Personalized finite element modelling of the proximal femur for fracture risk prediction

Because of the population ageing, osteoporosis and hip fractures are largely considered as a major health-care problem. The surgery, combined with the patients rehabilitation and their dependency management, involve a worrying socio-economical burden which must be reduced, but the high cost of any preventive strategy must also be taken into account. Bone mineral density measured by DXA is considered as the gold standard for fracture risk prediction, but this diagnostic tool is limited by the overlap that still exists between bone density values of fractured and non-fractured subjects. The purpose of this study was to develop Finite Element models in order to improve the hip fracture risk prediction. An anatomical subject-specific model that take into account the complex geometric and mechanical specificities of each specimen reached a good compromise between an accurate individual fracture load prediction (Fexp = 1,006 FMEF avec r = 0,87, IC95% = +- 2600 N) and a suitable numerical cost for further clinical use (numerical simulation 30mn). To improve the model personalisation, an experimental study on 92 bone samples (46 traction and 46 compression) led to a better prediction of the cortical mechanical properties from medical imaging (r = 0,42 a 0,72). A total of 40 pairs of human proximal femurs were also tested to failure in order to provide a robust database for the model validation. A parameterized model was simultaneously used in an experimental design study in order to understand the role of the geometrical parameters in the femoral strength. This work indicated the ability of the Finite Element modelling to become a complementary tool for fracture risk prediction, and underlined the major influence of the cortical bone and its mechanical strength.