UCLA

In this thesis, I identified and addressed four problems of network analyses: 1) arbitrary enforcing of network density, 2) network measures are not independent of each-other with three studies, 3) failure to account for structural information in shaping functional networks, and 4) network dynamics -- estimating network change over time. The first chapter introduces the problems and describes current methods within neuroimaging that attempt to address each problem. The second chapter introduces exponential random graph models (ERGM), which is a framework capable of addressing all four problems. The third chapter demonstrates the interaction of problem #2 and problem #3 on the estimation of functional in patients with disorders of consciousness (DOC) and Human Connectome Project (HCP) participants while allowing each patient and HCP to have both their functional and structural connectivity naturally vary (i.e., problem #1). The fourth and fifth chapters focused on the importance of network dynamics for DOC (i.e., problem #4). The fourth chapter describes the use of structural connectivity to investigate the network dynamics of formation and dissolution of connectivity associated with recovery of consciousness over complex behavior in patients with DOC. The fifth chapter describes the use of functional connectivity combined with structural connectivity metrics in a single model to investigate network dynamics in a smaller cohort of 12 patients with DOC. The sixth chapter describes general discussions about the final three chapters and possible future work.