Imamat hxt019

IMA Journal of Applied Mathematics (2013) 78, 729–749
doi:10.1093/imamat/hxt019
Advance Access publication on April 26, 2013
Three-dimensional effects in the lower urinary tract

Marios Tziannaros∗ , Stephen E. Glavin and Frank T. Smith
Department of Mathematics, University College London, Gower Street, London WC1E 6BT, UK
∗
Corresponding author: mtziannaros@gmail.com
Downloaded from http://imamat.oxfordjournals.org/ at The Chinese University of Hong Kong on March 25, 2016
[Received on 25 November 2012; revised on 18 March 2013; accepted on 19 March 2013]
The two major components of the human lower urinary tract, namely the bladder and the urethra, are
modelled by means of 3D theory accompanied by analysis and computations for the fluid flows evolv-
ing within the vessels. Both vessels undergo substantial changes in shape as they function. The time-
dependent bladder shape that is prescribed essentially directs fluid into the urethral opening but also
induces involved behaviour as the bladder collapses almost totally. The urethral internal shape by con-
trast interacts in full with its contained fluid motion, mainly by virtue of the fluid pressure response. Solu-
tions are presented for various configurations. A similar approach is taken for the influences of enlarged
prostate glands over a comparatively long scale, again incorporating three-dimensionality.
Keywords: bladder; urethra; fluid flow; lower urinary tract.
1. Introduction
The biomedical emphasis here is concerned with the lower urinary tract (LUT), consisting primarily
of the bladder and the urethra. The present theoretical contribution comprises modelling, analysis and
computation with a focus and hence model on 3D effects in spatial terms partly because of the practi-
cal considerations as distinct from the obvious limitations of 2D theory. The study follows pioneering
investigations by Griffiths (1971, 1980) and is based on the studies of Tziannaros (2011), Tziannaros &
Smith (2012) and Glavin (2012), which include 2D properties and provide a springboard for the inves-
tigations described in this contribution. Apart from the modelling, the aim of the analysis below is to
provide comparisons and checks and to yield extra insights, given that the aim of the computational
work is to be able to cope in a simple and reliable way with many different shapes and changes in
topology: bladder shapes for example vary enormously.
The research presented also has an industrial motivation. This is in view of an EPSRC-CASE award
for M. T. involving Astellas Pharma Europe Ltd and concerning incontinence, as follows.
Urinary incontinence, bladder instability and urethral complications including prostate problems are
of major importance. The number of patients with this problem and the potential economic impact are
staggering, as described in the following section. Other points also need an immediate remark. The
LUT contains unsteady fluid flow whose typical Reynolds number is large, suggesting inviscid theory
as a first model unless the length scales are relatively long. The two main components produce such
substantial shape changes that nonlinear theory is required, with the bladder essentially having shape-
driven motion of fluid whereas the urethra has shape-flow interaction (except perhaps when prostate-
gland effects matter considerably). Section 2 discusses the biomedical background while Sections 3–5
present modelling and study for the bladder, the urethra and prostate effects, respectively. Concluding
comments are given in Section 6. We dedicate this to Jean-Marc Vanden-Broeck, a fine colleague and
collaborator.

c The authors 2013. Published by Oxford University Press on behalf of the Institute of Mathematics and its Applications. All rights reserved.
730 M. TZIANNAROS ET AL.
2. The lower urinary tract

An excellent discussion of urology is given in Walsh (2002), further to Griffiths (1980). Problems of
the LUT are of major importance both socially and economically. It has been estimated that 17 million
people suffer from bladder control problems in the USA alone, costing $26 billion per year to manage
(Walsh, 2002). This figure suggests possible costs of £1.4 billion per year to the National Health Service
in the UK at 1990 prices (Brocklehurst et al., 1999).
The LUT comprises the bladder and the urethra and has, as its main functions, storage and mic-
turition (voiding) of urine. Injuries or diseases of the nervous system can impede voluntary control of
micturition causing bladder instability and urge incontinence (Walsh, 2002), but the latter may also be
due to intrinsic detrusor myogenic abnormalities resulting in motor detrusor instability (Brading, 1997).
Urethral dysfunction may also cause urinary incontinence. An increase in bladder outlet resistance (the
pressure required to open the urethra due to the material properties of the vessel) either caused by
enlargement of the prostate gland (hyperplasia) or due to detrusor-sphincter dyssynergia (spinal cord
injuries, multiple sclerosis) causes the bladder to remodel as it attempts to adapt. As a consequence, this
induces urge incontinence (Walsh, 2002). More complicated mixed effects may also be observed, for
example in patients with mixed stress and urge incontinence. Here stress incontinence causes leakage
of urine into the urethra, stimulating urethral effects that induce an involuntary voiding reflex (urge
incontinence) (Walsh, 2002).
The bladder body lies above the ureteral orifices, while its base consists of a smooth triangular region
within the bladder, whose vertices are formed of the two ureteral orifices and the internal urethral orifice
(this structure is known as the trigone), and the bladder neck (Walsh, 2002). The bladder outlet is formed
of the bladder base, the urethra and external urethral sphincter (a ring of circular muscle which can
constrict the urethra) (Walsh, 2002), see Figs 1–3. The male urethra is approximately 20 cm in length,
a sketch of which can be seen in Fig. 1(b) adapted from Griffiths (1980). While at rest the urethra is
closed down to a mere slit. The urethra follows a downward sloping path except at the spongiose part
where it turns upward in the bulb of the penis resulting in an abrupt angulation. Four regions can be
identified. The preprostatic part has a stellate lumen and is approximately 1–1.5 cm long. The prostatic
(a) (b) (c)
Fig. 1. Sketch of the lower urinary tract for (a) females and (b) males viewed from the front. Labels are D, detrusor smooth
muscle; T, trigone; SM, urethral smooth muscle; DS, distal intrinsic urethral sphincter; PS, periurethral sphincter; BN, bladder
neck; P, prostate gland; O, ureteral orifices; MU, membraneous urethra; PU, penile urethra; EM, external meatus; E, ejaculatory
duct; C, connective tissue. This diagram is adapted from Griffiths (1980). Sketch of a tube law (c) (pressure-area relation) for a
thin-walled elastic tube with approximate (simplified) wall shapes (Griffiths, 1980; Pedley & Luo, 1998; Sherwin, 2003).
THREE-DIMENSIONAL EFFECTS IN THE LOWER URINARY TRACT 731
(a) (b) (c) (d)
Fig. 2. Shapes of different bladders later on during micturition, taken by a Voiding Cystourethrogram (a) Radiology Malaysia, (b)
Croitoru, Gross & Barmeir (2007), (c) Radiology Malaysia, (d) Princeton University. The light patch in the middle is the bladder.
Also visible are the bones and in some instances such as (c) and (d) we can see the right ureter.
(b) (c)
(a)
(d) (e) (f)
Fig. 3. Cross-sections of a human female urethra taken from Pullan et al. (1982) illustrating the changes in shape of the urethra
from the external meatus (a) to the bladder neck (f).
part is 3–4 cm long and passes through the prostate; it is the widest and most dilatable part of the urethra
(Abrams, 2002). The membranous part follows and is the shortest and least dilatable section, has a
stellate cross-section and with the exception of the urethral orifice is the narrowest section. The final
section is the spongiose part and is contained within the corpus spongiosum penis. It is approximately
15 cm long and extends to the urethral orifice on the glans penis. It has a narrow lumen with a uniform
diameter of approximately 6 mm and forms a transverse slit.
The female urethra is between 3 and 4 cm long and runs from the internal meatus of the bladder
to the external meatus just above the vaginal opening (Abrams, 2002). See Fig. 1(a). Five sections
can be identified, each approximately one fifth of the total length. The first fifth is surrounded by the
bladder neck, the next two fifths are encircled by the external sphincter and smooth muscle, the fourth
section contains the compressor urethrae and urethrovaginal sphincter and the final section is passive
in nature, being non-muscular. Together the sphincter urethrae, compressor urethrae and urethrovaginal
sphincter form a structure called the striated sphincter which may be contracted, increasing the urethral
closure pressure and preventing leakage (Abrams, 2002). Micturition occurs when bladder pressure
exceeds urethral closure pressure, produced by contraction of the detrusor muscle within the bladder
wall along with relaxation of the striated sphincter. Smooth muscle, which is oriented longitudinally
along the urethra wall, contracts during micturition helping to open up the vessel and aiding in the
voiding of urine (Abrams, 2002). A cross-section of a female urethra may be seen in Fig. 3 taken
from Pullan et al. (1982) and helps us to illustrate the changing and non-uniform nature of the lumen
cross-section.
Regarding the aim of the study, the urethra, certainly in males, is a long winding and distensible tube
whose physical properties can also have profound influences over the efficiency of voiding. A rise in
the resistance offered by the urethra can have substantial effects on the whole LUT. Section 3 is devoted
to the bladder, motivated by the shapes in the cystourethrograms in Fig. 2, whereas Section 4 below
addresses the fluid–structure interaction associated with the urethral resistance and urinary flow.
Current urodynamic tests tend to be invasive (Abrams, 2009; Brading, 1999) and so it is helpful
to produce mathematical models that can aid in validating tests and identify underlying pathologies.
Typical urodynamic examinations involve measuring bladder pressure with a catheter within both the
abdomen and the bladder (Griffiths, 1980). The detrusor pressure at the bladder wall is estimated by sub-
tracting the abdominal pressure (which is taken via the rectum) from the bladder pressure. This of course
relies on an assumption of uniform pressures present within both the abdomen and bladder. During the
examination the bladder is slowly filled with X-ray contrast medium via a catheter while measurements
of pressure are made. Once the bladder is full the clinician will ask the patient to micturate into a device
that measures flow rates and volumes. Pressures and shapes within the bladder and urethra may be
recorded at the same time in order to examine pathologies (urethral profile profilometry Abrams, 2006).
Important quantities for a urologist are therefore pressure, volume and flow rate. It is also important
that informative noninvasive diagnostic techniques be devised in order to avoid discomfort for patients
(Blake & Abrams, 2004).
The aim is thus to help understand the physical properties of the LUT and the corresponding flow
behaviour in order to gain better insight and also to aid development of new predictive tests. On mod-
elling bladders, see Tziannaros & Smith (2012). On urethras, although the male urethra has been kept
in mind during this work with regard to a relatively long, distensible and tortuous tube, in principle,
the ideas can be applied to the female urethra also. Studied here or in Glavin (2012) are the effects of
urethral length, tortuosity and variable cross-sectional dimensions on the contained fluid flow as well
as the effects of distensible walls. The urethra is a distensible tube such that, when filled with fluid,
the cross-sectional area of the lumen depends on the pressure (Griffiths, 1980) and so a mathematical
relation is taken to exist between the fluid pressure and the cross-sectional area. A typical relation for a
uniform thin-walled elastic tube can be seen in Fig. 1(c) which holds for many such tubes, for example
large arteries (Pedley & Luo, 1998) or latex tubes. In such a relation, for a circular tube with undisturbed
radius a0 and wall thickness h, when an excess pressure within the tube pe is applied it induces a circum-
ferential tension a0 pe per unit length of the tube. This circumferential tension pulls on an area h of the
tube material per unit length, giving a circumferential tension stress (force pulling on unit area normal to
that force) of a0 pe /h. For a thin walled tube (h a) the magnitude of the circumferential stress greatly
exceeds the radial-stress magnitude. The corresponding circumferential strain is thus a0 pe /Ēh where the
elastic behaviour of the tube material under small stress is represented by modulus Ē. Typical numerical
values are discussed in Section 4, while values typical of the bladder can be found in Tziannaros &
Smith (2012).
3. Bladder modelling
The flow in the bladder is very much shape-driven rather than shape-affecting, as well as requiring non-
small alterations and being thoroughly 3D. Modelling, analysis and computations are described here.
The idea of the numerical approach in particular is to be able to cope in a simple and reliable way with
many different shapes and changes in topology since bladder shapes vary enormously. The idea of the
analysis is partly to provide comparisons and checks and partly to give extra insights.
3.1 Model
An axisymmetric situation is taken as a starting point before fully 3D properties are addressed later.
Hence, in terms of velocity components uR = −(1/R)(∂Ψ/∂z), uz = (1/R)(∂Ψ/∂R) in cylindrical coor-
dinates R, z the governing equation of concern is for the Stokes stream function Ψ ,
∂ 2Ψ 1 ∂Ψ ∂ 2Ψ
− + = 0. (3.1)
∂R2 R ∂R ∂z2
Despite the relative simplicity of (3.1) many extra features arise from the complicated boundary shapes
involved as mentioned already and found below. The Reynolds numbers for the collapsing bladder
flow have quite large values in general and as a consequence nominally the flow is effectively unsteady,
incompressible and inviscid. It is also irrotational because it starts from rest. Hence, there is an unknown
velocity potential and corresponding stream function to be found, satisfying (3.1).
The main required boundary conditions correspond to the effective sink at the urethra, the squeeze
velocity at the moving boundary and the fact that the urethra point remains fixed throughout.
3.2 Early-time or special-case analysis

A special-configuration analysis which is equivalent to an early-time analysis in general is used in order
to provide comparisons for our numerical work below. The exact solution for a sphere is formulated
here using the features of the boundary conditions described at the end of the previous subsection.
From Johnson (1998) the stream function due to a point source is
Q
Ψ =− (1 + cos θ ), (3.2)
4π
where Q is equal to the volumetric flow rate emitted from the source and θ is the horizontal angle with
respect to a point P. Once we have assigned our sources and sinks we proceed by employing Butler’s
theorem arising from the fact that a sphere is involved in the flow. The images are added in the current
flow and several geometric relationships are used in order to simplify the result. Adding up all terms
and cancelling appropriately, we obtain the solution

Q1 r Q2 a/r + cos θ
Ψ= (1 + cos θ ) − 1+ 2 2
4π a 4π (a /r + (2a/r) cos θ + 1)1/2

Q2 r r/a + cos θ
+ 1+ 2 2 + Ds , 0 < θ < π , (3.3)
4π a (r /a + (2r/a) cos θ + 1)1/2
where θ = θ1 and we have let c → a. The constant radial velocity on the sphere yields the inward squeeze
velocity as V = Q1 /4π a2 . This result makes good sense physically in terms of the surface area 4π a2 of
the sphere. Applying further simplifications and determining the downward flow stream proportional to
Ds to be Ψstream = 12 Ur2 sin2 θ, where U is the velocity component along the axis of symmetry, leaves
us with the overall analytical result for the flow in a sphere,

z+1
Ψ (R, z) = aV R +z +z +a V 1+ 2
2 2 2
(R + z2 + 2z + 1)1/2

z + R2 + z2 V
− aV R +z + 2
2 2 − R2 , (3.4)
(R + z2 + 2z + 1)1/2
2
in cylindrical polar coordinates where R = r sin θ , z = r cos θ .
3.3 Computations
The numerical method used is an iterative finite-difference method with boundary interpolation, similar
to Tziannaros & Smith’s (2012) but adjusted for axisymmetric configurations. The method is efficient
around the boundaries of the model shapes of interest which are highly distorted in general and it also
deals readily with topological changes (Bowles et al., 2008; Tziannaros, 2011; Tziannaros & Smith,
2012) suggesting application allied with accuracy and flexibility. A standard five-point form is used at
most internal points supplemented by an in–out criterion extending the idea of Bowles et al. (2008)
to refine boundary interpolation. The boundary interpolation with the iterative finite differencing han-
dles the present complex situations in terms
of the boundary condition on the tangential derivative of
the stream function ∂Ψ/∂s where ∂s = (∂R2 + ∂z2 ) is the distance element along the boundary. The
condition now only depends on the shape function F(R, z, t) which can in particular describe complex
shapes.
The form of the input boundary condition for shapes F = z − f (R, t) = 0 is, from the kinematic
condition,
∂Ψ Ft
=− 2 . (3.5)
∂s (Fz + FR2 )1/2
The modelled shapes that the vessel takes during micturition in the case of human bladder evacuation
are specified as follows and presented in Figure 4(b). The function for the collapse becomes

n
hm t2 e−8(R−km ) ,
2
Fu (R, z, t) = z − (1 − R2 )1/2 + (3.6)
m=1
where hm is the collapsing factor at the R = km position of the vessel, of the m out of n movements in
Bowles et al. (2008). This represents the upper half of the function, and we similarly have a lower one
given as

q
hp t2 e−8(R−kp ) ,
2
Fl (R, z, t) = z + (1 − R2 )1/2 − (3.7)
p=1
where hp is the collapsing factor at the R = kp position for the bladder, of the p out of q movements.
These shape structures start off as spheres at t = 0 and then evolve accordingly to emulate the human
bladder collapse. The evolving shape descriptions (3.6) and (3.7) are substituted in (3.5) to define the
complete boundary condition. Grids employed were in the range 502 to 2002 and checks on their effects
combined with the analytical comparisons below verified the accuracy of the numerical solutions.
(a) (b)
Fig. 4. (a) Set-up of the formulation for determining the flow inside a sphere (Section 3.2). (b) Shapes that the axisymmetric
vessel takes at certain times as it collapses with a cardioidal-like shape, in the model (Section 3.3).
The main computational results are presented in Figure 5 showing contour plots and velocity profiles
at different stages of the micturition process during the bladder collapse. An end-stage analysis similar
to that of Tziannaros & Smith (2012) applies as we see in Section 3.4 for the case where the bladder
becomes very thin. Flow velocities and lengths are given here relative to reference values of collapse or
squeeze velocity and bladder dimensions in turn for a typical micturition time of 17 s. Analytical checks
as in (3.4) and also further analysis from Tziannaros & Smith (2012) prove supportive.
3.4 Late-time analysis

The computational results in Figure 5 indicate that a substantial local change of the flow takes place near
the point sink (the urethra point) at late times. Mass conservation establishes the overall sink strength
entering the urethra but not the local behaviour above. In the present inviscid model (3.1) the late-time
response corresponds to the bladder virtually collapsing with vertical length scale small compared with
the total horizontal scale and so the controlling equation is merely
∂ 2Ψ
= 0, (3.8)
∂z2
to leading order. Here the scenario is analogous to the 2D of Tziannaros & Smith (2012), with the
system (3.8) therefore producing a momentum effect left over as the radial distance R tends to zero on
approach to the sink. The apparent clash which this suggests is actually smoothed out very close to the
sink inside an inner region whose radial scale is of the same order as the vertical z scale and where the
original equation (3.1) is reinstated in full. The corresponding matched asymptotic expansions proceed
in essence as in the above paper. The connection with the numerical solutions at late times shown in
figure 5(g–i) in particular is supportive.
(a) (b) (c)
(d) (e) (f)
(g) (h) (i)
Fig. 5. (a, d, g) Contour plot of the fluid motion within the modelled axisymmetric collapsing vessel during micturition. Also
plotted are the horizontal (b, e, h) and vertical (c, f, i) velocity along the horizontal centreline.
Essentially the same mechanism applies also for general 3D flow except that swirl is then likely
to be present and, in the case of a viscous fluid, the Navier–Stokes equations replace (3.1). The latter
reduce in scaled form to the 3D boundary-layer system
ux + vy + wz = 0, (3.9)
ut + uux + vuy + wuz = −Px + uzz , (3.10)
vt + uvx + vvy + wvz = −Py + vzz , (3.11)
in place of (3.8) at late times during the collapse process. The horizontal momentum balances in the sys-
tem are written in Cartesian coordinates for clarity and the pressure P is independent of z by virtue of
the vertical momentum balance. Along with (3.9–3.11) significant vorticity is present as the velocity
components must satisfy conditions of prescribed velocity at the moving vessel walls. Clearly one
extreme of the system (3.9–3.11) leads back to the form (3.8) associated with a predominantly uni-
form velocity profile in the core flow say. In broad terms however for the case (3.9–3.11) by analogy
with the above results the walls must presumably propel fluid horizontally towards the sink as at a rear
stagnation point and so flow separation may be induced. Within the setting (3.9–3.11) this is a regular
phenomenon since the pressure is unknown and so the Goldstein singularity is not encountered. In con-
trast the induced properties of left-over momentum and hence jet-like clash with separation close to the
sink can still arise even with viscous rotational effects present as Smith & Duck (1977) show.
4. Urethra modelling
Typical values of the quantities mentioned in Section 2 are
h0 = 5 × 10−4 m, Ē = 104 N m−2 , ν = 0.5, a0 = 3 × 10−5 m2 , (4.1)
where h0 and a0 are the vessel wall thickness and cross-sectional area when the tube is at rest, Ē(x) is
Young’s modulus and ν is the Poisson ratio of the vessel wall (Bykova & Regirer, 2005). The dynamic
viscosity and density of urine are both similar to that of water (Burton-Opitz & Dinegar, 1918), giv-
ing approximate values of ρ = 103 kg m−3 , μ = 10−3 kg m−1 s−1 , respectively. From Griffiths (1980),
the bladder volume is approximately 350–500 ml and typical maximum outflow is 25 ml s−1 with an
approximate urethra radius of 3 mm. A typical urethral velocity is given by u0 ≈ 1 m s−1 . In tests carried
out by Griffiths he found upper limits on flow speed to be approximately 2.1–2.2 m s−1 with a max flow
rate of 25.5(±1) ml s−1 (Griffiths, 1971). He also found (Griffiths & Rollema, 1979) that over much of
a micturition there is little variation of flow rate (being about 20–40 ml s−1 ) which is also independent
of initial volume. Also found was an opening/closing pressure of approximately 25 cm H2 O (2500 Pa).
If we now define the Reynolds number Re for the urethra as Re = ρu0 L/μ, we then find an approximate
Reynolds number of Re = 7000, see also Griffiths (1980) and Pullan et al. (1982).
Comments similar to those at the beginning of Section 3 thus apply for the urethra as well but
with the flexibility of the vessel and its interaction with the contained fluid motion clearly being very
significant now. The incident velocity at the most upstream end of the urethra model is taken to be
approximately uniform, in agreement with the sink effect within the bladder supply just ahead and the
negligible vorticity. The analytical and computational study presented below is guided by the findings
in 2D situations (Glavin, 2012; Griffiths, 1980) for certain types of wall response.
4.1 Model
The model is centred on the 3D Euler equations in non-dimensional form,
ux + vy + wz = 0, (4.2)
ut + uux + vuy + wuz = −px , (4.3)
vt + uvx + vvy + wvz = −py , (4.4)
wt + uwx + vwy + wwz = −pz , (4.5)
coupled with the pressure-shape relation
p(x, y, z, t) = f (b(x, y, z, t)). (4.6)

The functional form f is to be specified later for a particular example involving a parameter n (see (4.11))
and is based on the forms put forward by Griffiths (1971, 1980) in his original work. The unknown
functions p and b which interact throughout represent in turn the fluid pressure and the internal shape
of the vessel with b = y − a(x, z, t) say. The initial condition is that the vessel is closed, containing no
fluid, prior to the end x = 0 being opened at time zero. A moving front then develops in x > 0 whose
position is to be determined subject to a being zero at the front. The boundary conditions are kinematic
conditions at the unknown moving walls as in (4.12) and (4.13). In contrast with the previous section the
coordinate x here is measured from the start of the relatively long thin vessel and we take the opening
shape as given with velocity u there unknown but we are aware that there is likely an adjustment to
suppress jump effects near the opening of the vessel of the kind found in Smith et al. (2003).
4.2 Large-n analysis

This is based on 3D thin-region flow, appropriate in particular to n large (n is defined essentially in
(4.11)), which is equivalent to a fast-response wall. We assume a thin flow where (x, y, z) ∼ (1,,1)
along with the corresponding velocity components (u, v, w) ∼ (1, , 1), i.e. we have height and velocity
components laterally much smaller than in the cross-plane, for parameter values 1. In actuality, we
take ∼ n−1 . A schematic diagram is shown in Fig. 6. Thus, from (4.2)–(4.5) the governing equations
are
ux + vy + wz = 0, (4.7)
ut + uux + vuy + wuz = −px , (4.8)
0 = −py , (4.9)
wt + uwx + vwy + wwz = −pz , (4.10)
which we notice are of the 3D interactive inviscid boundary-layer type (Smith, 1991); they are equiva-
lent to the 3D shallow-water system (Vanden-Broeck, 2010) or that of 3D internal flows (Smith, 1976).
Coupled to this is the representative wall law
p(x, z, t) = p0 + β(x, z)(a(x, z, t)n − an0 ) (4.11)
along with a kinematic boundary condition on the unknown vessel surface y = a(x, z, t), impermeability
condition on the flat bottom surface y = 0, so that
v = at + uax + waz on y = a(x, z, t), (4.12)

v = 0 on y = 0, (4.13)
Fig. 6. A schematic diagram of the fast-response configuration. Here x, z are of order 1 with y being of order 1, where the
small parameter ∼ n−1 for large-n values.
and periodic boundary conditions on z = −D, D taken as the edges of the system in the z-direction, for
a constant D. In (4.11), the constants p0 , a0 (> 0) are given while the coefficient β(x, z) is positive.
With attention on large values of the parameter n, we introduce the variable E = an which we substi-
tute into the governing system (4.2)–(4.4). Taking n 1, the kinematic boundary condition (4.12) then
forces the ordering v ∼ n−1 while u and w remain order 1. We introduce the corresponding expansions
for u, v, w, E and p and substitute in our system (4.2)–(4.5), to obtain the leading order
u0x + w0z = 0, (4.14)
u0t + u0 u0x + w0 u0z = −p̄0x , (4.15)
w0t + u0 w0x + w0 w0z = −p̄0z , (4.16)
with p̄0 = p̄0 (x, z, t) = p0 + βE0 . Remarkably these are the 2D Euler equations acting in the plane of the
x–z surface. We now absorb the constant p0 into p̄0 for convenience, leaving
p̄0 = βE0 , (4.17)
as the pressure–shape relation.

Next we consider small values of β within the confines of system (4.14)–(4.17). We require here that
n−1 < β 1 in order not to violate the expansions for u, v, w, E and p. Then the velocities and pressure
can be expected to be small, (u, w, p) ∼ β, so as to balance the pressure terms. The system in such a
case reduces still further from (4.14)–(4.16) to
u0x + w0z = 0, (4.18)

u0t = −p̄0x , (4.19)
w0t = −p̄0z , (4.20)
along with (4.17). These are incidentally the equations of linearized unsteady potential flow.
Differentiating (4.19) with respect to x and (4.20) with respect to z and then adding the results
together yields, by use of (4.18), Laplace’s equation for p̄0 in the x–z plane,
∇ 2 p̄0 = 0. (4.21)
The boundary conditions are then the input condition:
p̄0 = g(z, t) at x = 0, (4.22)
for g(z, t) a known function found via (4.17) from the given opening E(0, z, t) ≈ E0 (z, t); periodicity on
the spanwise edges such that
p̄0 (x, −D, t) = p̄0 (x, D, t), p̄0z (x, −D, t) = p̄0z (x, D, t); (4.23)
and two constraints at the unknown moving front x = xf (z, t). The first of these two constraints,
p̄0 = 0 at x = xf (z, t), (4.24)
is derived from the requirement that E0 ∼ E(xf , z, t) = 0, a consequence of the definition of the front
as marking the moving position where the gap closes. The second constraint is a normal derivative
condition, akin to a kinematic boundary condition as in (4.12), given by
− Sx + Sz xfz = xft , (4.25)

t
where we have defined p̄0 = St for convenience. Here S = 0 p̄0 (x, z, t∗ ) dt∗ with p̄0 (x, z, 0) = 0 so that
the vessel is initially closed, by virtue of (4.17). Equations (4.19) and (4.20) therefore give that Sx = −u0
and Sz = −w0 . The unknown function −S(x, z, t) is clearly equivalent to a velocity potential, in keeping
with the original governing equations (4.18–4.20). The task here then is to solve (4.21–4.25).
4.3 Computations
The specific case of an opening shape given with non-uniform z-dependence such that
g ∝ q2 (t) + a(t) cos(z) (4.26)
is addressed now, with q2 (t) and a(t) depending on time only. This admits two exact solutions of much
interest. One exact solution contains a term proportional to exp(−x) and is defined by
πz
p̄0 (x, z, t) = xq1 (t) + q2 (t) + a(t) e−x cos , (4.27)
D
where q1 (t), q2 (t) and a(t) are functions of t to be found. This solution satisfies Laplace’s equation
(4.21) and the periodicity conditions (4.23) and predicts the front position for x to be

b̄ q̄tn+1
xf (0, t) = ln −1 + k1 exp . (4.28)
q̄ n+1
If, at t = 0, the front starts at x = 0 then we have the constant k1 = (1 + q̄/b̄).

Similarly, we may examine at z = π which gives

b̄ q̄tn+1
xf (π , t) = ln 1 + k2 exp , (4.29)
q̄ n+1
with the constant k2 = (−1 + q̄/b̄) if xf = 0 at t = 0. In order for both solutions xf to remain non-
negative for physical sense we require |b̄| < q̄. If we now examine t → ∞ we have that xf (0, t) →
q̄tn+1 /(n + 1): the solution is then approaching the 2D state. Comparisons of the 2D front and movement
of the 3D front along z = 0 can be seen in Fig. 7 where firstly we see a comparison for large times
showing the close agreement between the 2D and 3D results and secondly there is shown more clearly
the differences between the 2D and 3D results for front position at smaller times.
We now consider general z ∈ (0, π ) spanning across (around) the wall. The characteristics of the
equation are defined by
dz
= b̄tn e−xf sin(z), (4.30)
dt
and along the curves we have
dxf
= q̄tn + b̄tn e−xf cos(z). (4.31)
dt
(a) (b)
Fig. 7. For large-n theory in 3D, a comparison between front position xf as given by Equation (4.28) (solid line, for the 3D
evolution along z = 0) and (dotted line) for the 2D evolution, showing close agreement for (a) large values of t and (b) the clear
differences for order-1 values of t. Plots are for n = 1, 2, 3, 4. Here q̄ = 1, b̄ = 0.5.
Combining (4.31) and (4.30) gives the result that
dxf q̄ exp(xf )
= + cot(z), (4.32)
dz b̄ sin(z)
along characteristics. The solution to (4.32) for xf (z) on the characteristics is found by considering xf
as the sum of two functions of z such that xf = F(z) + G(z), say. Letting
dG
= cot(z) so that G = log | sin(z)|, (4.33)
dz
without loss of generality, from (4.32), it then remains to solve

dF q̄ exp(F)| sin(z)| q̄
= ; hence F = − log (z0 − z) , (4.34)
dz b̄ sin(z) b̄
for sin(z) > 0 as in our setting, with z0 a constant to be found. Therefore,
b̄ sin(z)
xf = log , (4.35)
q̄(z0 − z)
along the characteristics. If this is substituted into (4.30) we can then find characteristics given by z(t)
and the corresponding xf values. We have

dz q̄(z0 − z)
= b̄tn sin(z), (4.36)
dt b̄ sin(z)
and after some work we obtain the front solution

(a) (b) (c) (d)
(e) (f) (g) (h)
Fig. 8. 3D front position (a–b) plotted onto a tube with a normalized resting radius, for the case of (4.27). Here n = 10, b̄ = 0.9,
q̄ = 1. Opening pressure in the 3D case (c) of (4.27). Here n = 10, b̄ = 0.9, q̄ = 1. Dashed lines in (d) are the 2D solution, showing
the agreement between the 3D and 2D solution for larger times. 3D front position plotted onto a tube (e)-(f) with a normalized
resting radius, for the case of (4.39). Here n = 10, b̄ = 0.9, q̄ = 1 with t∞ given by (4.42). Opening pressure in the 3D case (g)
of (4.39). Here n = 10, b̄ = 0.9, q̄ = 1. We notice that on approach to t = t∞ the opening pressure tends to infinity (h). Here t∞ is
given by (4.42).
q̄tn+1 sin(zs + (b̄ sin(zs )/q̄)(1 − exp(−q̄tn+1 /n + 1)))

xf (t; zs ) = + log , (4.37)
n+1 sin(zs )
for 0 < z < π . From this we can see that, as in the z = 0 and π cases, as t → ∞, xf → q̄tn+1 /(n + 1).
The corresponding opening pressure function g(z, t) can now be found to be
g(z, t) = ntn−1 (q̄xf (z, t) + (1 − exp(−xf (z, t)))b̄ cos(z)). (4.38)
Again as a check, now letting t → ∞ leads us to the solution g(z, t) → nq̄2 t2n /(n + 1) for 0 z π .
Plots of front position and opening function can be seen in Figs 8(a,b,e,f).
A similar approach can be carried out with an alternative specific exact solution given by
πz
p̄0 (x, z, t) = xq1 (t) + q2 (t) + a(t) ex cos , (4.39)
D
and this exhibits different interesting behaviour. Performing a similar analysis leads to
q̄
xf (0, t) = log , (4.40)
b̄(1 + (−1 + q̄/b̄) exp{q̄tn+1 /(n + 1)})
q̄
xf (π , t) = log n+1 . (4.41)
b̄(−1 + (1 + q̄/b̄) exp{ q̄tn+1 })
As t → ∞ both solutions tend to a constant value log(q̄/b̄). However, it is evident that a finite-time
singularity occurs along the z = π position with the time to this singularity given by
1/(n+1)
n+1 q̄
t∞ = log 1 + . (4.42)
q̄ b̄
It is found (Glavin, 2012) that the opening-pressure function also exhibits irregular behaviour then and
so the solution is only valid for 0 t < t∞ at most. Analysis gives the front position
q̄tn+1 sin(zs − (b̄/q̄) sin(zs )(1 − exp{q̄tn+1 /(n + 1)}))
xf (t; zs ) = − log , (4.43)
n+1 sin(zs )
along curves

b̄ q̄tn+1
z(t; zs ) = zs − sin(zs ) 1 − exp . (4.44)
q̄ n+1
The general solution of (4.21)–(4.25) would have the single term in exp(−x) in (4.27) replaced by an
infinite series of terms containing exp(−mx) for integers m 1. It is also clear that the analysis here
applies to any shape of cross-section in general.
5. Prostate modelling
Following on from the 3D work of the previous section is another important case which is nonlinear and
allows for viscous effects in a steady-flow setting. We begin by taking a long slender rigid tube with a
smooth constriction formed by a relatively local collapse of the tube walls. The solution method adopted
follows that of Bowles et al. (2008) for tube branching; however in our case the constriction is not severe
enough to form two fully developed daughter tubes but only a local deformation in which the vessel
shape returns to the undisturbed cross-section downstream. The 3D effects of the constriction on the
flow pressure and velocities are to be found, where far upstream a unidirectional axial flow is imposed.
The present investigation, which is numerical and analytical, includes examining flow properties close
to the point of greatest constriction.
5.1 Governing equations

These are for a slender vessel with axial length scale L∗ along with radial scale R∗ L∗ , so that
R∗
α= 1
L∗
is a small aspect ratio, and Reynolds number Re = U ∗ L∗ ρ ∗ /μ∗ for U ∗ a representative axial flow veloc-
ity of the upstream laminar flow profile, ρ ∗ the fluid density and μ∗ the fluid viscosity. Then
ux + vy + wz = 0, (5.1)
uux + vuy + wuz = −px + R̄(uyy + uzz ), (5.2)
uvx + vvy + wvz = −Py + R̄(vyy + vzz ), (5.3)
uwx + vwy + wwz = −Pz + R̄(wyy + wzz ), (5.4)
are the non-dimensional governing equations where R̄ = 1/Re α 2 is an order-1 parameter so that for
fixed α a flow rate is determined. For a discussion on lubrication dominated (Re α −2 ) and inviscid
(1 α −2 Re) flow regimes see Bowles et al. (2008). The pressure is p(x) + α 2 P(x, y, z) + · · · here.
5.2 Analysis close to the point of minimum gap

There are two flow regions to consider close to the point of greatest constriction (or minimum gap or
carina), which is given by the point (x, y, z) = (xc , yc , zc ) say. One is the core flow in which we take
x − xc small with y − yc and z − zc both of order 1, and the second is the region very close to the carina
where y − yc and z − zc are then also small. The particular case considered herein will take a wall
that comes in quadratically both in the longitudinal- and cross-planes so that |z − zc | ∼ |y − yc |2 with
|z − zc | ∼ |x − xc |2 and so |y − yc | ∼ |x − xc |. The powers are a result of the known wall shape.
In the core flow region (firstly) we have x − xc = ˆ X say locally, where ˆ is a small positive param-
eter. Here (X , y, z) are all order-1 variables. We expect a regular expansion in the axial direction with
the pressure perturbation being mainly linear in x − xc : thus,
(u, v, w) = [u0 , v0 , w0 ](y, z) + ˆ X [u1 , v1 , w1 ](y, z) + · · · , (5.5)

(p, P) = (p0 , P0 (y, z)) + ˆ X (p1 , P1 (y, z)) + · · · . (5.6)
Substituting (5.5), (5.6) into (5.1–5.4) gives the 3D core-flow equations to leading order as
u1 + v0y + w0z = 0, (5.7)

u0 u1 + v0 u0y + w0 u0z = −p1 + R̄(u0yy + u0zz ), (5.8)
u0 v1 + v0 v0y + w0 v0z = −P0y + R̄(v0yy + v0zz ), (5.9)
u0 w1 + v0 w0y + w0 w0z = −P0z + R̄(w0yy + w0zz ). (5.10)
The velocity profile u0 and pressure constants p0 and p1 are known in effect from the upstream flow.
Near the saddle point of greatest constriction (secondly) the local smooth shape is assumed given
at the upper wall z = z+ (x, y) such that z+ − zc = A(x − xc )2 + B(y − yc )2 + γ for some constants
A, B, γ > 0. The lower wall is similarly given by z = z− (x, y) with z− − zc = −A(x − xc )2 − B(y −
yc )2 − γ . Hence, we have in this region an axial scaling as in the core section along with y − yc = ˆ Y
and z − zc = ˆ 2 Z. The cross-plane variables X and Y are of order 1. We expect p to scale as in the core
region, in order to match. Then continuity and momentum balances provide the following expansions:
(u, v, w) = [ˆ 4 U, ˆ 4 V , ˆ 5 W ](X , Y , Z) + · · · , (5.11)

(p, P) = (p0 , Q0 ) + (Xp
ˆ 1 , Q1 (X , Y , Z)) + · · · , (5.12)
with the X -dependence being more subtle here than in the core, as indicated. Here p0 and Q0 are con-
stants. Substitution into the system (5.1–5.4) then yields the lubrication equations
UX + VY + WZ = 0, (5.13)
0 = −p1 + R̄UZZ , (5.14)
0 = −Q1Y + R̄VZZ , (5.15)
0 = −Q1Z . (5.16)
These are subject to no-slip on the walls which are given by
Z+ = z+ − zc = AX 2 + BY 2 + 1, Z− = z− − zc = −AX 2 − BY 2 − 1, (5.17)
√
where ˆ is the minimum vertical half-gap width γ . In general ˆ represents a vertical scale.
Integrating (5.14) and applying the boundary conditions yields the axial velocity
p1 2
U(X , Y , Z) = − (Z − Z 2 ), (5.18)
2R̄ +
where we have used the fact that Z− = −Z+ . The wall shear stress on the upper wall is then given by
p1 2
UZ (X , Y , Z+ ) = (X + Y 2 + 1). (5.19)
R̄
Next, from (5.16) Q1 is independent of Z and so (5.15) gives the balance
Q1Y 2
V (X , Y , Z) = − (Z − Z 2 ). (5.20)
2R̄ +
Finally, we integrate (5.13) over the gap width, substituting in (5.19) and (5.20) and applying the no-slip
condition on the top wall, Z+ , to find W such that
∂(Z+2 ) ∂(Z+2 )
2R̄W (X , Y , Z) = p1 (Z − Z+ ) + Q1Y (Z − Z+ )
∂X ∂Y
(Z+3 − Z 3 )
+ Q1YY + Z+2 (Z − Z+ ) , (5.21)
3
where we have again used symmetry for simplification.

If we now apply the lower wall no-slip condition and introduce the gap width H(X , Y ) = Z+ − Z− =
2Z+ , then (5.21) becomes
∂H 3 ∂H 3
0 = p1 + Q1Y + Q1YY H 3 , (5.22)
∂X ∂X
which upon rearranging and integrating becomes a form of the Reynolds lubrication equation

∂(H 3 )
Q1Y H 3 = −p1 dY , (5.23)
∂X
essentially allowing the pressure correction Q1 (X , Y ) to be found given the known gap width H(X , Y )
and the pressure constant p1 .
5.3 Computational solutions

Numerical results are obtained using the method outlined in our Section 3 along with Bowles et al.
(2008) where a detailed description can be found. The system (5.1–5.4) is parabolic as long as the axial
velocity u remains positive and so there is no upstream influence. It is therefore possible to apply a
forward marching scheme in the axial direction, starting at some point upstream and continuing through
the carina. We apply boundary conditions of no slip on all walls and an initial velocity profile at the
starting point upstream. Mass flux is fixed with the flow far upstream taken to be fully developed parallel
(a) (b) (c) (d)
Fig. 9. Widening vessel exhibiting a symmetric constriction that briefly results in two daughter vessels, with the point of greatest
constriction taken to be x = 1. Here the grid step-sizes are dy = dz = 2.67−2 and dx = 5 × 10−4 . The solid line represents pressure
and the dashed the axial velocity.
flow. As the scheme marches forward towards the carina, a pinching of the vessel wall occurs. This is
handled numerically by modification of the grid at each x-station. See Tadjfar & Smith (2004).
Numerical results in Figs 9–10 are for two main cases. The first has vessels that are symmetric
about the plane given by Z = 0 and the second has non-symmetry where only the upper wall is allowed
to constrict, with the lower wall remaining circular and rigid.
For symmetric cases results were calculated for a widening vessel (Fig. 9), a severe constriction of
a non-tapering vessel (Fig. 10)(a–b) and a mild constriction of a non-tapering vessel (Fig. 10)(c–d). The
velocities along the midline and pressures were found in each case showing qualitatively similar results
with a pressure drop observed. In Fig. 10(i) the zero midline velocity is as a result of the constriction
completely obscuring that section of vessel. The equation used for the vessel wall is
2 1/2
Y hF(X )
Z+ = (1 + rF(X )) 1 − −
1 + rF(X ) 1 + 8Y 2
⎧ 2
⎪
⎪ X
⎪
⎨ for X X2 ,
X2
where F(X ) = (5.24)
⎪
⎪ 2X − X2
⎪
⎩ for X > X2 ,
X2
and Z+ = −Z− , with h being a parameter to be set and X2 = 1. In the expanding vessel r = 1.2 whereas
in the non-tapering vessels r = 0.
Non-symmetric cases were tackled with the aim of being closer to the reality of an enlarged prostate
whereby the constriction tends to intrude on one side of the vessel. Three cases were computed with
varying constriction (Figs 10(e–h)). Again, qualitatively similar results arise as the vessel provokes a
pressure drop through the constriction. The non-symmetric case has walls
hF(X )f (X )
Z+ = {1 − Y 2 }1/2 − , (5.25)
1 + 8Y 2
⎧ 2
⎪
⎪ X
⎨ for X X2 , AL
F(X ) = X2 , f (X ) = , (5.26)
⎪
⎪ 2X − X 2 1 + (X − X2 )2 /e2L
⎩ for X > X2
X2
(a) (b) (c) (d)
(e) (f) (g) (h)
(i) (j)
Fig. 10. Symmetric (a)–(d) and non-symmetric (e)–(h) constriction cases. Comparison of pressure and axial velocity along the
mid-line for the symmetric (i) and non-symmetric cases (j) y = z = 0 (for (i) the pressure for both symmetric cases is represented
by the lines starting at zero pressure (right axis) - top two lines from the left - and the axial velocity is represented by the other
two and similarly for the plot of non-symmetric case at (j)).
√
and Z− = 1 − Y 2 . Here eL = X2 /5 and X2 = 1. The strength of the 3D constriction is controlled by the
parameter AL . Fairly extensive grid tests were carried out in order to check for grid effects. There is also
agreement as in Bowles et al. (2008) with the local analysis near points of maximum constriction.
6. Other aspects and discussion

Space considerations rather require us to keep this short and (so perhaps) in the form of a list as
follows.
• Other aspects include weakly 3D effects in Glavin (2012) for the urethra.
• It would be helpful to extend the prostate-model work in Section 5 to encompass unsteady and
flexible-wall effects. Both Sections 4 and 5 apply for any cross-sectional shapes in principle.
• There is a need to consider more the nominally small influences of viscosity throughout the LUT
apart from those incorporated in Section 5. For example, lubrication or boundary-layer balances
may partly control the end-stage dynamics of the bladder addressed in Section 3.
• A major thrust is the accent on 3D properties, even though Sections 3–5 may represent somewhat
disparate elements so far and there is a wish to join the bladder and urethra models together.
• This type of study is really only just at its beginning and therefore perhaps is simple-minded. As
far as we know the present work in 3D taken together with that in Tziannaros (2011), Tziannaros &
Smith (2012), Glavin (2012) is the first to bring a combined analytical and computational effort to
bear on the LUT, although there have been a fair number of experimental studies and measurements.
More investigations along these lines may follow.
• The 3D regime here appears to produce a model and results which are fairly sensible physically
and of some potential as regards the medical context. The predictions for internal velocities and
pressures for instance are valuable in principle.
• Additional interesting points driven by the LUT application are on the interference caused by intro-
duction of a catheter to measure pressure, the emerging flow at the exit from the urethra and the
break-up of jets which is significant in some urodynamical tests.
Acknowledgement
The interest shown by the UCL Medical Modelling Group and Professor Alan Cottenden especially is
acknowledged gratefully.
Funding
This work was supported by EPSRC, Astellas Pharma and UCL.
References
Abrams, P. (2002) Incontinence: 2nd International Consultation on Incontinence, Paris July 1–3, 2001. Plymbridge
Distributors Ltd.
Abrams, P. (2006) Urodynamics. Berlin: Springer.
Abrams, P. (2009) Incontinence: 4th International Consultation on Incontinence, Paris, July 5–8, 2008. Health
Publications Ltd.
Blake, C. & Abrams, P. (2004) Noninvasive techniques for the measurement of isovolumetric bladder pressure.
J. Urol., 171, 12–19.
Bowles, R. I., Ovenden, N. C. & Smith, F. T. (2008) Multi-branching three-dimensional flow with substantial
changes in vessel shapes. J. Fluid Mech., 614, 329–354.
Brading, A. F. (1997) A myogenic basis for the overactive bladder. Urology, 50, 57–67.
Brading, A. F. (1999) .The physiology of the mammalian urinary outflow tract. Exp. Phys., 84, 215.
Brocklehurst, J., Amess, M., Goldacre, M., Mason, A., Wilkinson, E., Eastwood, A. & Coles, J. (1999)
Health Outcome Indicators: Urinary Incontinence. Report of a working group to the Department of Health.
Oxford: National Centre for Health Outcomes Development.
Burton-Opitz, R. & Dinegar, R. (1918) The viscosity of urine. Am. J. Phys.-Leg. Con., 47, 220.
Bykova, A. A. & Regirer, S. A. (2005) Mathematical models in urinary system mechanics (review). Fluid Dyn.,
40, 1–19.
Croitoru, S., Gross, M. & Barmeir, E. (2007) Duplicated ectopic ureter with vaginal insertion: 3D CT urography
with IV and percutaneous contrast administration. Am. J. Roent., 189, w272–w274.
Glavin, S. (2012) Mathematical modelling of urethral and other flows. Ph.D Thesis, UCL.
Griffiths, D. J. (1971) Hydrodynamics of male micturition—II measurements of stream parameters and urethral
elasticity. Med. Biol. Eng. Comp., 9, 589–596.
Griffiths, D. J. (1980) Urodynamics—The Mechanics and Hydrodynamics of the Lower Urinary Tract. Hilger.
Griffiths, D. J. & Rollema, H. J. (1979) Urine flow curves of healthy males: a mathematical model of bladder
and urethal function during micturition. Med. Biol. Eng. Comp., 17, 291–300.
Johnson, R. W. (1998) Handbook of Fluid Dynamics. Boca Raton, FL: Springer.
Pedley, T. J. & Luo, X. Y. (1998) Modelling flow and oscillations in collapsible tubes. Theor. Comp. Fluid Dyn.,
10, 277–294.
Princeton University, Images and Scans.
Pullan, B. R., Phillips, J. I. & Hickey, D. S. (1982) Urethral lumen cross-sectional shape: its radiological deter-
mination and relationship to function. Brit. J. Urol., 54, 399–407.
Radiology Malaysia. The Official Homepage of the College of Radiology, Academy of Medicine.
Sherwin, S. J. (2003) One-dimensional modelling of a vascular network in space-time variables. J. Eng. Math.,
47, 217–250.
Smith, F. T. (1976) On entry-flow effects in bifurcating, blocked or constricted tubes, J. Fluid Mech., 78, 709–736.
Smith, F. T. (1991) Steady and unsteady 3D interactive boundary layers, Comp. Fluids, 20, 243–268.
Smith, F. T. & Duck, P. W. (1977) Separation of jets or thermal boundary layers from a wall, Quart. J. Mech. Appl.
Maths, 30, 143–156.
Smith, F. T., Ovenden, N. C., Franke, P. & Doorly, D. J. (2003) What happens to pressure when a flow enters a
side branch? J. Fluid Mech., 479, 231–258.
Tadjfar, M. & Smith, F. T. (2004) Direct simulations and modelling of basic three-dimensional bifurcating.
J. Fluid Mech., 519, 1–32.
Tziannaros, M. (2011) Modelling bladder-collapse flow. Ph.D thesis, University College London.
Tziannaros, M. & Smith, F. T. (2012) Numerical and analytical study of bladder-collapse flow. Int. J. Differential
Equations (doi:10.1155/2012/453467)
Vanden-Broeck, J. M. (2010) Gravity-Capillary Free-Surface Flows. Cambridge: Cambridge University Press.
Walsh, P. C. (2002) Campbell’s Urology. Philadelphia, PA: Saunders.

Imamat hxt019

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

Imamat hxt019

Uploaded by

Copyright:

Available Formats

IMA Journal of Applied Mathematics (2013) 78, 729–749

Three-dimensional effects in the lower urinary tract

Keywords: bladder; urethra; fluid flow; lower urinary tract.

2. The lower urinary tract

(a) (b) (c)

(a) (b) (c) (d)

(d) (e) (f)

3.2 Early-time or special-case analysis

in cylindrical polar coordinates where R = r sin θ , z = r cos θ .

3.4 Late-time analysis

(a) (b) (c)

(g) (h) (i)

h0 = 5 × 10−4 m, Ē = 104 N m−2 , ν = 0.5, a0 = 3 × 10−5 m2 , (4.1)

coupled with the pressure-shape relation

p(x, y, z, t) = f (b(x, y, z, t)). (4.6)

4.2 Large-n analysis

p(x, z, t) = p0 + β(x, z)(a(x, z, t)n − an0 ) (4.11)

v = at + uax + waz on y = a(x, z, t), (4.12)

u0x + w0z = 0, (4.14)

p̄0 = βE0 , (4.17)

as the pressure–shape relation.

u0x + w0z = 0, (4.18)

The boundary conditions are then the input condition:

p̄0 = g(z, t) at x = 0, (4.22)

p̄0 = 0 at x = xf (z, t), (4.24)

condition, akin to a kinematic boundary condition as in (4.12), given by

− Sx + Sz xfz = xft , (4.25)

g ∝ q2 (t) + a(t) cos(z) (4.26)

If, at t = 0, the front starts at x = 0 then we have the constant k1 = (1 + q̄/b̄).

Combining (4.31) and (4.30) gives the result that

without loss of generality, from (4.32), it then remains to solve

for sin(z) > 0 as in our setting, with z0 a constant to be found. Therefore,

and after some work we obtain the front solution

(a) (b) (c) (d)

q̄tn+1 sin(zs + (b̄ sin(zs )/q̄)(1 − exp(−q̄tn+1 /n + 1)))

g(z, t) = ntn−1 (q̄xf (z, t) + (1 − exp(−xf (z, t)))b̄ cos(z)). (4.38)

5.1 Governing equations

5.2 Analysis close to the point of minimum gap

(u, v, w) = [u0 , v0 , w0 ](y, z) + ˆ X [u1 , v1 , w1 ](y, z) + · · · , (5.5)

u1 + v0y + w0z = 0, (5.7)

(u, v, w) = [ˆ 4 U, ˆ 4 V , ˆ 5 W ](X , Y , Z) + · · · , (5.11)

These are subject to no-slip on the walls which are given by

where we have again used symmetry for simplification.

5.3 Computational solutions

(a) (b) (c) (d)

(a) (b) (c) (d)

6. Other aspects and discussion

You might also like

(u, v, w) = [u0 , v0 , w0 ](y, z) + ˆ X [u1 , v1 , w1 ](y, z) + · · · , (5.5)

(u, v, w) = [ˆ 4 U, ˆ 4 V , ˆ 5 W ](X , Y , Z) + · · · , (5.11)