The master thesis deals with the research of intercellular couplings of
neuronal cell cultures. For brain research it is important to understand
causal relationships between neuron as well as the development of neural
networks. The transfer entropy is a suitable model-free tool for analysing
causal relationships. It is an information-theoretic measure quantifying
nonlinear directed interactions of two dynamic processes. The master thesis
is aimed to show how causally determined interactions between two time
series can be analyzed with the transfer entropy. Moreover delayed
source-target interactions were taken into account. Furthermore it was
investigated whether it is possible to form neural networks by using the
established causal linkages. The transfer entropy was calculated for binary
as well as for discrete time series in dependence on the interaction-delay.
It was computed for simulated neuronal time series and data from cultured
cardiomyocells obtained from an 8 x 8-multielectrode array. After all a
cluster analysis was used to select calculated transfer entropy results in
dependence on the interaction-delay and to determine potential neuronal
networks. As a result it is possible to identify casual relationships
between neuronal time series. Due to the cluster analysis it was possible
to determine networks out of the selected transfer entropy results. All in
all the findings show that the interaction-delay as a parameter for
transfer entropy calculation has a significant influence on the obtained
results and consequently has to be chosen carefully. In contrast to
transfer entropy of discrete time series, transfer entropy calculation for
binary time series is remarkable less computationally intensive.