Application of Neural Network and Finite Element Method for Multiscale Prediction of Bone Fatigue Crack Growth in Cancellous Bone - Multiscale Computer Modeling in Biomechanics and Biomedical Engineering - page 25

sites on radiographs of 18 specimens. They reported that the predicted fracture
sites were located in the sub-capital region in all specimens. FENN results pre-
dicted localized Cr.Dn in the femoral neck which is qualitatively in agreement
with possible fracture sites in femur necks [ 1 , 12 , 13 ].
In general, clinically available methods of estimating bone strength and the risk of
fracture of the femoral neck include bone densitometry or peripheral quantitative
computed tomography and imaging procedures. These techniques evaluate regional
bone density and morphology, which are partly related to fracture risk, but they are of
limited value for quantifying structural strength [ 12 , 13 ]. The FENN method, which
incorporates results averaged from real scanned 3D trabecular bone cores using
micro-CT FE models, could possibly achieve rapid multiscale non-invasive and
precise clinical assessments of the strength of the proximal femur and possible sites
of fracture. Previous micro-CT FE models [ 10 , 26 ] showed that microdamage ini-
tiation occurred prior to apparent yield at relatively low local principal strains in
compression. The authors suggested that local tissue yielding can in fact initiate at
very low apparent strains and that the apparent mechanical properties are degener-
ated through these localized effects. Morgan et al. [ 38 ] suggested that relatively small
amounts of microdamage have a major effect on the mechanical properties of bone.
Repetitive loading from everyday activities results in fatigue microdamage
accumulation leading to overall bone structural fragility. The histological evidence
indicates that fatigue damage occurs at the microstructural level (microcracks) and
at ultrastructural levels (diffuse damage) [ 50 ]. Bone microcracks resulting in the
disruption of osteocytic communication via the canalicular network may in fact be
an important stimulus providing spatial regulation of bone remodeling activity [ 8 ].
The balance between local remodeling and accumulation of trabecular bone fati-
gue microcracks is believed to play an important role in the maintenance of
skeletal integrity. However, the local mechanical parameters associated with the
initiation and propagation of fatigue microcracks are not well understood. Despite
the high clinical relevance of trabecular bone damage, the relationship between
local multiple microcracks and local stresses and strains at trabecular level is not
well understood. A quantitative assessment of trabecular level stresses and strains
associated with Cr.Dn and Cr.Le accumulation may provide insight into the
improvement of fracture risk assessment methods and bone repair simulations.
7 Summary and Conclusions
In this chapter, the focus was on the development and the implementation of a novel
multiscale approach referred to as the FENN method for crack density and crack
length accumulation in trabecular bone using finite element simulation and neural
network computation. The transition from mesoscale to macroscale is done by
means of the trained NN. The mesoscale model (NN) able to predict detailed
responses was incorporated into the macroscale model as a material formulation on
the integration point level at every FE iteration—e.g. the behaviour law needed to