Development of MicroCT-based Finite Element models for the prediction of the mouse femur mechanical properties

Murine models are widely employed in biomechanics research for different preclinical applications, such as evaluating the effectiveness of bone drugs, studying the effect of genetic mutations on bone mechanical properties, and more. Finite Element (FE) models can be used, after appropriate validation, for the non-destructive assessment of bone mechanical properties, thus reducing the number of animals and refining preclinical experimentation. The aim of this project was to develop and validate FE models of the mouse femur to predict stiffness and failure load in four-point bending configuration. Experimental measurements of femur mechanical properties were conducted at the Mechanical Testing Laboratory of Istituto Ortopedico Rizzoli (IOR, Bologna) and used to evaluate the accuracy of FE predictions. Mouse femurs (N=5) were scanned with micro-Computed Tomography (micro-CT) imaging at 10.76 μm voxel size before four-point bending tests. Micro-CT scans were used to reconstruct the femur geometry. A Gaussian filter was applied to reduce high frequency noise, and subsequently images were segmented using a global threshold calculated with the Otsu method. FE Models were generated by meshing the bone volume with tetrahedral elements with global size of 0.07 mm, and assuming homogeneous isotropic linear elastic material properties. Boundary conditions were applied to replicate the experimental set up. The bone extremities were constrained using a Multi-Point Constraint with two pilot nodes corresponding to the lower supports, to replicate the embedding in polymeric resin; one of the supports was fixed to avoid rigid body motion, while the other was free to move along the femur longitudinal direction. Two areas were selected on the posterior surface of the femur to simulate the contact with the upper supports, and a displacement of 0.2 mm was imposed in the postero-antero direction. Stiffness was calculated as the sum of reaction forces at the lower constraints divided by the displacement at the midshaft. Failure load was estimated using different strain-based failure criteria. A sensitivity analysis was carried out to evaluate the effect of the main model parameters on predicted mechanical properties. Lastly, FE predictions were compared with the experimental data. Stiffness was predicted with average error of 12%. Failure load was highly dependent on failure criterion, with differences up to 69%. Future developments will be focused on the optimization of model parameters as well as increasing sample size. In conclusion, FE models offer a valuable non-invasive approach to estimating the mechanical properties of bone. Nevertheless, these results highlight the importance of validation with respect to the relevant context of use.