Knee osteoarthritis (OA) is a prolific and devastating disease. The processes leading to the development of OA are poorly understood, but periarticular bone adaptations are theorized to play an important role. High-resolution peripheral quantitative computed tomography (HR-pQCT) can resolve human bone microarchitecture in vivo, and is uniquely suited to investigate these hypotheses. This thesis develops automated methods for quantifying bone microarchitecture at the knee with HR-pQCT. First, a deep learning workflow is developed for segmenting bone compartments in HR-pQCT images of the distal radius and tibia using data from a large normative cohort, achieving excellent accuracy and equal or better precision than the established semi-automated workflow. Building on this, a deep learning algorithm for segmenting subchondral and trabecular bone in HR-pQCT images of the knee is developed using similar methods and an atlas-based segmentation workflow is developed for automatic contact surface identification, again achieving excellent accuracy and improved precision compared to the established workflow. Separately, a large dataset is leveraged to analyze the statistical and geometric interdependence of four parameters that quantify trabecular microarchitecture, and evidence is found that these four parameters represent only two true degrees of freedom, with important implications for inferential testing and study design. Next, an alternative, model-based method for measuring cortical thickness and subchondral bone plate thickness on HR-pQCT is developed by reformulating Cortical Bone Mapping as a global optimization with spatial regularization of fitted parameters. Finally, methods developed herein are applied to study the one-year changes in periarticular bone microarchitecture in the knee following both ACL injury and surgery. The main effects for the injured side are consistent with theory, but the contrasts for OA risk factors are not significant while contrasts for baseline microarchitectural factors are. This evidence suggests that the effects of OA risk factors on OA pathogenesis may be mediated through the state of the microarchitecture at the onset of OA development, rather than directly impacting tissue changes that lead to OA from a common baseline state.