6-20130828 / $I:(DE-82)080010_20140620$},
pnm = {571 - Connectivity and Activity (POF3-571) / HBP SGA1 - Human Brain Project Specific Grant Agreement 1 (720270) / SPP1665-Ohl(2) - Kausative Mechanismen mesoskopischer Aktivitaetsmuster in der auditorischen Kategorien-Diskrimination (DFG-GF-1753/4-2) / SMHB - Supercomputing and Modelling for the Human Brain (HGF-SMHB-2013-2017)},
pid = {G:(DE-HGF)POF3-571 / G:(EU-Grant)720270 / G:(DE-82)DFG-GF-1753/4-2 / G:(DE-Juel1)HGF-SMHB-2013-2017},
typ = {PUB:(DE-HGF)1},
url = {https://juser.fz-juelich.de/record/845221},
}