Efficient Large-Scale Informatics for Diffusion Magnetic Resonance Imaging in Alzheimer’s Disease

Abstract

Alzheimer's disease (AD) is the leading cause of dementia worldwide, with white matter (WM) phenotypes being biomarkers of cognitive decline. The canonical datatype to quantify WM in the brain in vivo is diffusion magnetic resonance imaging (dMRI); however, derived dMRI data often require many image processing steps before reaching the phenotypes used in AD studies. Furthermore, formats, acquisitions, processing, and quality control of these derived and raw dMRI data are heterogenous and can introduce unwanted variability in downstream analyses. For large-scale studies aiming to detect smaller biological effects, these informatics considerations become nuisance issues when scaling up sample sizes that need to be accounted for. This dissertation has two main contributions: 1) development of a scalable informatics framework that standardizes data formats, processing, and quality control of heterogenous dMRI data, and 2) using this informatics foundation to characterize and perform large population analyses of WM that facilitate future large-scale studies of AD. Together, these contributions provide a reproducible framework for transforming heterogeneous dMRI data into reliable WM phenotypes, enabling better-powered analyses of WM changes in aging and AD.