Integrating micro CT indices, CT imaging and computational modelling to assess the mechanical performance of fluoride treated bone

Abstract

In this study we evaluate the influence of low-dose fluoride treatment on 23 patient biopsies. Computational finite element (FE) models of each biopsy were subjected to a range of loads including compression, shear and torsion. The modelling framework was validated against three 3D printed models with known material properties subjected to compression till failure using an Instron machine. The primary outcomes from this study were that mechanical strength was not significantly correlated to low-dose (<10mg/day) of fluoride levels (one-way ANOVA, P-values of 0.78, 0.69 and 0.62 for compression, shear and torsion, respectively). However, when bulk bone material properties were derived from DXA bone mineral density (BMD) from each patient's proximal femur a non-significant linear decline in mechanical strength with increase in fluoride was predicted. When the same material property was used for all bones (to evaluate bone architecture influence) then mechanical strength showed a characteristic concave upwards trend, consistent with the variation of micro CT derived percentage bone volume (BV/TV). The secondary outcomes from this study were that in compression, BV/TV was observed to be a strong surrogate measure for mechanical strength (R2=0.83), while bone surface density (R2=0.6), trabecular thickness (R2=0.5) and intersection surface (R2=0.6) also explained the variation of mechanical strength well. However, trabecular separation and trabecular number were mildly correlated with mechanical strength (R2 of 0.31 and 0.35, respectively). Compression was the loading mode most strongly correlated to micro CT indices. Material properties adapted from the proximal femur reduced the CT index correlations by up to 58% indicating that bulk density from a near proximity is a poor representation of specific localised density. Substituting the 3D micro CT indices with 2D histomorphometric data decreased correlations by at least 33% indicating that structural identification on a plane is not representative of the full 3D architecture necessary for a complete bone strength analysis. The presented computational framework may be used to assess the roles that bone architecture and loading modes play in bone quality, and which micro CT indices are good surrogate measures for mechanical strength.