DEVELOPMENT AND VALIDATION OF A PROBABILISTIC FINITE ELEMENT MODEL OF THE SPINE

INTRODUCTION: The objective of this effort was to develop and validate with experimental data a probabilistic finite element model (FEM) of the cervical and lumbar spines. The FEM would then be used to predict the probability of injury due to vertical loading. A rigorous hierarchical verification and validation methodology was followed to ensure that there is a high degree of confidence that model predictions reflect human behavior. METHODS: Starting with tissue level models, experimental data was used to calibrate and validate material properties. Next, component level models (i.e., disc, ligaments and vertebrae) were combined to create sub-system models (i.e., two vertebral motion segments) that were validated against experimental results. Finally, the sub-systems were jointed to create full cervical spine and lumbar spine models. As an additional check, a component level validation methodology was developed in which the cumulative distribution function (CDF) from the probabilistic response of the model was compared to the CDF of a separate set of experimental data to calculate a quantitative model performance metric. Validation data included single ligament tension tests, full cervical spine dynamic tests, live human subject tests, and whole cadaver sled tests. RESULTS: The resulting model is capable of not only predicting the mean response but also the expected variation of that response. Furthermore, by applying probabilistic analysis techniques, the model can determine the probability of a specific injury occurring. DISCUSSION: A high fidelity probabilistic FEM of the cervical and lumbar spines was developed using a rigorous hierarchical verification and validation methodology. A large body of experimental data relevant in size and age to the military male and female population was used to validate the FEM at each level of its construction. The resulting model is able to predict the probability of injury during high C vertical loading events.