|
Abstract:
|
Recently, the focus of brain imaging studies has shifted from region based activation analysis to a connectome based interpretation. It is now widely believed that different areas of the brain exhibit functional connections which may differ with clinical, demographic, environmental factors, as well as the brain structure. We investigate the influence of the brain anatomical structure on brain functional connectivity. We fuse Diffusion Tensor Imaging (DTI) knowledge with functional magnetic resonance (fMRI) measurements to estimate the brain functional network via a novel Bayesian, covariate adjusted Gaussian graphical model and an associated EM algorithm. The approach allows structurally connected regions to inform functional connections while also allowing additional functional connections supported by the data. Through simulations, we show that the approach has higher power to detect functional connections in comparison to unstructured network approaches not accounting for supplementary brain anatomical knowledge. We apply our approach to the Philadelphia Neurological Cohort study, which provides DTI and fMRI data on children and young adults of varying cognitive abilities.
|
