Details

Autor(en) / Beteiligte

• Wendelberger, Barbara Ann

Titel

Exploiting biology's structure - function relationship to improve effective connectivity estimates in neuroimaging

Ort / Verlag

ProQuest Dissertations & Theses

Erscheinungsjahr

2016

Quelle

ProQuest Dissertations & Theses A&I

Beschreibungen/Notizen

• Biology consistently demonstrates the correlation between anatomical structure and physiological function. Blood oxygenation levels measured in functional magnetic resonance imaging (fMRI) are used to infer regions of functional activation in the brain’s gray matter, while measured water diffusion in diffusion tensor imaging (DTI) can be used to infer the structural location of myelinated white matter tracts. Effective connectivity modeling in neuroimaging, which estimates directed neural network models, has historically focused almost exclusively on the analysis of fMRI. The well-established association between anatomy and physiology suggests that incorporating structural information into functional data models could improve both the estimation and understanding of neurobiological networks. In neuroimaging, this idea can be tested by combining structural information from DTI with fMRI to investigate effective connectivity estimates using dynamic causal modeling (DCM). DCM incorporates statistical inference with biophysical modeling to estimate directed neural networks from fMRI data using an input-state-output model and a fully Bayesian approach. Default DCM analyses in Matlab utilize Gaussian shrinkage priors in the initialization of the neuronal state equations. A previous study suggests that increasing the variance of these shrinkage priors, based on the probability of an anatomical connection, improves the functional MRI effective connectivity estimates, as determined by Bayesian model selection (BMS). The statistical methods presented here explore the impact of the DCM prior means on neuronal connectivity estimates and investigate the degree to which DCMs might profit from the inclusion of detailed quantitative anatomical connectivity knowledge. Modeling that more accurately represents both the brain’s anatomy and its physiology will improve the estimates of and inferences on brain connectivity networks. Further understanding of network connectivity in the brain paves the way for more effective treatments and therapies aimed at improving patients’ quality of life, particularly in pathological situations.