Escolar Documentos
Profissional Documentos
Cultura Documentos
Abstract
Cardiovascular disease is one of the major problems in todays medicine
and the number of patients increase worldwide. To find the correct treatment,
prior knowledge about function and dysfunction of the cardiovascular system
is required and method need to be developed that identify the disease in an
early stage.
Mathematical modeling is a powerful tool for prediction and investigation
of the cardiovascular system. It has been shown, that the Windkessel model,
drawing an analogy between electrical circuits and fluid flow, is a simple but
effective method to model the human cardiovascular system.
In this paper we have applied parametric local sensitivity analysis to a linear elastic model of the arm arteries including anastomosis to find and rank
sensitive parameters that may be helpful in clinical diagnosis. A computational model for anastomosis (anterior ulnar recurrent) is carried out to study
the effects of some clinically relevant haemodynamic parameters like blood
resistance and terminal resistance on pressure and flow at different locations
of the arm artery. In this context we also discuss the spatio-temporal dependency of local sensitivities. The percentage effect on the measurable state
variables pressure and flow, with respect to percentage change in cardiovascular input parameters, is quantified using norms. This novel methodology
allows us to verify the qualitative results obtained by sensitivity analysis.
The sensitivities with respect to flow resistance (R), arterial compliance
(C) and flow inertia (L), reveal that the flow resistance and diameter of the
vessels are most sensitive parameters. Those parameters play a key role in
diagnoses of severe stenosis and aneurysms. In contrast, wall thickness and
elastic modulus are found to be less sensitive. Results from time dependent
sensitivities show that R is sensitive at diastole and systole of the flow and
pressure wave respectively. While C and L are most sensitive at systole of
both flow and pressure waves respectively.
Sensitivities w.r.t. R and Z are obtained from inferior ulnar collateral anastomosis with anterior ulnar recurrent (IUC-AUR), which show that by increasing
R or equavalently by decreasing diameter of the anastomosis, the flow within
anastomosis decreases and the pressure difference yields back flow near endto-side anastomosis.
Keywords: computational cardiovascular model, cardiovascular parameters,
sensitivity analysis, anastomosis, Windkessel model.
Fachbereich
Introduction
viscous and inertial effects and also elastic properties of two vessels [13]. Details of
the valveless flow mechanism are given in [14]. In this paper, we present a computational model of anastomosis around elbow joint (interior ulnar collateral anastomosis
with anterior ulnar recurrent, IUC-AUR). As luminal diameter or equavalently the
blood resistance play an impotant role in blood flow through anastomosis, so, here
we will study the effects of blood resistance and terminal impedance on pressure
and flow through anastomosis.
The methods developed in this paper, are seen as the first step towards cardiovascular system identification from cardiovascular measurements. Within this work we
derive a computational cardiovascular model for the arm arteries (with and without
anastomosis) by using the lumped parameter approach. In a first instance we apply
local sensitivity analysis (without anastomosis) to study the effects of cardiovascular
parameters on the hemodynamic state variables. We study time dependent sensitivities w.r.t. R, C and L to find sensitive regions in the pressure and flow waves.
Due to the network structure of the cardiovascular system it is helpful to determine
and discuss the location dependent sensitivities i.e which locations in arm arteries
are sensitive w.r.t. R, C and L. Finally to quantify and compare our results, we
apply the concept of norms.
Under the assumption that the arterial tree is decomposed into short arterial segments of length l with a constant circular cross-section and linear elastic wall behavior, the following one dimensional flow equations can be derived from the linearized
Navier-Stokes equation, the equation of continuity and the shell-equation for thin
walled, linear elastic tubes [1, 16]
p
q
= Rq + L ,
x
t
(1)
q
p
p
=
+C .
(2)
x
Z
t
Within these equations the state variables are the flow q and the relative pressure p.
The viscosity and inertial forces of the blood flow are described by the viscous flow
resistance R and blood inertia L per unit length respectively. The elastic properties
of the wall are modeled by a compliance C per unit length, while Z is terminal
resistance [16]. Integration of the two partial differential eqns. (1) and (2) along
flow axis leads to a system of equations (3) commonly used to describe electrical
circuits (see Figure ??). In this type of model each segment of the arterial system is
described by a set of two equations that are known as Windkessel equations. Here
(pin , qout) and (pin, pout) are the boundary conditions for non-terminal and terminal
nodes respectively. To model a mean venous pressure with a value of 15mmHg for
terminal nodes the equation system is setup by including an additional terminal
resistance Z. The matrix form of the Windkessel eqns. with boundary conditions
is
dX
= AX + B
(3)
dt
Where, X = {p1 q1, ..., pNs qNs } is a state vector, A is a state matrix and B contain
the boundary conditions. A complete description of state-space representaion is
Figure 1: Linear elastic model for fluid flow in nonterminal vessel segments (left)
and for terminal vessel segments (right).
given in section 2.3
For non-terminal segments (Figure1:left) X = (qin, pout)T
R 1
pin
L
L
L
A=
,
B(pin , qout) =
1
qout
0
C
C
For terminal segments (Figure1:right) X = (qin, qout)T
R
pin
1
L
L
L
,
B(pin , pout) =
A=
1
1
out
pZC
ZC
C
The electrical parameters for ith segment in the arterial tree are related to the
physiological parameters of the fluid and vessel wall by:
Ri =
8l
,
r4
Li =
l
,
r2
Ci =
2r2l
.
Eh
(4)
where
E =
h =
Young modulus,
wall thickness,
l = length of vessel,
= blood viscosity,
r = radius of vessel
= blood density
2.1
While modeling arm arteries (with and without anastomosis), we adopt domain
decomposition (DD) approach in which the domain of whole cardiovascular systen is
decomposed into number of subdomains of cardiovascular segments. Each segment
4
Nodes
units
E
kgm2 s2
105
4
4
4
4
4
4
8
8
8
8
8
16
8
8
8
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
l
m
102
6.1
5.6
6.3
6.3
6.3
4.6
7.1
7.1
7.1
2.2
6.7
7.9
6.7
6.7
3.7
d
m
103
7.28
6.28
5.64
5.32
5
4.72
3.48
3.24
3
2.84
4.3
1.82
4.06
3.48
3.66
h
m
104
6.2
5.7
5.5
5.3
5.2
5
4.4
4.3
4.2
4.1
4.9
2.8
4.7
4.6
4.5
R
kgs1 m4
106
3.539
5.868
10.15
12.82
16.43
15.10
78.90
105
142.9
55.11
31.94
1173
40.19
50.22
33.60
C
m4 s2 kg1
1011
7.454
4.778
4.035
3.514
2.974
1.9
0.667
0.531
0.448
0.1207
1.067
0.0834
0.9366
0.80
0.3958
L
kgm4
106
1.539
1.898
2.648
2.976
3.369
2.76
7.838
9.042
10.55
3.647
4.844
31.88
5.434
6.075
3.693
Table 1: Numerical values of parameters for each node of the arm arteries(shown
in Fig. 2). The value of terminal resistance (Z) on three terminal nodes is 3.24
109 kgs1 m4 , = 0.004 kgs1 m1 and = 1050 kgm3 [4].
of the arm arteries in a network structure as given in Figure 2 is represented by an
electrical circuit as shown in Figure (1, top).
To built a cardiovascular network, relations between the arterial segments need to
be defined like at bifurcation flows diverge i.e. q1 = q2 +q3 and at anastomosis flows
merge i.e. q3 = q1 + q2. Pressure at bifurcation and at anastomosis remain same
(see Figure 1, bottom).
2.1.1
q4
q7
q10
q13
pin p1 R1 q1
p1 p2 R2q2
p2 p3 R3 q3
, q2 =
,
q3 =
L1
L2
L3
p3 p4 R4q4
p4 p5 R5q5
p5 p6 R6 q6
, q5 =
,
q6 =
L4
L5
L6
p6 p7 R7q7
p7 p8 R8q8
p8 p9 R9 q9
, q8 =
,
q9 =
L7
L8
L9
p9 p10 R10q10
p6 p11 R11q11
p11 p12 R12q12
, q11 =
, q12 =
L10
L11
L12
p11 p13 R13q13
p13 p14 R14q14
p14 p15 R15q15
, q14 =
, q15 =
L13
L14
L15
(5)
Pressure equations:
p1
p5
p9
2.1.2
q1 q2
q2 q3
q3 q4
q4 q5
, p2 =
,
p3 =
, p4 =
C1
C2
C3
C4
q5 q6
q6 q11 q7
q7 q8
q8 q9
, p6 =
, p7 =
, p8 =
C5
C6
C7
C8
q9 q10
q10 (p10 pout )/Z1
q11 q12 q13
, p10 =
, p11 =
C9
C10
C11
p12
p15
p14 =
q14 q15
,
C14
(6)
It was shown we have the AUR anastomosis in the model of arm artery (see Figure
2), the flow will split at nodes 3 and 6, and will merge at node 11. the inlet pressure
at node 11 is same as nodes 6 and 33. As we add three additional nodes so we have
now 5 new equations for pressure and flow. Additional equations for pressure and
flow of the ulnar recurrent artery are,
Anastomosis flow equations:
q3
q33
p2 p3 R3q3
p3 p31 R31 q31
, q31 =
L3
L31
p32 p6 R33 q33
L33
q32 =
q3 q4 q31
,
C3
p31 =
q31 q32
q3 q33
q3 + q6 q11
, p32 = 2
, p33 = 3
(8)
C31
C32
C6 + C33
2.2
Blood flow at junctions plays an important role in normal and pathological conditions of the cardiovascular system. In this section we briefly discuss pressure
and flow modeling at junctions for both diverging (at bifurcations) and merging
blood flows. In the arterial system merging flow conditions appear in the context
of anastomosis.
2.2.1
In the above model structure diverging flows occur at the bifurcation of the vessels
at nodes 3, 6 and 11 (see Figure 2). Flows at junction mainly depend on angle and
radius of the daughter vessels. In this paper we limit our study to symmetric and
asymmetric bifurcations with respect to radius of the daughter vessels. For diverging flows the mass conservation equation at node 3 is, q3 = q31 + q4. For linearized
6
Figure 2: Simplified anatomy of the arm arteries (left) and model geometry of
brachial, anterior ulnar recurrent, ulnar and radial arteries, with number of nodes
Ns = 15 (without anastomosis), Nas = 18 (with anastomosis) and number of terminal nodes Nt = 3 (right).
system (eqns. 9, 10) the total pressure through bifurcation remain unchanged, i.e.
the output pressure at node 3 is the input pressure for both nodes 31 and 4.
2.2.2
To model the merging flows at junctions has a great importance to understand the
effect of anastomosis in bypass in the cardiovascular system. These type of flows
occur also in vascular grafting and arteriovenous fistula (AVF). According to the
conservation of mass, the flow at node 11 is, q33 + q6 = q11 and according to the law
of conservation of momentum, the total pressure remains continuous at nodes 11.
2.3
State-space representation
= Axt1 + But ,
= Cxt + Dut .
(9)
(10)
Here xt is the state vector of the system, ut the input vector and yt the observation
vector. The dynamics of the system is described by the state dynamics matrix
A M (nn). The input matrix B M (ni) specifies the time dependency of the
in- and outflow boundary values and the observation matrix C M (m n) defines
the observation locations within the state-space system, i.e. the nodal location
in the network. Here m denotes the number of observations. Finally the input
to observation matrix D M (m i) adds the influence of the input vectors to
the observation vectors. Besides the computational advantage the state-space form
allows the integration of experimental measurements (observations) into the model
building process. This step is essential for the adjecent model parameter estimation
from experimental measurements, that are planned in a future study.
7
The state vector xt contains the flow and pressure functions at all network
locations, whereas the output vector yt contains the flow and pressure at selected
nodes i. For a m = 4 dimensional observation vector, the output vector is e.g.
y(t) = (q5(t), p5 (t), q6(t), p6 (t))T where y Rm at nodes 5 and 6. The state-space
system for the arm artery given in Figure 2, using eqn. (5) and eqn. (6) is defined
by
Aij =
Bij =
R
i+1
L
i+1
Ci
L i+1
12
L i+1
i = 1, 3, 5, ...29, j = i
i = 2, 4, 6, ...30, j = i 1
i = 1, 3, 5, ...29, j = i + 1
i = 3, 5, 7, ...29, j = i 1 and i 6= 21, 25
Ci
Zk C 2i
Li
Zj11 C i
Zj1 C i
Zj1 C 2i
i=j=1
i = 20, j = 2
i = 24, j = 3
(
1
, Cij =
0
i = 1, 2, 3, 4, j = i + 8
otherwise
i = 30, j = 4
otherwise
and Dij = 0.
For AUR anastomosis we expand our state-space model by using eqns. (7, 8). The
system is solved using the MATLAB built in solvers ODE 45 and lsim.
Sij =
yi
j
(11)
3.1
yi ( + ) yi ()
yi
= lim
,
0
(12)
where yi is the i-th model output, is the model input parameter and is the
change in model parameters. There are various methods to compute the sensitivity
coefficients in eqn. (12), within this work we use the method of finite difference:
Si =
yi
yi ( + ) yi ()
'
(13)
Equation (13) produces a set of two sensitivity time series Si (t) (one for pressure
and one for flow) per parameter and per network node (see figure 4).
Here we discuss three different scenarios of sensitivity analysis, (i) sensitivity analysis with respect to structural parameters ELdh, (ii) electrical parameters RCL
and (iii) location dependent sensitivities. In this context we also discuss the time
dependence of the sensitivities and compare the results of the two scenarios with
and without anastomosis
The sensitivity results are finally compared to the 2-norm of the distance vector of
the state variables of two time series.
3.1.1
To study the effects of Eldh on pressure and flow, we first solve our model with
both 10% variation of Eldh from its nominal values (see table 1). The sensitivities
are computed using eqn. (13). In this case is either Eld or h. The sensitivities for
pressure and flow w.r.t Eldh were calculated for node 7 (radial artery) and plotted
in figure 3.
3.1.2
Sensitivity time series with respect to R, C and L, can be used to find sensitive
regions in pressure and flow waves at each location of the arm arteries. For this
we first solve eqn. (9) and eqn. (10) with Matlab built in solvers, then calculate
sensitivity time series for both pressure and flow at nodes 5 and 7 by using eqn.
(13). Results are presented in figure 4
3.1.3
and Li are identical for each node. This allows us to analyze the influence of the
network structure onto sensitivity values at different node locations.
3.1.5
For section 3.1.3 and section 3.1.4, we calculate mean of sensitivity time series for
all arm arteries nodes. Therefore we compute the nodal sensitivity time series Si for
all nodal parameters as described in previous section. For each node, n, we obtain
two sensitivity time series (one for the pressure and one for the flow) per parameter
variation. The total number of possible parameter variations is 3n2, so we end up
with a set of 6n2 time series. To reduce the complexity, we average out the time
dependency by computing the mean of the absolute value of every time series. The
6n2 real values are used for further analysis in section 4. They are displayed into 6
matrices that characterize the sensitivity of pressure and flow (sensitivity nodes) in
the network structure based on changes in the electrical parameters RCL (selected
nodes). Each cell in Figure 5 and 6, represents the mean absolute value of time series
(Si ). Due to the fact that we use the pressure as an input and output boundary
condition, the change in pressure with any parameter at all inlet and terminal nodes
will be zero.
3.1.6
Blood flow in AUR anastomosis (collateral circulation) depends on size and mainly
on diameter of anastomosis, smaller the diameter lesser the flow in anastomosis and
vice versa. In this paper we limit our study to end-to-side AUR anastomosis and
show the sensitivity of anastomosis structure on pressure and flow by changing blood
resistance Ra and Rb and terminal resistance (Z) [1114]. Here we consider Ra =
R31 + R32 + R33 and Rb = R4 + R5 + R6, where Ra and Rb are the total resistance
through AUR anastomosis and through its parallel brachial artery respectively.
Further more we discuss 5 different scenarios (a) when Ra ' Rb, (b) Ra Rb, (c)
Ra Rb, (d) Z is small and (e) Z is large. The sensitivity is computed as earlier
stated.
3.2
k yi (2 , t) yi (1 , t) k2
kyi (1 , t)k2
i = 1, 2, 3...2 Ns ,
10
x 10
4.5
4
pressure [Pa]
flow [ml/s]
2.5
Phase and
amplitude shift
1.5
1
0
0.1
0.2
x 10
0.3
0.4
time [s]
0.5
Phase and
amplitude shift
0.9
0.6
0.7
0.8
0
0.8
0.2
1
0
0.1
pressure [Pa]
0.2
0.3
0.4
time [s]
0.5
0.6
0.7
0.8
0
0.8
pressure [Pa]
3
2.5
0.1
1.4
A7
pressure [Pa]
1.3
3
2.5
Norm sensitivity=0.58
0.2
0.3
Phase and
amplitude shift
1.5
0.1
0.2
0.3
0.4
time [s]
0.5
0.5
10%
+10%
Nominal values
0.6
0.7
0.8
Norm sensitivity=0.13
A8
1.2
1.1
Phase and
amplitude shift
0.4
time [s]
x 10
Norm sensitivity=0.36
3.5
1
0
1.5
10%
+10%
Nominal values
0.8
0.8
0
0.8
x 10
0.7
0.7
1.1
1
0.6
0.6
0.9
0.5
0.5
1.2
1.5
0.4
time [s]
0.4
time [s]
1.3
0.3
0.3
10%
+10%
Nominal values
1.4
0.2
0.2
x 10
1.5
4.5
0.1
Norm sensitivity=1.34
0.1
Phase and
amplitude shift
0.9
3.5
1
0
0.8
1.1
1.6
10%
+10%
Nominal values
0.7
1.2
x 10
4.5
0.6
Norm sensitivity=0.32
1
Phase and
amplitude shift
1.5
0.5
1.3
0.4
time [s]
10%
+10%
Nominal values
1.4
2.5
0.3
x 10
Norm sensitivity=0.95
A3
0.1
1.5
10%
+10%
Nominal values
3.5
flow [ml/s]
1.1
flow [ml/s]
1.2
flow [ml/s]
Norm sensitivity=0.13
1.3
10%
+10%
Nominal values
1.4
x 10
Norm sensitivity=0.36
3.5
1.5
10%
+10%
Nominal values
0.9
0.6
0.7
0.8
0
0.8
0.1
0.2
0.3
0.4
time [s]
0.5
0.6
0.7
0.8
Figure 3: Pressure and flow for a 10% change in E, l, d and h in the radial artery
at node 7. The results reveal that d and l are sensitive parameters while E and h
are less sensitive, for the reason of linear appearence E and h have same effects on
cardiovascular pressure and flow.
4.1
The sensitivity for pressure and flow are obtained by a variation of the cardiovascular
parameters E, l, d and h of arm arteries by 10%. From figure 3 it is directly evident,
that diameter and length are most sensitive parameters, while the elastic modulus
and wall thickness are comparatively less sensitive parameters.
11
4.2
4.3
12
15
x 10
0.5
Node 5
Node 7
B1
x 10
0.5
1.5
Node 5
Node 7
B2
R sensitivity
R Sensitivity
2
0
100
5
1.5
x 10
200
300
400
500
Time [ms]
600
700
100
13
Node 5
Node 7
C1
2
0
800
300
400
500
Time [ms]
600
700
800
x 10
Node 5
Node 7
C2
0.5
200
0.5
2
1
C sensitivity
1.5
2
0
100
14
x 10
1.5
200
300
400
500
Time [ms]
C sensitivity
600
700
6
0
800
Node 5
Node 7
D1
100
x 10
300
400
500
Time [ms]
600
700
800
Node 5
Node 7
200
L sensitivity
0.5
0
0
2
0.5
L sensitivity
1
1.5
0
100
200
300
400
500
Time [ms]
600
700
4
6
0
800
100
200
300
400
500
Time [ms]
600
700
800
Figure 4: Flow and pressure waves (A1 , A2) and sensitivity time series with respect
to R, C and L (B1 , B2 , C1, C2, D1 , D2) at nodes 5 and 7. Results shows that R is
sensitive at diastolic and systolic part of flow and pressure wave respectively. While
C and L are most sensitive at systolic part as compared to the diastolic part of
pressure and flow waves at respective nodes.
thus a change of C in any node has a small effect on pressure and flow in the whole
arm artery. In contrast a variation of the arterial compliance C in the ulnar and
radial arteries have a large (local) effect on pressure and large (global) effect on flow
especially in the brachial artery.
From eqn. (4) it is obvious that viscous resistance and blood inertia are inversely
related to r4 and r2 respectively. Which means in large arteries blood inertia plays
an important role, while in small arteries viscous resistance is more important. A
variation of blood inertia in the first node of the brachial artery has large influence on
pressure and flow of all following nodes of the arm arteries (see Figure 5 (bottom)).
However, we observe only minor local (within brachial artery) effects on flow when
we change L in each segment of the brachial artery. Further, due to the fact that
1
1
1
the total inductance Ltotal
= Lulnar
+ Lradial
reduces at the furcation, the flow and
pressure in the ulnar and radial arteries are less sensitive with respect to L.
13
14
x 10
6
Selected nodes
Selected nodes
Brachial artery
4
10
2
10
12
10
Radial artery
12
10
8
10
C sensitivities of flow
on brachial artery
12
14
2
6
8
10
Sensitivity nodes
12
2
14
7
6
10
12
2
1
2
6
8
10
Sensitivity nodes
12
14
12
x 10
Selected nodes
Brachial artery
1.5
8
1
10
12
13
x 10
14
Selected nodes
12
14
6
8
10
Sensitivity nodes
x 10
14
Selected nodes
Selected nodes
14
Sensitivity nodes
10
14
14
12
Ulnar artery
14
16
Brachial artery
12
12
x 10
Brachial artery
x 10
5
4
6
8
Radial artery
10
12
0.5
Ulnar artery
14
14
2
6
8
10
Sensitivity nodes
12
14
6
8
10
Sensitivity nodes
12
14
4.4
14
13
x 10
6
2
Brachial artery
10
Selected nodes
Selected nodes
12
6
8
10
Sensitivity nodes
12
Radial artery
10
2
Ulnar artery
2
x 10
12
14
Selected nodes
1.5
1
12
4
6
8
2
10
12
0.5
14
2
6
8
10
Sensitivity nodes
12
Selected nodes
8
6
10
4
12
6
8
10
Sensitivity nodes
12
12
14
4
x 10
15
4
6
10
8
10
12
14
6
8
10
Sensitivity nodes
10
12
x 10
12
14
14
13
x 10
4
10
6
8
10
Sensitivity nodes
2.5
4
6
14
14
Selected nodes
12
1
14
Selected nodes
Brachial artery
x 10
14
14
6
8
10
Sensitivity nodes
12
14
4.5
15
flow decreases. Results show some back flow at node 33 relate to the pressure drop
resulting from the increased blood resistance through the AUR anastomosis (see
Figure 7 C, D).
(c). Ra Rb:
A similar behavior of pressure and flow distribution can be seen by decreasing the
diameter of nodes (4 , 5 and 6) of brachial artery. Figure (7 E, F) shows back flow
at node 6 due to the pressure drop by setting Rb = 50Ra.
(d). Z is small:
Physical activities lead to a reduction in terminal resistance (Z), which increases
the blood flow and mean cardiovascular pressure. As a result the cardiac output
will increases. These are the temporary changes which appear only when we do
some physical exercise. Another artificial reason for low terminal resistance is the
implantation of arteriovenous fistula (AVF), which is abnormal connection between
a peripheral artery and vein. The decrement in terminal resistance increases blood
pressure and flow and finally increases cardiac output. In this study we consider
Z
small terminal resistance, Zs = 10
(see Figure 7 G, H).
(e). Z is large:
In this section we consider large terminal resistance (Z) at terminal node 15 of
the ulnar artery i.e. ZL = 10Z, which appears when we have no physical activity.
Figure (7 I, J) shows that if the person is at rest, we have large value of Z, then
flow will reduce (Ohms law of hydrodynamics). Results are obtained at different
locations of the arm artery, which reveal that by increasing Z the flow will reduce
and pressure will increase near the terminal.
4.6
Finally we compare the results obtained by sensitivity analysis with those obtained
by using norms. We found that the diameter and length of vessel are most influential
parameters and that the norm computed for the wall thickness and elastic modulus
has identical values (see table 2 and 3).
16
0.8
1.5
Node 5
Node 6
Node 11
Node 32
1
Flow [ml/s]
x 10
B1
Node 33
0.6
0.4
Merging flows
0.2
100
6
200
300
400
500
Time [ms]
600
700
1.3
1.2
Node 3
2
3
1.1
1
0.8
0
800
x 10
1.5
B3
x 10
1.3
200
300
400
500
600
700
Time [ms]
Pressure at selected nodes with Ra>>Rb
Node
Node
Node
Node
1.4
Node 33
10
100
4
Pressure [Pa]
Node 5
Node 6
Node 11
Node 32
15
Flow [ml/s]
Node 5
Node 6
Node 11
Node 3
0.9
0
0
20
x 10
1.4
Pressure [Pa]
1.2
5
6
11
3
800
B4
Node 33
1.2
1.1
1
0.9
Backflow
100
20
x 10
300
400
500
Time [ms]
600
700
0.8
0
800
15
Flow [ml/s]
200
5
6
11
32
1.5
B5
x 10
300
400
500
Time [ms]
600
700
800
B6
1.3
200
1.4
Node 33
10
100
4
Pressure [Pa]
5
0
Node 3
1.2
1.1
1
0.9
Backflow
5
0
100
200
300
400
500
Time [ms]
600
700
0.8
0
800
Node
Node
Node
Node
Node
Flow [ml/s]
1.5
6
11
14
15
3
1.4
x 10
7
1.3
1
0.5
0
1.2
200
300
400
500
Time [ms]
600
700
800
Node
Node
Node
Node
Node
Pressure [Pa]
2.5
100
4
6
11
14
15
3
1.1
1
0.9
Backflow
100
6
20
x 10
Flow [ml/s]
15
200
300
400
500
Time [ms]
600
700
0.8
0
800
Node
Node
Node
Node
Node
6
11
14
15
3
100
4
1.5
x 10
1.4
1.3
Pressure [Pa]
0.5
0
10
200
300
400
500
Time [ms]
600
700
800
6
11
14
15
3
B10
1.2
1.1
1
0
0.9
Backflow
5
0
100
200
300
400
500
Time [ms]
600
700
0.8
0
800
100
200
300
400
500
Time [ms]
600
700
800
17
E
l
d
h
N0
q0
0.45
0.54
1.96
0.45
N1
q1
0.44
0.53
1.76
0.44
N2
q2
0.42
0.47
1.75
0.42
N3
q3
0.36
0.43
1.55
0.36
N4
q4
0.34
0.39
1.43
0.34
N5
q5
0.27
0.32
1.16
0.27
N6
q6
0.24
0.31
1.06
0.24
N7
q7
0.36
0.95
1.34
0.36
N8
q8
0.39
0.86
1.19
0.39
N9
q9
0.19
0.48
0.80
0.19
N10
q10
0.18
0.46
0.78
0.18
N11
q11
0.18
0.25
0.87
0.18
N12
q12
0.11
0.13
0.45
0.11
N13
q13
0.24
0.34
0.10
0.24
N14
q14
0.14
0.19
0.63
0.14
N15
q15
0.12
0.15
0.54
0.12
Table 2: 10% change in E, l, d and h at node 7 of the arm artery (see Figure 2, without anastomosis) and their corresponding
percentage change in flow at each node.
18
E
l
d
h
N0
p0
0
0
0
0
N1
p1
0.02
0.03
0.13
0.02
N2
p2
0.03
0.05
0.15
0.03
N3
p3
0.05
0.06
0.21
0.05
N4
p4
0.06
0.08
0.27
0.06
N5
p5
0.08
0.09
0.33
0.08
N6
p6
0.08
0.10
0.36
0.08
N7
p7
0.13
0.32
0.58
0.13
N8
p8
0.13
0.32
0.58
0.13
N9
p9
0.14
0.37
0.63
0.14
N10
p10
0
0
0
0
N11
p11
0.09
0.11
0.38
0.09
N12
p12
0
0
0
0
N13
p13
0.09
0.11
0.40
0.09
N14
p14
0.10
0.12
0.44
0.10
N15
p15
0
0
0
0
Table 3: 10% change in E, l, d and h at node 7 of the arm artery (see Figure 2, without anastomosis ) and their corresponding
percentage change in pressure at each node. Change in pressure with respect to any parameter is zero, because we use pressure as a
boundary condition.
Conclusion
Future work
The methods applied, give satisfactory results if the cardiovascular parameters are
independent, in the real scenarios however, they are often interdependent like e.g.
the observation of a high correlation between the extension of the elastic walls and
the tangential tension caused by transmural pressure. To study these type of effects
in a more general way, we plan to apply global sensitivity analysis to a closed loop
and open loop cardiovascular system models, which deal with variations in many
parameters at a time.
Acknowledgments
The author would like to thank the HEC/DAAD for their financial support also to
Rudolf Huttary for fruitful discussions and encouragement.
References
[1] Quarteroni, A., Ragni, S. and Veneziani, A., Coupling between lumped and
distributed models for blood flow problems. Computing and Visualization in
Science. Volume 4, pages 111-124 (2001)
[2] Milisic, V. and Quarteroni, A., Analysis of lumped parameter models for blood
flow simulations and their relation with 1D models. Mathematical Modeling
and Numerical Analysis (1999)
19
[3] Westerhof, N., Bosman, F., De Vries, C.J. and Noordergraaf, A., Analog studies
of the human systemic arterial tree. Journal of Biomechanics, Volume 2, pages
121-143 (1969)
[4] Noordergraaf, A., Verdouw, P.D. and Boom, H.B.K., The use of an analog
computer in a circulation model. Progress in Cardiovascular Diseases. Volume
5, number 5 (1963)
[5] Phillips, C., A simple lumped parameter model of the cardiovascular system.
PhD thesis, Colorado State University Fort Collins, Colorado (2011)
[6] Shim, E. B., Sah, J. Y. and Youn, C. H., Mathematical modeling of cardiovascular system dynamics using a lumped parameter method. Japanese Journal
of Physiology. Volume 54, pages 545-553 (2004)
[7] Yobing, S., Lumped parameter modeling of cardiovascular system dynamics under different healthy and diseased conditions. PhD thesis, University of Sheffield
(2013)
[8] Sherwin, S. J., Franke, V., Peiro, V., and Parker, K., One dimentional modeling
of a vascular network in space time variables. Kluwer Academic Publishers.
Printed in the Netherlands. (2003)
[9] Sato, T., Yamashiro, S.M., Vega, D. and Grodins, F.S., Parameter sensitivity analysis of a network model of systemic circulatory mechanics. Annals of
Biomedical Engineering. Volume 2, pages 289-306 (1974)
[10] Yu, Y.C., Boston, J.R., Simaan, M.A. and Antaki, J.F., Sensitivity analysis of
cardiovascular models for minimally invasive estimation of systemic vascular
parameters. Proceedings of American Control Conference, San Diago, California, (1999)
[11] Harvey, W., Exercitatio anatomica de motu cordis et sanguinis in animalibus
, Frankford, chapter 14 (1628)
[12] Weber, EH., De pulsu, resorptione, auditu et tactu. Annotationes Anatomicae
et Physiologicae, Lipsiae, (1834)
[13] Liebau, G., Die bedeutung der tragheitskrafte fur die dynamik des blutkreislaufs. Zs kreislaufforschung. Volume 4, pages 428-438 (1957)
[14] Maximilian, M., Johnnie, W. H., Graham, S. S., Thomas, K., Noordergraaf,
A.,Impedance defined flow: Generalization of William Harveys concept of the
circulation-370 year later. International Journal of Cardiovascular Medicine
and science, Volume 1, pages 205-2011 (1998)
[15] Laguy, C.A.D., Bosboom, E.M.H., Belloum, A.S.Z., Hoeks, A.P.G. and van de
Vosse, F.N., Global sensitivity analysis of a wave propagation model for arm
arteries. Medical Engineering and Physics. Volume 33, pages 1008-1016 (2011)
[16] Jager, Gerad S., Westerhof, N., Noordergraf, A., Oscillatory flow impedance
in electrical analog of arterial system: representation of sleeve effect and nonnewtonian properties of blood. Circulation research. Volume XVI, pages 121133 (1965)
20
21