New statistical methods to derive functional connectivity from multiple spike trains

Abstract

Analysis of functional connectivity of simultaneously recorded multiple spike trains is one of the major issues in the neuroscience. The progress of the statistical methods to the analysis of functional connectivity of multiple spike trains is relatively slow. In this thesis two statistical techniques are presented to the analysis of functional connectivity of multiple spike trains. The first method is known as the modified correlation grid (MCG). This method is based on the calculation of cross-correlation function of all possible pair-wise spike trains. The second technique is known as the Cox method. This method is based on the modulated renewal process (MRP). The original paper on the application of the Cox method (Borisyuk et al., 1985) to neuroscience data was used to analyse only pairs and triplets of spike trains. This method is further developed in this thesis to support simultaneously recorded of any possible set of multiple spike trains. A probabilistic model is developed to test the Cox method. This probabilistic model is based on the MRP. Due to the common probabilistic basis of the probabilistic model and the Cox method, the probabilistic model is a convenient technique to test the Cox method. A new technique based on a pair-wise analysis of Cox method known as the Cox metric is presented to find the groups of coupled spike trains. Another new technique known as motif analysis is introduced which is useful in identifying interconnections among the spike trains. This technique is based on the triplet-wise analysis of the Cox method. All these methods are applied to several sets of spike trains generated by the Enhanced Leaky and Integrate Fire (ELIF) model. The results suggest that these methods are successful for analysing functional connectivity of simultaneously recorded multiple spike trains. These methods are also applied to an experimental data recorded from cat’s visual cortex. The connection matrix derived from the experimental data by the Cox method is further applied to the graph theoretical methods.