Dynamic organization of human brain function and its relevance for psychosis vulnerability

The brain is the substrate of a complex dynamic system providing a remarkably varied range of functionalities, going from simple perception to higher-level cognition. Disturbances in its complex dynamics can cause an equally vast variety of mental disorders. One such brain disorder is schizophrenia, a neurodevelopmental disease characterized by abnormal perception of reality that manifests in symptoms like hallucinations or delusions. Even though the brain is known to be affected in schizophrenia, the exact pathophysiology underlying its developmental course is still mostly unknown. In this thesis, we develop and apply methods to look into ongoing brain function measured through magnetic resonance imaging (MRI) and evaluate the potential of these approaches for improving our understanding of psychosis vulnerability and schizophrenia. We focus on patients with chromosome 22q11.2 deletion syndrome (22q11DS), a genetic disorder that comes with a 30fold increased risk for developing schizophrenia, thus enabling the study of developmental alterations and risk factors that precede the onset of full-blown psychosis. We first examine temporal variability of the blood oxygenation level dependent (BOLD) signal, a voxelwise measure of dynamic fluctuations. As static functional connectivity (sFC) is scaled for variance, we also compare BOLD signal variability with altered sFC. The broad pattern of altered BOLD signal variability in 22q11DS partly, but not entirely, overlaps with altered sFC, suggesting a complex non-linear relationship between the two measures. Further, testing for altered BOLD signal variability in patients with higher psychotic symptoms, we find reduced values in the dorsal anterior cingulate cortex (dACC), a central node of the salience network (SN). Going beyond a voxel-wise measure of brain dynamics, we next look into aberrant dynamics of large-scale functional brain networks. We use innovation-driven co-activation patterns (iCAPs), an approach that stands out by its ability to recover spatial and temporal overlap of functional brain networks. As in the original iCAPs framework such spatial overlap can introduce spurious temporal activations, we propose a novel spatio-temporal regression framework that relies on transient-based constraints to overcome this limitation. With this improved scheme, we probe into clinical risk factors of psychosis in 22q11DS and find aberrant activation and coupling of functional brain networks, again implicating the SN, as well as the amygdala and hippocampus. Finally, we explore the implications of structural network topology on functional dynamics, by combining iCAPs with network control theory, an approach that relies on a dynamic model to predict the energy required for controlling the brain's state given its structural connectivity. We find that the brain operates in an energetically optimal way, spending less time in brain states that require higher control energy. In patients with 22q11DS, this relationship is less pronounced, suggesting less efficient functional brain dynamics in these patients. In summary, our results confirm that the dynamic nature of brain function contains essential information and warrants further attention for the development of clinically relevant imaging markers for psychosis vulnerability. Moreover, we provide initial evidence for aberrant relationship between brain function and structure in patients with 22q11DS that merits further exploration.