The human mirror neuron system - Effective connectivity and computational models

In this thesis, I attempt to gain a deeper insight into the details of the human mirror neuron system by finding the effective connectivity of its central regions and simulating them with computational modeling. To achieve this aim, I have used the measured functional magnetic resonance imaging (fMRI) data for healthy participants while performing key tasks of social cognition (imitation of emotional faces, theory of mind, and empathy). Using a self-developed firing-rate-based extension of the statistical analysis procedure dynamic causal modeling (DCM), I was able to determine the effective network structure of the human mirror neuron system from the fMRI data and compare it between the different tasks of social cognition. In particular, far more complex processing occurs in imitation than in the other two tasks, which seems plausible given that imitation involves matching observed and self-performed emotional expression. Furthermore, we were able to show that the extended DCM procedure allows for significantly better model evidence, both for our novel data and for previously established datasets from other research groups. Thus, in addition to the substantive insight, this project has provided an important methodological advance for all users of the widely used DCM procedure. Furthermore, the main regions of the mirror neuron system are modeled in detail by a modification of an existing, completely data-driven spiking network model of the prefrontal cortex. Here, I use the estimated parameters of the modified DCM to match the time series of the simulated and observed data. This two-stage approach allows both to account for the neural mass signals measured by fMRI and assess the fine-scale temporal dynamics of the local dynamics, and thus derive predictions about the physiological details that cannot be obtained from non-invasive recordings alone