Computational Analysis of Bone Fracture

Abstract

Age-related non-traumatic fractures are a major public health problem with fracture of the proximal femur resulting in significant patient mortality. Currently, the clinical gold standard for assessing an individual’s fracture risk is dual-energy x-ray absorptiometry (DEXA). Although DXA-based bone mineral density (BMD) measurements have been shown to correlate with fracture risk, BMD distributions describing normal individuals and those who have suffered a hip fracture contain significant overlap, thereby reducing the specificity of BMD based fracture risk categorization. Since bone strength results from a combination of bone mass, bone micro-architecture and material properties, and overall bone geometry, two-dimensional imaging based diagnostic approaches such as DEXA are limited in their ability to specifically predict bone strength. Engineering models of skeletal structures that combine descriptions of three dimensional bone geometry, bone mass distribution, and bone material behavior into a high fidelity simulation have been shown to predict bone strength with an improved accuracy compared to DEXA alone, albeit with some limitations. Specifically, voxel based finite element models derived from QCT image data are based on correlations between predicted structural stiffness and structural failure and therefore are generally only valid for simple uniaxial loading; it is unlikely this approach would be valid for more complex applied loads. Furthermore, these models generally use simplified bone material descriptions that do not capture the demonstrated non-linear damaging behavior of bone and do not model bone material failure. Another major limitation is that these models require three dimensional QCT image data, which can be expensive and limited in its availability, and they are not parametric – they are specific to an individual. Finally, all models used to date are deterministic and do not account for uncertainty and/or variability in bone structure and/or material properties, and cannot be used to predict the probability of bone failure. Each of these limitations can be addressed using advanced computational methods such as statistical shape and bone density modeling, non-linear continuum damage mechanics modeling, and probabilistic analysis methods with the goal of improving the identification of individuals who are at a greater risk of fracture and focusing resources on those areas where treatment can be directed to significantly improve an individual’s fracture risk. Ultimately, this approach will significantly improve the ability to clinically quantify the risk of fracture in an individual and allow treatment to be administered in a timely manner.