Dokument

Summary:
The recognition and processing of faces is a core competence of our human brain, in which many neuronal areas are involved. Faces are not only a means to recognize and distinguish between individuals, but also a means to convey emotions, intentions, or trustworthiness of our counterpart. The processing of faces is an orchestrated interaction of a multitude of neuronal regions. This interplay can be quantified at the neuronal level using so-called e↵ective connectivity analysis. The most common e↵ective connectivity analysis, which is also used in the present work, is called Dynamic Causal Modeling. With its help, interregional interactions are modelled at the neuronal level, and at the measurable level – such as with functional magnetic resonance imaging – evidence is found for the probability of the presence of neuronal connections and also their quantitative expression. E↵ective connectivity analyses can thus reveal the couplings between brain areas during specific cognitive processes, such as face perception. The way we process faces also changes when, for example, mental illness is present. Thus, negative emotions such as fear may be perceived disproportionately more intense, or positive emotions such as joy less intense. The evaluation of neuronal parameters in face processing could be used in clinical practice, e.g. for the early detection of mental illnesses or the quantification of therapy success. A prerequisite for clinical application is the reliability of the modeling method. Thus, results of models should be generalizable and not depend on certain nuances of the modeling. Furthermore, the interpretability of many model parameters turns out to be dicult. However, this is necessary to be able to describe causal relationships. In the present dissertation, so-called Dynamic Causal Models are applied in the field of neural face processing. In a first study a clinical context is used. Here, neural models of emotion regulation in face processing were used to identify potential consequences of risk factors for the development of mental illness. In another study, the generalizability of neural network models was tested in a healthy population. Here, many limitations of the method as a whole were revealed. In a final study, both observed and simulated data were used to uncover more limitations in the interpretation of model parameters.

Das Dokument ist im Internet frei zugänglich - Hinweise zu den Nutzungsrechten