Over the last years, the research focus in neuroscience has shifted from functional localization to functional connectionism. In the context of the questions of how the neurons
are connected and what the underlying dynamics are, effective connectivity (causal interconnections) becomes more and more important.
Pernice et al. (2013) published a method to infer connectivity between neurons from covariances of their activities. Applied to brain recordings, like electrocorticography (ECoG), one hence could gain insight into the connections beneath the electrode grid. To validate, test and investigate its sensitivity towards specific aspects, like recording length, both the ECoG data and the anatomical connections have to be available. This is often not the case.
Based on the experimental setup we built a computational model using NEural Simulation Tool (NEST)(Gewaltig and Diesmann, 2007). With this network of networks, we combined two levels of connectivity: local (like beneath an electrode) and global (between electrodes). Thereby we aimed for mimicking ECoG recordings. On the local scale we dealt with balanced networks, whereas the global connectivity was assumed to be distance-dependent. The parameter settings were adjusted on the basis of experimental results. The process of modelling, as well as the model itself, were expected to give insight into the ECoG signal, by addressing questions like which biophysical aspects contribute to the recordings and what features does the model ought to have to show the same temporal or spatial correlation pattern seen in experiments.
By performing simulations with various artificial global network structures imposed, instead of the distance-dependency, findings confirmed that this model can be used to test the reconstruction of connectivity from covariances.
The talk is open to the public. Guests are cordially invited!