Influence of bone microstructure distribution on developed mechanical energy for bone remodeling using a statistical reconstruction method

  • Sheidaei A.
  • Kazempour M.
  • Hasanabadi A.
  • Nosouhi F.
  • Pithioux Martine
  • Remond Yves
  • Georges Daniel

  • Finite element model
  • Mechanical behavior
  • Statistical reconstruction
  • Bone microstructure

ART

The development of a predictive model for bone remodeling is becoming increasingly important for medical applications such as bone surgery or bone substitutes like prostheses. However, as bone remodeling is a complex multiphysics phenomenon and difficult to quantify experimentally, predictive numerical models remain, at best, phenomenologically driven. Patient dependency is often ignored as its influence is usually considered secondary, although it is known to play an important role over long periods of time. Another difficulty to study this patient dependency is the availability of experimental samples to carry out extensive analyses. Using our recently developed statistical reconstruction framework, a set of ''bone like'' microstructures with variety of distributions has been created to study pseudo ''patient variabilities. '' The method provides similar effective stiffness tensor, equivalent stresses, and strain energy distributions for the original and the statistically reconstructed samples. The main outcome of this study is the correlation of similar effective mechanical properties between samples when bone remodeling will depend on the local strain energy distribution as a function of each bone microstructure. It is expected that two different micro-structures with equivalent bone volume fraction will lead to identical bone remodeling in a short period of time, whereas this needs to be proven for long term evolution. This work could be used to develop precise predictive numerical models while developing parametric studies on an infinite number of virtual samples and correlating patient dependency with more precise mechanobiological numerical models.