Statistical methods for network association and discovery in neuroimaging data
Alexandria Jensen, PhD Candidate in Biostatistics University of Colorado, Denver
Applications of quantitative network analysis to neurogimaging data have become popular in the last decade due to their ability to describe the general topological principles of brain connectivity. The brain is the quintessential example of a complex system, whose organization can be seen on a variety of scales, from microscopic interactions between neurons to the global property of lobular division. However, many issues arise when applying standard techniques to represent brain networks or when conducting statistical analysis. Previous studies have shown that brain connectivity is a hierarchy of mesoscopic interactions and these regions tend to communicate in a tree-based structure of homogenously clustered areas. As such, we have proposed a multimodal, hierarchical spinglass algorithm that leverages both functional connections estimated from resting state functional magnetic resonance imaging (fMRI) and white matter pathways estimated from diffusion spectrum imaging. To then analyze these network objects, we propose a novel semiparametric kernel-based regression scheme to better understand the association between these network objects and biological phenotypes, while controlling for potential confounders.