Multi-level characterization and information extraction in directed and node-labeled functional brain networks
Current research in computational neuroscience puts great emphasis on the computation and analysis of the functional connectivity of the brain. The methodological developments presented in this work are concerned with a group-specific comprehensive analysis of networks that represent functional interaction patterns. Four application studies are presented, in which functional brain network samples of different clinical background were analyzed in different ways, using combinations of established approaches and own methodological developments. Study I is concerned with a sample-specific decomposition of the functional brain networks of depressed subjects and healthy controls into small functionally important and recurring subnetworks (motifs) using own developments. Study II investigates whether lithium treatment effects are reflected in the functional brain networks of HIV-positive subjects with diagnosed cognitive impairment. For it, microscopic and macroscopic structural properties were analyzed. Study III explores spatially highly resolved functional brain networks with regard to a functional segmentation given by identified module (community) structure. Also, ground truth networks with known module structure were generated using own methodological developments. They formed the basis of a comprehensive simulation study that quantified module structure quality and preservation in order to evaluate the effects of a novel approach for the identification of connectivity (lsGCI). Study IV tracks the time-evolution of module structure and introduces a newly developed own approach for the determination of edge weight thresholds based on multicriteria optimization. The methodological challenges that underly these different topological analyses, but also the various opportunities to gain an improved understanding of neural information processing among brain areas were highlighted by this work and the presented results.
Use and reproduction:
All rights reserved