Biomedical Engineering Reference
In-Depth Information
In an effort to create more realistic and sophisticated
in silico
models researchers
started incorporating true geometries obtained by various imaging methods (Mc-
Garry et al.,
2005
; Anderson and Knothe Tate,
2008
). The study by Anderson and
collaborators, for example, was based on high-resolution transmitted electron micro-
graphs to analyze stresses imposed on osteocytes by fluid drag, while McGarry and
colleagues incorporated previously reported images of cell spreading (Frisch and
Thoumine,
2002
) to assess the effect of fluid shear stress and strain on the mechani-
cal response of bone cells using FE analysis. Nevertheless, while computational and
experimental components of the investigations are rarely carried out within a single
integrative study addressing the same research question using both
in silico
and
in
vivo
modalities concomitantly, the possibility of direct validation remains slim.
A computational model for signaling pathways and interactions between os-
teoblast and osteoclasts has attempted to predict the effects of catabolic treatment
with parathyroid hormone (PTH), as well as to simulate the interaction between
receptor activator of nuclear factor-
κβ
, its ligand, and osteoprotegerin (RANK-
RANKL-OPG pathway), which is essential for osteoclast formation (Lemaire et
al.,
2004
). This complex
in silico
framework has reportedly been able to correctly
predict cellular interaction, and the effects of the common metabolic diseases, such
as estrogen deficiency, calcitriol deficiency, senescence and glucocorticoid excess.
The results of the simulation find convincing evidence in the extensive comparison
with literature; however no other direct validation has been undertaken. Other the-
oretical models with the focus on the prediction of molecular signaling pathways
and mechanobiology have also been presented (Potter et al.,
2005
; Pivonka et al.,
2008
; Lio et al.,
2011
); unfortunately, despite the fact that all of them strive to pre-
dict bone adaptation on the micro scale, none of them have been verified against
corresponding
in vivo
data, and thus are lacking confirmation of the level of fidelity.
The need for validation has also been emphasized for
in silico
models of cellular
chemotaxis and cytoskeletal reorganization (Loosli et al.,
2010
; Landsberg et al.,
2011
). Both investigations compare results of the computational simulations with
the
in vitro
experiments. In both reports sample geometries and boundary condi-
tions for the models were derived directly from the experimental data. For example,
the study by Landsberg and colleagues used a tetrahedral mesh for micro-CT re-
construction, as a starting point for the chemotaxis simulation, while Loosli and
colleagues reconstructed the shapes of the adhesive islands from the
in vitro
study
to computationally predict the adhesion sites of the cells (Fig.
27.1
). Such comple-
mentary experimental and
in silico
studies tend to enable better understanding of the
model limitations. For example, Landsberg and colleagues refer to a similar ongo-
ing experimental study, utilizing signaling molecules, for further model validation.
Loosli and colleagues, on the other hand, mention the algorithm's failure to predict
adhesion formation at curved geometries, due to a missing model parameter, as one
of the limitations, requiring further improvements.
The overall lack of adequate validation for the predictive value of microscale
models in bone mechanobiology has also been noted by other authors (Jacobs and
Kelly,
2011
; Webster and Müller,
2011
; Isaksson,
2012
). A particular concern of
validating models with
in vivo
data provided by collaborating investigators is that
