English
 
User Manual Privacy Policy Disclaimer Contact us
  Advanced SearchBrowse

Item

ITEM ACTIONSEXPORT

Released

Meeting Abstract

Unsupervised identification of neural events in local field potentials

MPS-Authors
/persons/resource/persons75278

Besserve,  Michel
Department Physiology of Cognitive Processes, Max Planck Institute for Biological Cybernetics, Max Planck Society;
Max Planck Institute for Biological Cybernetics, Max Planck Society;

/persons/resource/persons84193

Schölkopf,  Bernhard
Dept. Empirical Inference, Max Planck Institute for Intelligent Systems, Max Planck Society;

/persons/resource/persons84063

Logothetis,  Nikos K
Max Planck Institute for Biological Cybernetics, Max Planck Society;
Department Physiology of Cognitive Processes, Max Planck Institute for Biological Cybernetics, Max Planck Society;

Fulltext (public)
There are no public fulltexts stored in PuRe
Supplementary Material (public)
There is no public supplementary material available
Citation

Besserve, M., Schölkopf, B., & Logothetis, N. K. (2014). Unsupervised identification of neural events in local field potentials. In 44th Annual Meeting of the Society for Neuroscience (Neuroscience 2014).


Cite as: http://hdl.handle.net/21.11116/0000-0001-31FC-C
Abstract
Local Field Potentials (LFP) recordings carry information about a large variety of dynamical network mechanisms occurring at multiple scales. While these mechanisms are hypothesized to be instrumental to information processing, identifying them without prior information is challenging. A standard approach to achieve this goal is to extract information from specific frequency bands reported in the literature (such as Theta or Gamma bands). However, the variability of these bands across species, brain structures and individuals is a major difficulty. We propose an unsupervised technique to automatically identify and detect relevant dynamical events. The methodology is based on a Non-negative Matrix Factorization (NMF) of the time varying LFP spectrum that decomposes the signals into a small number of dynamical components with specific spectral signatures. Large transient events in each component are further detected with a statistical test assuming a Gaussian null distribution of time course. We applied this methodology on LFPs recorded from the CA1 subdivision of hippocampus in 3 anesthetized macaques, totalizing 12 recording sessions. In each session, we quantified the stability of the factorization by computing the correlation between the spectral signatures obtained different blocks of LFP data. Setting a threshold on the minimum average correlation to .8, we concluded that one could extract on average 6 stable dynamical components from these signals. The obtained spectral components, clustered across sessions using a graph clustering algorithm, are represented on Figure 1 (left panel), as well as example time courses of components in one session (Figure 1 right panel). The dark rectangles indicate detected dynamical events in each component, among which classical hippocampal sharp-wave and ripple events (first and second component from the top) are well isolated. In sum, our approach offers a principled way to isolate key dynamical events in LFP data without prior frequency band definition and can be applied in a wide range of experimental settings and brain structures.