This title appears in the Scientific Report :
2017
Please use the identifier:
http://hdl.handle.net/2128/15260 in citations.
Statistical analysis tools for assessing the functional relevance of higher-order correlations in massively parallel spike trains
Statistical analysis tools for assessing the functional relevance of higher-order correlations in massively parallel spike trains
Understanding how the brain processes information is a major goal in neuroscience. A crucial step towards achieving this goal is to elucidate how information is processed by networks of neurons and how this is expressed in their firing activities. It has been hypothesized that neurons communicate th...
Saved in:
Personal Name(s): | Rostami, Vahid (Corresponding author) |
---|---|
Contributing Institute: |
Jara-Institut Brain structure-function relationships; INM-10 Computational and Systems Neuroscience; IAS-6 Computational and Systems Neuroscience; INM-6 |
Imprint: |
Jülich
Forschungszentrum Jülich GmbH Zentralbibliothek, Verlag
2017
|
Physical Description: |
X, 176 S. |
Dissertation Note: |
RWTH Aachen, Diss., 2017 |
ISBN: |
978-3-95806-251-1 |
Document Type: |
Book Dissertation / PhD Thesis |
Research Program: |
Theory, modelling and simulation Connectivity and Activity |
Series Title: |
Schriften des Forschungszentrums Jülich. Reihe Schlüsseltechnologien / Key Technologies
152 |
Link: |
OpenAccess |
Publikationsportal JuSER |
Understanding how the brain processes information is a major goal in neuroscience. A crucial step towards achieving this goal is to elucidate how information is processed by networks of neurons and how this is expressed in their firing activities. It has been hypothesized that neurons communicate through temporally precisely coordinated spiking, a long-standing hypothesis investigated by several studies. This view has been supported by the insight that a neuron would fire most efficiently if it receives synchronous (in a range of a few milliseconds) spike input. Experimental studies have endorsed this hypothesis and shed light on the functional relevance of synchrony in cortical information processing. However, much of this work has been carried out on handfuls of neurons, thus neglecting features of coordinated activity which emerge only at the level of statistically representative neuron populations while remaining invisible in data sampled from small populations. Over the last few decades electrophysiological techniques have improved significantly, enabling nowadays to record from hundreds of neurons simultaneously. Such advances allow us to study the mechanism behind cortical information processing on a large scale by incorporating the interaction of the constituents of neuronal ensembles. This, however, calls for advanced analysis techniques which can handle increasingly larger numbers of recorded neurons. It is therefore very important to check whether the effectiveness of small-scale data analysis tools also extends to large-scale data. In this thesis we begin with investigating two widely used statistical analysis tools in statistical neuroscience, namely the Unitary Event (UE) analysis and the Maximum Entropy (ME) model. These analysis tools have been designed to assess the presence, magnitude and neuronal composition of correlations in parallel spike train data, and the relevance of these correlations to cortical information processing. We address the emergent features and shortcomings of these analysis tools when applied to massively (i.e. up to hundreds of) parallel spike trains. Then, in line with the statistical frameworks employed by each of these analysis tools, we present two novel statistical analysis tools, i.e. Inhibited Maximum Entropy (IME) and Population Unitary Events (PUE), that can handle the high complexity of large scale data and are able to provide insights about the correlations within an ensemble of neurons. In particular, they aim at segregating the correlations which emerge as a direct consequence of network connectivity, e.g. shared inputs to pairs of neurons, and those with a functional role, e.g. task-dependent modulation of correlations in relation to behavior. In the application of our newly proposed methods to massively parallel spike train recordings from motor cortex of macaque monkey, we verify the insights that they can bring about the importance of correlations (pairwise and higher order) incortical information processing. To summarize, this work proposes two novel statistical tools to study the presence and functional relevance of correlations in massively parallel spike trains. In studying the mechanism behind cortical information processing, these tools shed new light on the distinction between correlations that emerge as a consequence of network connectivity, and those that have a functional role. |