Osteoporosis is a degenerative bone disease characterized by reduced bone strength, with the primary clinical outcome being fragility fractures. The current clinical diagnosis and management of osteoporosis relies on areal bone mineral density (aBMD), measured using dual-energy X-ray absorptiometry (DXA);​ however, there remains a critical gap between those diagnosed with osteoporosis and those who ultimately fracture. High-resolution peripheral quantitative computed tomography (HR-pQCT) has shown promise in improving fracture stratification through detailed measures of bone microarchitecture. Nonetheless, improvements to date have been modest. The aim of this dissertation was to develop and evaluate new methodologies for assessing bone health, with a particular focus on HR-pQCT. First, the e!ects of di!erent methods for defining the HR-pQCT scan region were investigated, allowing for broader evaluation of data across varying protocols. This was followed by validation of two existing models originally trained on first-generation HR-pQCT data:​ a bone phenotyping model and the microarchitecture fracture risk assessment calculator (µFRAC), which use HR-pQCT-derived parameters as input. Both models demonstrated consistent performance and excellent reproducibility when applied to second-generation HR-pQCT data. The µFRAC model was further validated in external cohorts, including older females scanned using first-generation HR-pQCT and older males scanned using second-generation systems, showing good generalizability and performance in fracture risk prediction. Its utility for longitudinal tracking of fracture risk was also established, demonstrating sensitivity to changes in bone strength over time. A new metric termed void space, designed to quantify localized bone loss, was then implemented in a longitudinal population-based cohort. This analysis revealed sex- and age-specific trends in localized bone loss and strong associations with bone strength and fracture risk. To facilitate clinical translation, an algorithm was developed to quantify void space using standard clinical computed tomography (CT), which was applied to multiple skeletal sites (hip, spine, radius, tibia). Interestingly, void space was found not to correlate across skeletal sites, suggesting that localized bone loss may evolve independently at di!erent anatomical locations. Using the same longitudinal cohort, transitions in bone phenotypes over time were also examined. This analysis revealed distinct sex- and size-dependent trajectories. Finally, a new method for characterizing bone health from HR-pQCT, termed skeletal age, was developed. This model predicts skeletal age directly from unprocessed HR-pQCT images, allowing it to be compared against chronological age as a concise and interpretable summary of bone health for both patients and clinicians. Collectively, these findings validate several tools and models for assessing bone health using HR-pQCT and demonstrate a range of potential use cases for this imaging modality in both research and clinical contexts. Together, they represent an important step toward improving the accessibility, interpretability, and clinical applicability of advanced bone imaging.