The Influence of Structural Changes and Population Interactions on the Entropy Based Synchronicity

Abstract

Event Abstract Back to Event The Influence of Structural Changes and Population Interactions on the Entropy Based Synchronicity Fikret E. Kapucu1*, Inkeri Vornanen1, Jarno M. A. Tanskanen1, Francois Christophe2 and Jari Hyttinen1 1 Tampere University of Technology, Computational Biophysics and Imaging Group, Finland 2 Tampere University of Technology, Department of Pervasive Computing, Finland Motivation Neuronal synchrony is generally defined as the simultaneous activity of neuronal cells or cell assemblies. Most common way to analyze simultaneous activity is to assess simultaneous activities of spikes [1,2], or temporal occurrence of bursts [3,4]. These methods are widely employed especially for in vitro neuronal networks whereas other methods such as frequency/phase coupling between signals [5] are also employed for in vivo synchrony analysis, e.g. from electroencephalograms (EEG). A reason for that is because brain oscillation frequency bands are well-defined in EEG studies as opposed to developing neuronal cultures. However, for in vitro neuronal signals, spike and burst information is not always adequate due to the challenges in recognition of spikes and bursts in recordings, for such cases analyzing synchrony based on temporal activity is not preferable. This issue motivates the method presented in this paper which evaluates synchrony by means of temporal correlations of spectral complexity for neuronal recordings. In this work, we investigate the influence of structural changes in neuronal networks on this suggested synchronicity method. Material and Methods We performed initial tests with different neuronal interactions and network structures based on the computational model introduced by Tsodyks et al [6]. The parameters for neurons and synapses are the same as in [6]. The network was driven with Gaussian current. Three minutes spike trains of the neurons were simulated using the NEST simulator [7] and population activity is formed by accumulating spike time points of each neuron. Then a Sinc kernel is convolved to obtain a continuous function representing the population spike firing continuously. Next we calculated spectral entropy (SE) as described in [8] based on previous work [9]. After SE is calculated for every time window, we obtain time-variant SE(w), w∈[1 N], where N is the total number of windows where SE is calculated as in Fig.1A. The equation in Fig.1A calculates correlation between pairwise time-variant SEs, SE-sub-X and SE-sub-Y, where SE-sub-X bar and SE-sub-Y bar are the sample means of the corresponding SEs and sigma is the standard deviation. Correlations by means of time-variant entropies are calculated for different network parameters, such as direct vs. indirect connections, different levels of connectivity, number of interacted populations. Results An exemplary result is presented in Fig. 1 for interrelations of three neuronal populations with different connectivity levels. The influence of indirect connection (between 1 and 3) on the measured synchrony for the corresponding levels of connectivities (50% and 10%) can be seen in Figure 1B. Discussion We obtained results from 100 simulations for each tested neuronal network structure and connectivity level. Consequently we are able to validate the relations of neuronal interactions statistically. On the other hand since natural neuronal networks are more complex than the ones defined here, further studies including natural neuronal networks are necessary for the evaluation of the method’s feasibility with MEA recordings. Conclusion Method is able to detect different structural changes and connectivity levels. Future studies with more parameters and tests with natural neuronal cultures would be beneficial for method’s development for practical usability.