Progressive bone loss and weakening bone strength associated with aging predispose the elderly population to osteoporosis and millions of costly fragility fractures. Micro finite element (µFE) analysis based on clinical high-resolution skeletal imaging provides an accurate computational solution to assessing the mechanical properties of bone, which can be used as the dominant factors for fracture risk. However, the current µFE analysis technique is impractical for clinical use due to its prohibitive computational costs, which result from the “voxel-to-element” approach of modeling human bone regardless of its microstructural pattern. I developed a novel plate-rod microstructural modeling technique for highly efficient patient-specific µFE analysis and translated it to clinical research for the assessment of bone strength in osteoporosis and fragility fractures.

Trabecular microstructure is composed of interconnected plate-like and rod-like trabeculae. Instead of converting every image voxel directly into an element, the plate-rod modeling approach created mechanical characterization for every individual trabecular plate and rod. The validation studies demonstrated that the PR model was able to reproduce the morphology and mechanical behavior of the original trabecular microstructure, while reducing the size of the µFE model and improving the efficiency of µFE simulations. First, the PR models of trabecular bone were developed based on high-resolution micro computed tomography (µCT), and evaluated in comparison with computational gold standard-voxel µFE models and experimental gold standardmechanical testing for estimating Young’s modulus and yield strength of human trabecular bone. Results suggested that PR model predictions of the trabecular bone mechanical properties were strongly correlated with voxel models and mechanical testing results. Moreover, the PR models were indistinguishable from the corresponding voxel models constructed from the same images in the prediction of trabecular bone Young’s modulus and yield strength. In addition, PR model nonlinear µFE analyses resulted in over 200-fold reduction in computation time compared with voxel model µFE analyses.

In the effort of studying the heterogeneous bone mineralization in trabecular plates and rods, I developed an individual trabecula mineralization (ITM) analysis technique that allows quantification of the tissue mineral density of each individual trabecular plate and rod. By examining the variation of mineral density with trabecular types and orientations, it was found that trabecular plates were higher mineralized than trabecular rods. Furthermore, trabecular plate mineral density varied with trabecular orientation, increasing from the longitudinal direction to the transverse direction. ITM provided measurement of mineral density of each trabecular plate and rod, which was converted to trabecula-specific tissue modulus and used in the PR models to incorporate mineral heterogeneity in µFE simulations. Results suggested that heterogeneous PR models did not differ from the homogeneous PR models or specimen-specific PR models in their predictions of apparent Young’s modulus and yield strength of the human trabecular bone specimens from non-diseased donors.

Based on the trabecular bone PR model, a whole bone PR model was developed for assessing whole bone mechanical strength at the distal radius and the distal tibia from highresolution peripheral quantitative computed tomography (HR-pQCT). The accuracy of the whole bone PR model was evaluated on human cadaver radius and tibia specimens which were imaged using HR-pQCT and µCT, respectively, and tested to failure. The whole bone stiffness and yield load of the radius and tibia segments predicted by HR-pQCT PR models were strongly correlated with those predicted by corresponding HR-pQCT voxel models, µCT voxel models, and mechanical testing measurements. The PR models µFE results were indistinguishable from the voxel models constructed from the same HR-pQCT images. Moreover, the PR models significantly reduced the computational time for nonlinear µFE assessment of whole bone strength. After evaluating the accuracy and efficiency of the newly developed whole bone PR model, it was employed in a clinical study aimed at characterizing the abnormalities of trabecular plate and rod microstructure, cortical bone, and whole bone mechanical properties in postmenopausal women with vertebral fractures. Women with vertebral fractures had thinner cortical bone, and larger trabecular area compared to their non-fractured peers. ITS analyses suggested vertebral fracture subjects had deteriorated trabecular microstructure, evidenced by fewer trabecular plates, less axially aligned trabeculae and less trabecular connectivity at both radius and tibia. These microstructural deficits translated into reduced whole bone stiffness and yield load at radius and tibia as predicted by PR model nonlinear µFE simulation. More importantly, logistic regression indicated that whole bone yield load was effective in discriminating the vertebral fracture subjects from the non-fractured controls.