Reliability of multiscale models of bone is related to the accuracy of the experimental information available on bone microstructure. X-ray-based imaging techniques allow to inspect bone structure and mineralization in vitro at the micrometre scale. However, spatial resolution achievable in vivo is much coarser and can produce blurry, uncertain information on bone microstructure. Working with uncertain data calls for new modelling paradigms able to propagate uncertainty through the scales. In this paper we investigate the effects of uncertain bone mineralization on the elastic coefficients of the bone matrix. To this aim, some stochastic concepts were developed and compared with one another in order to identify the best way to account for uncertain input data. These concepts step from a deterministic micromechanical model of bone matrix which was extended in order to account for uncertain bone composition. Uncertainty was introduced by assuming to know only mean value and dispersion of the parameters describing bone composition. Thus, these parameters were modelled as random variables and their distribution functions were obtained using the maximum entropy principle. Either the tissue mineral density (TMD) or the ensuing volume fractions of collagen and mineral were used to describe uncertain bone composition. Moreover, mean value and dispersion were estimated at the scales of either 10 or a few 100 [Formula: see text]m, representative of standard in vitro and in vivo spatial resolutions, respectively. Analysis of these modelling concepts suggests that TMD measured at the sub-millimetre scale can be used to obtain reliable statistical information about the elastic coefficients of bone matrix.
Keywords: Bone matrix; CT; Continuum micromechanics; Elastic properties; Homogenization; MaxEnt principle; Stochastic modelling; Uncertainty quantification.