Prediction of the Vertebral Strength Using a Finite Element Model Derived From Low-Dose Biplanar Imaging

Abstract

A finite element analysis on osteoporotic vertebrae. This study aims to validate subject-specific finite element models (FEMs) derived from a low-dose imaging system (EOS, Biospace Med, France) for the prediction of vertebral strength. The vertebrae are submitted to an eccentric compression force leading to compression and anterior bending. Given the aging population, osteoporosis and vertebral fractures are a major public health issue. A low bone mineral density (BMD) does not always explain incident fractures, and multifactorial analyses are required. In this context, FEMs based on quantitative computed tomography (QCT) have been proposed to predict vertebral strength in vitro or quantify effects of treatments. However, the clinical use of such a model for the in vivo follow-up of the whole spine is limited by the high-radiation dose induced by QCT and the lying position, which does not allow postural assessment with the same modality. Fourteen vertebrae were modeled using a parametric meshing method. The mesh was subject-specific using geometric parameters computed on the 3-dimensional (3D) reconstructions obtained from the EOS biplanar radiographs. The contribution of cortical bone was taken into account by modeling a cortico-cancellous shell whose properties were derived from experimental data. The effect of subject-specific bone Young's moduli derived from EOS vertebral areal BMD was quantified. The 3D position of the point-of-load application and the 3D orientation of the force was faithfully reproduced in the model to compare the predicted strength and experimental strength under the same loading conditions. The relative error of prediction decreased from 43% to 16% (2.5 times) when subject-specific mechanical properties, derived from EOS areal BMD, were implemented in the FEM compared with averaged material properties. The resulting subject-specific FEMs predicted vertebral strength with a level of significance close to the QCT-based models (r adjusted = 0.79, root mean square error = 367 N). This work underlines the potential of low-dose biplanar x-ray devices to make subject-specific FEMs for prediction of vertebral strength.