Pinheironeto 2017

Proceedings of the 20th World Congress
Proceedings of the
The International 20th World
Federation of Congress
Automatic Control
Proceedings
The of the
International 20th Worldof Congress
Toulouse, France,Federation Automatic
July 9-14, 2017 Control
Available online at www.sciencedirect.com
The International Federation of
Toulouse, France, July 9-14, 2017Automatic Control
Toulouse, France, July 9-14, 2017
ScienceDirect
IFAC PapersOnLine 50-1 (2017) 8066–8071
Variable
Variable Order
Order Differential
Differential Models of Bone
Differential Models of Bone
Variable OrderRemodelling Models of Bone
Remodelling
Remodelling
∗ ∗ ∗
Joana Pinheiro Neto ∗ Rui Moura Coelho ∗ Duarte Valério ∗
JoanaSusana Pinheiro
VingaNeto ∗ ∗ Rui Moura Coelho
∗∗ ∗ Duarte Valério∗∗
∗ Dominik Sierociuk ∗∗ Wiktor Malesza
∗ ∗
Joana Pinheiro Neto Rui Moura Coelho Duarte Valério ∗∗
Susana Vinga Michal ∗ Dominik
Macias ∗∗Sierociuk ∗∗
Andrzej Wiktor ∗∗
Dzieliński Malesza ∗∗
∗
Susana Vinga Dominik∗∗Sierociuk Wiktor ∗∗ ∗∗ Malesza ∗∗ ∗∗
Michal Macias ∗∗ ∗∗ Andrzej Dzieliński ∗∗ ∗∗
Michal Macias Andrzej Dzieliński
∗
∗ IDMEC, Instituto Superior Técnico, Universidade de Lisboa, Lisboa,
∗ IDMEC,
∗ Instituto Superior
IDMEC, Instituto Superior
Técnico,
Portugal Universidade de Lisboa, Lisboa,
(e-mail:
Técnico, Universidade de Lisboa, Lisboa,
Portugal (e-mail:
{joana.neto,rui.coelho,duarte.valerio,susanavinga}@tecnico.ulisboa.pt).
Portugal (e-mail:
∗∗
∗∗ Institute of Control and Industrial Electronics, Warsaw University
∗∗ Institute
∗∗
Institute
of Control
of Technology,
of Control
and Industrial
Warsaw, Electronics,
Poland (e-mail:
and Industrial Electronics,
Warsaw
{dsieroci,
Warsaw
University
wmalesza,
University
of Technology, Warsaw,
michal.macias, Poland (e-mail: {dsieroci,
adziel}@ee.pw.edu.pl) wmalesza,
of Technology, Warsaw, Poland (e-mail: {dsieroci, wmalesza,
michal.macias, adziel}@ee.pw.edu.pl)
michal.macias, adziel}@ee.pw.edu.pl)
Abstract: Bone is a living tissue undergoing a constant remodeling process that involves
Abstract:
different cellBone types,is a living tissue undergoing a constant This remodeling
process process that involves
Abstract: Bone isinaparticular
living tissue osteoclasts and osteoblasts.
undergoing a constant remodeling can be modeled
process that involvesusing
different
differential cellequations,
types, in particular
accounting osteoclasts
for the and osteoblasts.
biochemical coupling This processthe
between canreferred
be modeled bone using
cells.
different cell types, in particular osteoclasts and osteoblasts. This process can be modeled using
differential
These models equations,
have also accounting
been extended for thetobiochemical
include thecoupling
effects ofbetween
bone diseasesthe referred
in its bone
dynamics,cells.
differential equations, accounting for the biochemical coupling between the referred bone cells.
These
such asmodelstumor have also been extended to include the effects of bone diseases in its dynamics,
metastizations.
These models have also been extended to include the effects of bone diseases in its dynamics,
such
Due as the
to tumor high metastizations.
number of parameters involved in the existing physiological models for bone
such as tumor metastizations.
Due to the high
remodeling, a newnumber
approach of parameters
to this systeminvolved in the
is here existing Variable
presented. physiological ordermodels for bone
derivatives are
Due to the high number of parameters involved in the existing physiological models for bone
remodeling,
introduced asa anew a new approach
method to
to simplify this system is here presented. Variable order derivatives
that leadare
remodeling, approach to thisitssystem
structure, providing
is here presented.moreVariable
compactorder models derivatives to
are
introduced
similar results as atomethod
those ofto simplify
originalits structure, providing more compact models forthat lead to
introduced as a method tothesimplify itsformulations. These new
structure, providing moremodels
compact also allow
models anomalous
that lead to
similar
diffusionresults
to be to those of the
accounted for. original formulations. These new models also allow for anomalous
similar results to those of the original formulations. These new models also allow for anomalous
diffusion
These models to be accounted for.
diffusion to be are expected
accounted for.to serve in clinical decision systems such as personalized therapy
These
schemes. models are expected to serve in clinical decision systems such as personalized therapy
These models are expected to serve in clinical decision systems such as personalized therapy
schemes.
schemes.
© 2017, IFAC (International Federation of Automatic Control) Hosting by Elsevier Ltd. All rights reserved.
Keywords: Bone remodeling, Bone metastasis, Fractional derivatives, Variable order derivatives
1. INTRODUCTION However, as the model grows to include the nefarious
1. INTRODUCTION However, as the model
action of metastases, it alsogrowsgrows to ininclude
generalthe nefarious
complexity.
1. INTRODUCTION However, as the model grows to include the nefarious
action of metastases,
Bone is not made of static tissue: it undergoes spatially action of metastases, it also grows in general complexity. it also grows in general complexity.
Bone is not made of static tissue: it undergoes spatially Biological processes often present anomalous diffusion
heterogeneous
Bone is not made and of asynchronous
static tissue:remodeling
it undergoes cycles. This Biological processes often present anomalous diffusion
spatially
heterogeneous and asynchronous remodeling cycles. This (Magin, 2004),
Biological and fractional
processes often present derivatives
anomalous can be intro-
diffusion
tightly coupledand mechanism involves cells termed
cycles.osteo-
This (Magin,
heterogeneous
tightly coupled
asynchronous
mechanism involves
remodeling
cells termed osteo- duced
(Magin, to2004),
betterand
2004), and
fractional
simulate
fractional thosederivatives
derivatives
can be intro-
diffusion characteristics
can be intro-
clasts, responsible
tightly coupled mechanism for bone involves
resorption, cellsand cells osteo-
termed called duced to better simulate those diffusion characteristics
clasts, responsible for bone resorption, and cells called (Sierociuk
duced to et al., simulate
better 2013; Rahimy, those 2010). However,
diffusion many
characteristics
osteoblasts,
clasts, that account
responsible for bone forresorption,
bone formation (Crockett
and cells called (Sierociuk et al., 2013; Rahimy, 2010). However, many
physical processes appear to exhibit a fractional order
osteoblasts,
et al., 2011).that
osteoblasts,
that
In the account
presence
account
for bone
forofbone
a tumor,formation
the bone
formation micro- physical processes appear to exhibit a fractional many
(Crockett
(Crockett
(Sierociuk et al., 2013; Rahimy, 2010). However,
order
et al., 2011). In the presence of a tumor, the bone micro- behavior processes
physical that variesappear with time to or space
exhibit a (Valério and
fractional Sá
order
environment
et al., 2011). In is the
disrupted
presence andof aseverely
tumor, affected
the bone by the behavior that varies with time or space (Valério and
micro- Sá
environment is disrupted and severely affected by the da Costa,
behavior 2013;
that Lorenzo
varies with andtimeHartley,
or 2002).
space (Valério and Sá
disease. For an
environment is osteolytic
disrupted tumor type, bone
and severely mass density
affected by the da Costa, 2013; Lorenzo and Hartley, 2002).
disease.
decreasesFor For an
as an osteolytic
theosteolytic tumor
bone micro-environment type, bone mass density da Costa, 2013; Lorenzo and Hartley, 2002).
disease. tumor type, bonebenefits tumor In this paper, the referred models for bone remodeling
mass density
decreases
growth (Casimiroas the bone et al.,micro-environment
2009). Studying benefits tumor In
such behavior in this paper, the referred models for boneByremodeling
decreases as the bone micro-environment benefits tumor In athis tumorous
paper, the environment
referred models are changed. including
for bone remodeling
growth
is important(Casimironot onlyet al., 2009).
because Studying such behavior in a tumorous environment are changed. Bythe including
growth (Casimiro et al., 2009).ofStudying
primary such tumors in the fractional
behavior variable order derivatives,
in a tumorous environment are changed. By including to explain altered
is important
bone, but alsonot not only
to account because of
for metastaticprimary tumors
developments in the fractional variable order derivatives, to explain the altered
is important only because of primary tumors in due healthy behavior
the fractional variable of orderbone, more succinct
derivatives, to explain equations
the alteredare
bone,
to other but also to account for metastatic developments due healthy behavior of bone, more succinct equations are
bone, buttypes
also ofto cancer,
accountwhere breast and
for metastatic prostate cancer
developments achieved.behavior
due healthy Hence, this of bone,paper’s more novelty
succinct consists in using
equations are
to
are other types
specially of cancer,
prone to such where breast and
development prostate
(Suva et al., cancer
2011). achieved.
variable Hence,
order this paper’s
derivatives, for novelty
the first consists
time in in using
bone re-
to other types of cancer, where breast and prostate cancer achieved. Hence, this paper’s novelty consists in using
are specially
Duespecially prone to
to the severity such
of such development
this type of disease, (Suva etmodels
theseet al., 2011).
are variable
modeling,order derivatives,
to provide similar forresults
the first to time
those in bone re-
are prone to development (Suva al., 2011). variable order derivatives, for the first time inof bone
existingre-
Due to
expected the severity of this type of disease, these models are modeling, to provide similar results to those of existing
Due to thetoseverity
provideofclinical
this type decision systems
of disease, with
these efficient
models models buttowith
are modeling, fewersimilar
provide variables and parameters.
results to those of existing
expected
personalized to provide
therapies. clinical decision systems with efficient models but with fewer variables and parameters.
expected to provide clinical decision systems with efficient models Only local butmodels
with fewer
are herevariables and parameters.
considered, as diffusion of bone
personalized therapies.
personalized therapies.
A healthy remodeling mathematical model was initially Only Only local
cells islocal models
leftmodels
for futureare here considered,
In all thatas
work.considered, diffusion D11of bone
are here asfollows,
diffusion ofisbone
the
A healthyinremodeling
proposed Komarova mathematical
et al. (2003), andmodel wasextended
further initially cells
first is lefttime
order for future
derivativework. ( dIn all that follows, D 1 is the
); the variable order is given
A healthy remodeling mathematical model was initially cells is left for future work.dt dIn all that follows, D is the
1
proposed
in Ayati etin inal.
Komarova
(2010) toet etinclude
al. (2003), and further
a tumorous extended first order timeα(t)
metastization derivative ( dt d );
d the variable order is given
proposed Komarova al. (2003), and further extended by firstα(t)
order andtime derivative
Dα(t) represents ( dt
dt );
thethetimevariable
dependentorder variable
is given
in Ayati et al. (2010) to include a tumorous metastization
a non-local by α(t) and Dα(t)
influence
in Ayati etinal. bone dynamics.
(2010) to include It also proposed
a tumorous metastization by α(t) and D α(t) represents
dα(t)
represents
the time dependent variable
the time dependent variable
influence in bone dynamics.with It also proposed a non-local order derivative ( α(t) ); all models here presented use
one-dimensional
influence in boneapproach
dynamics. It also diffusion,
proposed for both cases. order derivative ( dt
a non-local dα(t)
α(t)
α(t) ); allparameters,
models here presented use
one-dimensional approach with diffusion, for both cases. order dimensionless variables
derivative ( dtα(t)d
d
dt
α(t)
and including
); all models here presented use the cell
one-dimensional
This work was supportedapproach by withFCT,diffusion,
throughfor both cases.
IDMEC, under dimensionless
populations (see variables
dtα(t) and parameters, including the cell
Table 1).
This work was supported by FCT, through IDMEC, under dimensionless variables and parameters, including the cell
LAETA,
projects
was UID/EMS/50022/2013, BoneSys, joint Polish- populations (see Table 1).
This work
LAETA, projects
supported by FCT, through IDMEC, under populations (see Table 1).
Section 2 presents previously published models. Section 3
Portuguese
LAETA, projectsproject UID/EMS/50022/2013,
Modelling and controllingBoneSys,
UID/EMS/50022/2013, BoneSys,
joint Polish-
cancer evolution using
joint Polish-
Portuguese project Modelling and(PTDC/EMS-SIS/0642/2014)
controlling cancer evolution using Section
concerns22variable
presentsorderpreviously published
derivatives. Section models.
4 givesSection
the new3
fractional calculus, PERSEIDS
Portuguese project Modelling and controlling cancer evolution using and Section presents previously published models. Section 3
fractional calculus, PERSEIDS (PTDC/EMS-SIS/0642/2014) and
IF/00653/2012. concerns variable order derivatives. Section 4 gives the new
fractional calculus, PERSEIDS (PTDC/EMS-SIS/0642/2014) and
IF/00653/2012.
concerns variable order derivatives. Section 4 gives the new
IF/00653/2012.
Copyright
2405-8963 ©© 2017,
2017 IFAC 8400Hosting by Elsevier Ltd. All rights reserved.
IFAC (International Federation of Automatic Control)
Copyright
Peer review©under
2017 responsibility
IFAC 8400Control.
of International Federation of Automatic
Copyright © 2017 IFAC 8400
10.1016/j.ifacol.2017.08.1233
Proceedings of the 20th IFAC World Congress
Toulouse, France, July 9-14, 2017 Joana Pinheiro Neto et al. / IFAC PapersOnLine 50-1 (2017) 8066–8071 8067
models developed, and simulation results are provided. 100
Section 5 sums up future work. 90
80
2. EXISTING MODELS FOR TUMOROUS BONE 70
REMODELING
Tumor [%]
60
50
40
This section overviews some of the published models for 30
healthy and tumorous bone remodeling environments that 20
were the starting point for this paper. 10
In the model for healthy bone tissue remodeling, proposed 0 200 400 600
Time [days]
800 1000 1200
in Komarova et al. (2003) and given by Equations 1,

bone remodeling takes the form of an S-system (Savageau Fig. 1. Tumor evolution, according to Equation 2c. Pa-
and Arbor, 1988). Coupling the behavior of osteoclasts, rameters are presented in Table 1, having the same
C(t), and osteoblasts, B(t), is done through biochemical meaning and value as in Ayati et al. (2010).
autocrine (gCC , gBB ) and paracrine (gBC , gCB ) factors
expressed implicitly in the system’s exponents. Bone mass verified in the first two plots of Figure 2. However, as the
density, z(t), is assumed to be determined by the extent to tumor continuously promotes the activity of osteoclasts
which normalized values of C(t) and B(t) exceed nontrivial and inhibits osteoblasts (Suva et al., 2011), bone loss
steady state levels, CSS and BSS , respectively. Below such occurs.
values, the populations of osteoclasts and osteoblasts are
assumed to consist of less differentiated cells that are 3. VARIABLE ORDER DERIVATIVES
unable to resorb or build bone, but are able to partici-
pate in autocrine and paracrine signaling. Constants κC The approximation of fractional constant order derivative,
and κB represent bone resorption and formation activity, in an iterative form, according to the Grünwald-Letnikov
respectively. definition, is as follows.
Definition 1. The iterative fractional order derivative is
D1 C(t) = αC C(t)gCC B(t)gBC − βC C(t) (1a) defined as
n
D1 B(t) = αB C(t)gCB B(t)gBB − βB B(t) (1b) α 1 α
0 Dt f (t) ≈ α (−1)r f (t − rh),
D1 z(t) = −κC max [0, C(t) − Css ] h r=0 r
+κB max [0, B(t) − Bss ] (1c) where h ∈ R+ is a step time, and n = t/h.
An osteolytic tumor burden in the healthy bone remod- There also exists a recursive equivalent of this constant
eling environment, as proposed in Ayati et al. (2010) order definition, given as follows:
and presented in Equations 2, acts through the autocrine
and paracrine regulations pathways in the form of rij Definition 2. The recursive fractional order derivative ap-
parameters. New variable T (t) represents the tumor cells proximation is defined as follows:
density at time t, with a Gompertz form of constant n
growth γT > 0, independent of bone loss, and translating rec α f (t) r −α r α
0 Dt f (t) ≈ − (−1) D f (t − rh) .
a possible maximum tumor size of LT that can be seen hα r=1
r 0 t−rh
in Figure 1. Bone mass is the same as the healthy case
(Equation 1c). Both definitions, for an implementation length equal to
the duration of the simulation, are equivalent (Sierociuk
T (t)
T (t)
et al., 2015b).
g 1+rCC L g 1+rBC L
D1 C(t) = αC C(t) CC T B(t) BC T
−βC C(t) (2a) The recurrent relation allows to write the following vari-
able order definition:
g
CB
T (t)
T (t)
Definition 3. (Sierociuk et al., 2015b) The D-type of frac-
1 1+r gBB −rBB
D B(t) = αB C(t) CB LT
B(t) LT
tional variable order derivative approximation is defined
−βB B(t) (2b) as follows:
L n

D1 T (t) = γT T (t) log T (2c) D α(t) f (t) −α(t) D α(t)
T (t) 0 Dt f (t) ≈ − (−1)r 0 Dt−rh f (t) .
hα(t) r=1 r
In Figure 2, simulation results are presented for both The D-type derivative, given by Definition 3, corresponds
Equations 1 and 2. Osteoclasts, osteoblasts and bone mass to the input-reductive switching scheme as presented in
can be compared in a healthy (presented in red & bold) (Sierociuk et al., 2015b). The input-reductive switching
and in a tumorous (presented in blue) environment. For strategy assumes that, in a chain of differentiators, for
the healthy case, it can be seen that periodical oscillations changing the order, the differentiators from the input are
are sustained. However, being the tumor here considered rejected. This strategy means that the effect of order
of an osteolytic nature, it tends to diminish bone mass switching (changing) starts immediately after switch and
as it grows. This is achieved by promoting osteoclasts without output switches (as it is characteristic for example
and inhibiting the action of osteoblasts, which can be in the A-type definition). More about other types of
8401
8068 Joana Pinheiro Neto et al. / IFAC PapersOnLine 50-1 (2017) 8066–8071
Osteoclasts Population
Tumorous Case [cells]

20
Healthy Case [cells]

15 15
10 10
5 5
0 0
0 200 400 600 800 1000 1200
Time t [days]
Osteoblasts Population

0 200 400 600 800 1000 1200

Time t [days]
Bone Mass
Tumorous Case [%]

100 100
Healthy Case [%]
80 80
60 60
40 40
20 20
0 0
0 200 400 600 800 1000 1200
Time t [days]
Fig. 2. Simulation results for the osteoclasts, osteoblasts, and bone mass, respectively. In red & bold, a healthy
periodic remodeling response is presented, triggered by a rise of the osteoclast 10 units above its steady-sate
(Equations 1); in blue, the bone micro-environment evolution when a tumorous action is taking place according
to Figure 1 (Equations 2). Used parameters have the same meaning and value as described in Ayati et al. (2010),
and presented in Table 1.
definitions and equivalent switching schemes is presented

in (Sierociuk et al., 2015a; Macias and Sierociuk, 2014).
The D-type definition has been successfully used for mod-
elling the heat transfer process in a media such that struc-
ture is changed in time (Sakrajda and Sierociuk, 2016).
That is why the D-type definition will be used in our paper.
Additionally, this recursive definition allows also to define
an infinite length of constant initial conditions, which will
be necessary during simulation process. The D-type defi-
nition of variable order derivative, with initial conditions
in the form of an infinite length constant function, can be
written as follows: Fig. 3. Results of α = 0.5 integrators output with initial
Definition 4. (Sierociuk et al., 2015b) The fractional vari- conditions c = 0.5 and different values of T — the
able order derivative approximation for initial condi- length of initial conditions function
tions in the form of infinite length constant function
D α
−∞ Dl f (t) = c = const, for l = (−∞, 0), is given by the
following relation: 4. BONE REMODELING WITH VARIABLE ORDER
DERIVATIVES
As the tumor equation presented is independent of the

D α(t) f (t) bone microenvironment (Ayati et al., 2010), and bone
−∞ Dt f (t)≈ mass is but a reflection of osteoclasts and osteoblasts ac-
hα(t)
 tivity (Komarova et al., 2003), a new model is formulated
n
−α(t) α(t−jh) by changing the derivative’s order of the osteoclasts and
− (−1)j D
−∞ Dt−jh f (t) − c + c . osteoblasts equations only. This was done in an effort to
j=1
j
reduce the number of parameters involved in the original
equations with the tumor action.
It is the variable order, influenced by the tumor dynamics,
As it can be seen in Fig. 3, the results of variable order that is now responsible for inducing in the original healthy
derivative are different for different lengths of the initial model (Equations 1) the same response as the tumorous
conditions function. Only the results for an infinite length bone one (Equations 2). The tumor relates to the imposed
of initial conditions function given by Definition 4 are, in order through Equation 3e, where ρ(t) takes a linear form
such a case, expected to be a constant function. in time, with a constant θ gradient. Consequently, the
8402
Table 1. Variables and parameters used for simulation of Equations 1 and 2. Column A presents
the values used for the healthy bone simulation; Column B addresses the tumorous bone
remodeling micro-environment values. Parameters, in both meaning and value, follow what
was presented in Ayati et al. (2010).
Variables Description Units
t Time days
C(t) Number of osteoclasts —
B(t) Number of osteoblasts —
z(t) Bone mass %
T (t) Bone metastases size %
Parameters Description A B Units
αC Osteoclasts activation rate 3 3 day−1
αB Osteoblasts activation rate 4 4 day−1
βC Osteoclasts apoptosis rate 0.2 0.2 day−1
βB Osteoblasts apoptosis rate 0.02 0.02 day−1
gCC Osteoclasts autocrine regulator 1.1 1.1 —
gBC Osteoclasts paracrine regulator −0.5 −0.5 —
gCB Osteoblasts paracrine regulator 1.0 1.0 —
gBB Osteoblasts autocrine regulator 0 0 —
rCC Tumorous effect in osteoclasts autocrine regulation — 0.005 —
rBC Tumorous effect in osteoclasts paracrine regulation — 0 —
rCB Tumorous effect in osteoblasts paracrine regulation — 0 —
rBB Tumorous effect in osteoblasts autocrine regulation — 0.2 —
κC Bone resorption rate 0.0748 0.0748 day−1
κB Bone formation rate 0.00063952 0.00063952 day−1
LT Maximum size of bone metastases — 100 %
γT Metastases growth rate — 0.005 % day−1
C(0) Initial number of osteoclasts 11.16 15 —
B(0) Initial number of osteoblasts 213.72 316 —
z(0) Initial bone mass percentage 100 100 %
T (0) Initial bone mass percentage — 1 %
Css Steady-state osteoclasts number 1.06 5 —
Bss Steady-state osteoblasts number 213.72 316 —
action of the tumor in bone dynamics is represented by (magenta for Equations 3), both having a tumor evolu-
a simpler model, with less variables. tion according to Figure 1. It can be seen that, for the
osteoclasts, osteoblasts and bone mass, the response is
D α(t)
similar in both cases. However, the osteoclast population
−∞ Dt C(t) = αC C(t)gCC B(t)gBC − β C C(t) (3a) grows less in the new variable order case, and the remod-
D α(t) gCB gBB eling cycle duration is prolonged. Due to the longer cycle
−∞ Dt B(t) = αB C(t) B(t) − β B B(t) (3b)
1 and the newly calculated resorption-formation ratio, R,
D z(t) = −κC [0, C(t) − Css ] + κB [0, B(t) − Bss ] (3c)
bone mass loss is quickened in the beginning of the tumor
LT
D1 T (t) = γ C T (t) log (3d) growth. As the bone mass suffers severe losses, both bone
T (t) cells gradually decrease as a consequence of a lesser bone
α(t) = 1 − T (t)ρ(t) = 1 − T (t)θt (3e) availability due to tumor maximization.
However, some adaptations were required. As the new An overall analysis to the models’ structure results in 3
model’s steady-state is the same as that of the model suppressed parameters, as all 4 rij were removed and θ
for healthy bone, and most parameters remain with the was added to the orders equation. One additional equation
same value, the associated activity of osteoclasts and had to be added, to impose the order behavior.
osteoblasts in bone mass must differ from the origi-
nal case. Then, a new bone resorption-formation ratio, 5. CONCLUSIONS AND FUTURE WORK
t
max[0,C(t)−Css ]
R = 0t , is determined between 0 and t, Biochemical models for bone remodeling are expected to
max[0,B(t)−Bss ]
0
that corresponds to the completion time of a single cycle provide valuable insights about the bone complex system
of C(t) and B(t) in the new model. This method is the and to support the development of clinical decision sys-
same followed in Ayati et al. (2010), for a healthy bone tems for bone pathologies with efficient targeted therapies.
environment. Bone resorption and formation activities are Hence, a simplified model that mimics bone behavior with
then given by κC = rR and κB = r, respectively. a tumorous action is but an essential tool for a tailored
treatment for each patient. Here, new approaches to the
In Figure 4, a comparison is presented between the original existing biochemical models (Ayati et al., 2010; Komarova
tumorous model (blue for Equations 2, as in Figure 2), et al., 2003) were proposed. Variable order derivatives
and the new variable order model for tumorous action were introduced, for the first time in local biochemical
8403
8070 Joana Pinheiro Neto et al. / IFAC PapersOnLine 50-1 (2017) 8066–8071
Osteoclasts Population

20

15
10 10
0 0
0 200 400 600 800 1000 1200
Osteoblasts Population
1000 1000

800 800
600 600
400 400
200 200
0 0
0 200 400 600 800 1000 1200
Bone Mass
Tumorous Case [%]

100 100
Healthy Case [%]
80 80
60 60
40 40
20 20
0 0
0 200 400 600 800 1000 1200
Time t [days]
Fig. 4. Simulation results for the model with variable order derivatives inducing the tumorous behavior on the bone
micro-environment (osteoclasts, osteoblasts and bone mass, respectively). In blue, the evolution of the tumor
disrupted equations of Equations 2, as presented in Ayati et al. (2010); in magenta & bold, evolution of the same
populations simplified through variable order derivatives, as in Equations 3. For the latter, the resorption rate used
was R = 0.0044. Parameters given in Table 2.
Table 2. Additional variables and parameters of the proposed models with variable order
derivatives (Equations 3), as used for the simulation results of Figure 4. Unmentioned parameters
can be found in column B of Table 1.
Variables/Parameters Description Values Units
α(t) Time dependent variable order — —
θ Variable order gradient 8.33 × 10−9 —
κC Bone resorption rate 0.1548 % day−1
κB Bone formation rate 6.8176 × 10−4 % day−1
γT Bone metastases growth rate 0.004 % day−1
bone remodeling models, allowing for a reduced set of chanical solicitations in the equations (Belinha et al.,
parameters to describe similar tumorous results to those 2015a,b).
from the original formulations. A more compact model,
that promptly highlights tumorous bone interactions, is Finally, it would be highly desirable to find model co-
then achieved, with good qualitative simulations of what efficient values from experimental data — possibly data
is known for the bone dynamics. extrapolated from experiments with animals, since col-
lecting the data from humans may be too expensive and
Future work includes: unethical, or eventually data resulting from new medical
imaging techniques (Birkhold et al., 2015). In fact, model
• Applying these techniques to models that are not
parameters in this paper are little more than educated
local (Ayati et al., 2010; Valério et al., 2016) That is
guesses by clinicians and oncobiologists at the magnitude
to say, to models where variables depend not only of
of the values, leading to reasonable results from the qual-
time but also of space (e.g. C(t, x, y, z), B(t, x, y, z,
itative point of view. Finding actual experimental values
etc.), and where diffusion processes are explicitly
is probably the biggest challenge facing the model.
present.
• Applying these techniques to models that include a The comprehensive analysis and simulation of these mod-
more detailed description of the biochemical processes els are expected to bring new insights on bone physiology
involved (Coelho et al., 2016; Valério et al., 2016). and to provide a better understanding of cancer control
• Applying these techniques to models with the effects strategies, and therefore supporting the relief of the mil-
of available treatments of cancer, and their pharma- lions of patients diagnosed yearly.
cokinetic and pharmacodynamic effects (Ayati et al.,
2010; Coelho et al., 2016).
• Applying these techniques to models that incorporate REFERENCES
as well biomechanical effects, i.e. the effects of me- Ayati, B.P., Edwards, C.M., Webb, G.F., and Wikswo,
J.P. (2010). A mathematical model of bone remodeling
8404
dynamics for normal bone cell populations and myeloma plied Mathematical Modelling, 39(13), 3876–3888.
bone disease. Biology Direct, 5(28). Http://dx.doi.org/10.1016/j.apm.2014.12.009.
Belinha, J., Dinis, L.M.J.S., and Natal Jorge, R.M. Sierociuk, D., Malesza, W., and Macias, M. (2015b). On
(2015a). The mandible remodelling induced by dental the recursive fractional variable-order derivative: Equiv-
implants: a meshless approach. Journal of Mechanics in alent switching strategy, duality, and analog modeling.
Medicine and Biology, 15(4), 1550059. Circuits, Systems, and Signal Processing, 34(4), 1077–
Belinha, J., Dinis, L.M.J.S., and Natal Jorge, R.M. 1113. doi:10.1007/s00034-014-9895-1.
(2015b). The meshless methods in the bone tissue Suva, L.J., Washam, C., Nicholas, R.W., and Griffin, R.J.
remodelling analysis. Procedia Engineering, 110, 51–58. (2011). Bone metastasis: mechanisms and therapeutic
Birkhold, A.I., Razi, H., Weinkamer, R., Duda, G.N., opportunities. Nature Reviews Endocrinology, 7, 208–
Checa, S., and Willie, B.M. (2015). Monitoring in 218.
vivo (re)modeling: a computational approach using 4D Valério, D., Coelho, R.M., and Vinga, S. (2016). Frac-
microCT data to quantify bone surface movements. tional dynamic modelling of bone metastasis , microen-
Bone, 75, 210–221. vironment and therapy. In International Conference
Casimiro, S., Guise, T.A., and Chirgwin, J. (2009). The on Fractional Differentiation and its Applications. Novi
critical role of the bone microenvironment in cancer Sad, Serbia.
metastases. Molecular and Cellular Endocrinology, Valério, D. and Sá da Costa, J. (2013). An Introduction to
310(1-2), 71–81. Fractional Control. IET, Stevenage. ISBN 978-1-84919-
Coelho, R.M., Lemos, J.M., Valério, D., Alho, I., Ferreira, 545-4.
A.R., Costa, L., and Vinga, S. (2016). Dynamic model-
ing of bone metastasis, microenvironment and therapy
— integrating parathyroid hormone (PTH) effect, an-
tiresorptive treatment and chemotherapy. Journal of
Theoretical Biology, 391, 1–12.
Crockett, J.C., Rogers, M.J., Coxon, F.P., Hocking, L.J.,
and Helfrich, M.H. (2011). Bone remodelling at a
glance. Journal of cell science, 124, 991–998. doi:
10.1242/jcs.063032.
Komarova, S.V., Smith, R.J., Dixon, S.J., Sims, S.M., and
Wahlb, L.M. (2003). Mathematical model predicts a
critical role for osteoclast autocrine regulation in the
control of bone remodeling. Bone, 33, 206–215.
Lorenzo, C.F. and Hartley, T.T. (2002). Variable Frac-
tional Order and Distributed Order Operators. National
Aeronautics and Space Administration (NASA), (Febru-
ary).
Macias, M. and Sierociuk, D. (2014). An alternative
recursive fractional variable-order derivative definition
and its analog validation. In Proceedings of Interna-
tional Conference on Fractional Differentiation and its
Applications. Catania, Itally.
Magin, R.L. (2004). Fractional Calculus in Bioengineering.
Begell House.
Rahimy, M. (2010). Applications of Fractional Differential
Equations. Applied Mathematical Science, 4(50), 2453–
2461.
Sakrajda, P. and Sierociuk, D. (2016). Modeling heat
transfer process in grid-holes structure changed in time
using fractional variable order calculus. In Proceedings
of the 8th Conference on Non-integer Order Calculus
and its Applications.
Savageau, M.A. and Arbor, A. (1988). Introduction to S-
systems and the underlying power-law formalism. II(3),
546–551.
Sierociuk, D., Dzieliński, A., Sarwas, G., Petras, I., Pod-
lubny, I., and Skovranek, T. (2013). Modelling heat
transfer in heterogeneous media using fractional cal-
culus. Philosophical Transactions of the Royal Society
of London A: Mathematical, Physical and Engineering
Sciences, 371(1990). doi:10.1098/rsta.2012.0146.
Sierociuk, D., Malesza, W., and Macias, M. (2015a).
Derivation, interpretation, and analog modelling of
fractional variable order derivative definition. Ap-
8405

Pinheironeto 2017

Uploaded by

Copyright:

Available Formats

You might also like

Pinheironeto 2017

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Pinheironeto 2017

Uploaded by

Copyright:

Available Formats

Proceedings of the 20th World Congress

models developed, and simulation results are provided. 100

Section 5 sums up future work. 90

2. EXISTING MODELS FOR TUMOROUS BONE 70

This section overviews some of the published models for 30

healthy and tumorous bone remodeling environments that 20

were the starting point for this paper. 10

in Komarova et al. (2003) and given by Equations 1,

Tumorous Case [cells]

Healthy Case [cells]

Tumorous Case [cells]

0 200 400 600 800 1000 1200

Tumorous Case [%]

definitions and equivalent switching schemes is presented

 As the tumor equation presented is independent of the

Tumorous Case [cells]

Healthy Case [cells]

Tumorous Case [cells]

Tumorous Case [%]

You might also like

As the tumor equation presented is independent of the