Mathematical Modelling in Real Life Problems

Case Studies from ECMI Modelling Weeks Ewald

Lindner pdf download
https://textbookfull.com/product/mathematical-modelling-in-real-life-problems-case-studies-from-
ecmi-modelling-weeks-ewald-lindner/

★★★★★ 4.7/5.0 (20 reviews) ✓ 132 downloads ■ TOP RATED

"Amazing book, clear text and perfect formatting!" - John R.

DOWNLOAD EBOOK
 Mathematical Modelling in Real Life Problems Case Studies
from ECMI Modelling Weeks Ewald Lindner pdf download

TEXTBOOK EBOOK TEXTBOOK FULL

Available Formats

■ PDF eBook Study Guide TextBook

EXCLUSIVE 2025 EDUCATIONAL COLLECTION - LIMITED TIME

INSTANT DOWNLOAD VIEW LIBRARY

Collection Highlights

Mathematical Problems in Meteorological Modelling 1st

Edition András Bátkai

Mathematical Modelling in Solid Mechanics 1st Edition

Francesco Dell'Isola

Mathematical Modelling 1st Edition Seppo Pohjolainen

(Eds.)

Mathematical Modelling of Meteoroid Streams Galina O.

Ryabova
Modelling in Mathematical Programming: Methodology and
Techniques José Manuel García Sánchez

Mathematical Modelling Education and Sense making Gloria

Ann Stillman

Mathematical Modelling and Analysis of Infectious Diseases

Khalid Hattaf

Cognitive Informatics, Computer Modelling, and Cognitive

Science: Theory, Case Studies, and Applications: Volume 1:
Theory, Case Studies, and Applications 1st Edition Ganesh
R. Sinha (Editor)

Mathematical Methods and Modelling in Applied Sciences 1st

Edition Mehmet Zeki Sarikaya (Editor)
Mathematics in Industry 33
The European Consortium for Mathematics in Industry

Ewald Lindner
Alessandra Micheletti
Cláudia Nunes Editors

Mathematical
Modelling
in Real Life
Problems
Case Studies from ECMI-Modelling
Weeks
Mathematics in Industry

The European Consortium for Mathematics in Industry

Volume 33

Managing Editor
Michael Günther, University of Wuppertal, Wuppertal, Germany

Series Editors
Luis L. Bonilla, University Carlos III Madrid, Escuela, Leganes, Spain
Otmar Scherzer, University of Vienna, Vienna, Austria
Wil Schilders, Eindhoven University of Technology, Eindhoven, The Netherlands
The ECMI subseries of the Mathematics in Industry series is a project of The
European Consortium for Mathematics in Industry. Mathematics in Industry
focuses on the research and educational aspects of mathematics used in industry and
other business enterprises. Books for Mathematics in Industry are in the following
categories: research monographs, problem-oriented multi-author collections,
textbooks with a problem-oriented approach, conference proceedings. Relevance to
the actual practical use of mathematics in industry is the distinguishing feature of
the books in the Mathematics in Industry series.

More information about this subseries at http://www.springer.com/series/4651

Ewald Lindner • Alessandra Micheletti •
Cláudia Nunes
Editors

Mathematical Modelling
in Real Life Problems
Case Studies from ECMI-Modelling Weeks
Editors
Ewald Lindner Alessandra Micheletti
Institute of Computational Mathematics Department of Environmental Sciences and
Johannes Kepler University of Linz Policy and Data Science Research Center
Linz, Austria Milano
Università degli Studi di Milano
Milano, Italy

Cláudia Nunes
Center for Computational and Stochastic
Mathematics
Instituto Superior Técnico (IST),
Universidade de Lisboa
Lisboa, Portugal

ISSN 1612-3956 ISSN 2198-3283 (electronic)

Mathematics in Industry
The European Consortium for Mathematics in Industry
ISBN 978-3-030-50387-1 ISBN 978-3-030-50388-8 (eBook)
https://doi.org/10.1007/978-3-030-50388-8

Mathematics Subject Classification: 97Mxx, 97M10, 00A69

© Springer Nature Switzerland AG 2020

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of
the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation,
broadcasting, reproduction on microfilms or in any other physical way, and transmission or information
storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology
now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors, and the editors are safe to assume that the advice and information in this book
are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or
the editors give a warranty, expressed or implied, with respect to the material contained herein or for any
errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional
claims in published maps and institutional affiliations.

This Springer imprint is published by the registered company Springer Nature Switzerland AG.
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Preface

In recent years, numerous reports and studies have demonstrated that Mathemat-
ics is an essential tool to improve industrial innovation, and mixed academic–
industrial consortia and networks, like ECMI (European Consortium for Mathe-
matics in Industry—http://ecmi-indmath.org) and MI-Net (Mathematics in Industry
Network—https://mi-network.org/) are working to foster the recognition of Mathe-
matics as an enabling technology.
In this framework, an increasing need of mathematicians trained to work in
an industrial environment has been observed and has pushed the academic world
to provide novel training formats, able to respond to industrial needs. ECMI
in particular has established an educational programme, offered by the ECMI
Educational Centres, which is aimed to provide such training. Its main ingredients
are mathematical modelling activities, in particular the International Modelling
Weeks and the modelling seminars.
The International Modelling Weeks format is an international workshop where
master students and/or early-career investigators (PhD students and postdocs)
receive hands-on training in problem-solving, teamwork, and in learning to exploit
their different skills to model efficiently non-mathematical problems. During
modelling Weeks are training workshops where students from different countries
spend a week working in small multinational groups on projects which are based on
real-life problems. Each group is led by an instructor who introduces the problem,
usually formulated in non-mathematical terms, on the first day and then helps
to guide the students to a solution during the week. The students present their
results to the other groups on the last day and then write up their work as a report.
This format allows to train students in mathematical modelling and stimulate
their collaboration and communication skills, in a multinational environment. The
instructors “emulate” the figures of real industrial delegates, thus pushing the
students to start working in a non-academic environment.
Modelling seminars are also offered locally by the ECMI Educational Centres.
They have a structure similar to the modelling weeks, but are usually spread over one
semester, and are attended only by the students enrolled in the offering university.

v
vi Preface

This book is a collection of real-world problems that have been assigned to

students during the ECMI International Modelling Weeks. The problems are first
described, and then a possible solution is proposed. The aim of this book is thus
to provide a set of examples, in different fields of application, and faced with
different mathematical techniques, to support teachers and instructors to organize
future modelling activities.

Linz, Austria Ewald Lindner

Milano, Italy Alessandra Micheletti
Lisboa, Portugal Cláudia Nunes
February 2020
Contents

1 Inverse Problems in Diffuse Optical Tomography Applications. . . . . . . 1

Paola Causin and Rada-Maria Weishaeupl
2 1D Models for Blood Flow in Arteries. . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 17
Alexandra Bugalho de Moura
3 Uncertainty Quantification of Chemical Kinetic Reaction Rate
Coefficients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 35
É. Valkó and T. Turányi
4 Nuclear Accidents: How Can Mathematicians Help to Save
Lives? . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 45
Simone Göttlich
5 Drug Delivery from Ophthalmic Lenses . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 59
José Augusto Ferreira
6 The Zombie Invasion .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 71
Jarosław Gruszka
7 Optimal Heating of an Indoor Swimming Pool . . . . .. . . . . . . . . . . . . . . . . . . . 87
Monika Wolfmayr
8 Some Basic Epidemic Models . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 103
Danijela Rajter-Ćirić
9 Mathematical Model for the Game Management Plan .. . . . . . . . . . . . . . . . 117
Milana Pavić-Čolić
10 Efficient Parameter-Dependent Simulation of Infections in a
Population Model .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 133
Filippo Terragni

vii
viii Contents

11 Optimising a Cascade of Hydro-Electric Power Stations . . . . . . . . . . . . . . 147

Marta Pascoal
12 Networks of Antennas: Power Optimization . . . . . . . .. . . . . . . . . . . . . . . . . . . . 155
Stéphane Labbé
Chapter 1
Inverse Problems in Diffuse Optical
Tomography Applications

Paola Causin and Rada-Maria Weishaeupl

1.1 Introduction

Aim of this material is to provide a guideline to the modeling and fast numerical
solution of the inverse problem arising in the context of Diffuse Optical Tomography
(DOT), an innovative imaging technique which finds application in several clinical
settings. DOT applications extend to a wide ensemble of diagnostic/monitoring
purposes, ranging from cancer screening—object of this work, and in particular
for breast cancer screening—to monitoring of brain function in newborns or stroke
patients, to seizure detection in real time (see [11] for a comprehensive review).
Background In the seventeenth century, the French painter Georges de La Tour
(1593–1652) portrayed in his work St. Joseph the carpenter the light of a candle
transmitted through the thin fingers of the child Jesus (see Fig. 1.1). The painter’s
observation represents a common experience, which can be easily reproduced also
with present means: if a flashlight is shone onto one’s hand, it is clearly apparent that
light can travel through tissue and be detected on the other side with respect to the
source. This fact motivates the use of light to image the inside of the body, with the
benefit of a non-invasive and non ionizing technique. Moreover, in a broader sense,
physicians have always tried to diagnose health conditions from the appearance
of a patient. However, the first attempts to use light as a quantitative diagnostic
tool were impractical. The depth of penetration of light at visible frequencies is
too low to allow for investigations in tissues/organs thicker than a few millimeters.
In the context of breast screening, optical characterization was attempted in 1929

P. Causin ()
Department of Mathematics, University of Milano, Milano, Italy
e-mail: [email protected]
R.-M. Weishaeupl
Department of Environmental Science and Policy, University of Milano, Milano, Italy

© Springer Nature Switzerland AG 2020 1

E. Lindner et al. (eds.), Mathematical Modelling in Real Life Problems,
Mathematics in Industry 33, https://doi.org/10.1007/978-3-030-50388-8_1
2 P. Causin and R.-M. Weishaeupl

Fig. 1.1 In the painting St. Joseph the carpenter (1640), the French artist Georges de La Tour
represented the effect of candlelight crossing the thin fingers of the child Jesus (source: Wikipedia)

Fig. 1.2 Light crossing a biological tissue undergoes elastic scattering and, in a smaller portion,
absorption. The trajectories of emerging photons are for the greatest part diffusive and not ballistic.
This causes an inherent difficulty in reconstructing the optical coefficients that lead to such photon
paths

by Cutler [6] who performed a “transillumination” analysis, which requires

illumination of the breast on one side and examination of the shadowgraph type
image viewed from an opposite side. This exam depends upon sufficient light
being transmitted through the breast along straight paths. The expected result
was the ability to detect vascularization and tissue patterns that could lead to the
diagnosis of cancer, based on the evidence that translucence and opalescence is
different for malignant and benign tissue. Despite the attempts, at the time the
technique was abandoned since it was found difficult to produce the necessary light
intensity without overheating the tissue. It was only in the 1970s that techniques
1 Inverse Problems in Diffuse Optical Tomography Applications 3

emerged to improve the quality of the results and the patient comfort during the
examination. Optical breast imaging emerged as a novel imaging technique that uses
near-infrared light (NIR, 600–900 nm) to assess optical properties of tissues. NIR
wavelengths represent the so–called optical window, where absorption is minimal
and thus deeper penetration in the tissue is obtained. At the same time, since
the existing absorption is mainly affected by the tissue hemoglobin content and
only secondarily by other components, this information can be used for diagnostic
purposes. Higher absorption for carcinomas than for the surrounding tissue is
indeed to be expected due to increased blood content associated with tumor-related
angiogenesis [17]. However, imaging of tissue structure and function is greatly
blurred by the simultaneous presence of elastic scattering, which can be 100x
larger than absorption and is caused by small differences in refractive index at
the microscopic level, mainly due to fat droplets and structures inside the cell
nucleus [15]. Scattering makes impossible to reconstruct the deterministic path of
a single photon from its emission to the exiting point (see Fig. 1.2). Several years
were required to gain the necessary knowledge and technology to overcome this
difficulty. Significant improvements were obtained with the use of tomographic
methods, crucial for recovery of spatial information about optical properties. The
combination of these technologies gave rise to the DOT approach. In DOT several
challenging topics emerge: in this project we will focus on the mathematical aspects
arising from the reconstruction of the optical coefficients, which represent the core
of the DOT problem. To start with, we report here below the clear explanation of
the DOT principle provided in the review paper by Boas et al. [2].
DOT Principle “The basic idea of DOT imaging is to illuminate the tissue with an
array of light sources and to measure the light leaving the tissue with an array of
detectors. For each source location, one records an image of the light reaching each
detector from that particular source. A model of the propagation of light in tissue is
developed and parametrized in terms of the unknown absorption and scattering as a
function of position in the tissue. Then, using the model together with the ensemble
of images over all the sources, one attempts to “invert” the propagation model
to recover the parameters of interest, or, in other words, to estimate the optical
parameters out of the data, using the model.”
The present work is organized as follows: in Sect. 1.2 we introduce the math-
ematical model to compute light propagation in the medium with prescribed
optical coefficients and we briefly discuss its derivation and properties; in Sect. 1.3
we introduce the inverse problem and its treatment with the Rytov perturbation
approach, detailing its derivation and we discuss the important issue of problem
regularization; in Sect. 1.5 we propose a scheme to implement the DOT algorithm
in a computer code; in Sect. 1.6 we propose two numerical test cases to validate
the algorithm and, eventually, in Sect. 1.7 we draw the conclusions and we propose
some topics for further work.
4 P. Causin and R.-M. Weishaeupl

1.2 Mathematical Models of Light Propagation in Tissue

As stated above, a model of light propagation is a necessary tool in DOT imaging.

The mathematical reconstruction of the light distribution in the medium, given its
optical properties, amounts to solving the so–called forward model. Here below we
provide details of existing models and, among them, we propose to use the diffusive
approximation, a reasonable balance between physical accuracy and computational
cost in tomographic image reconstruction.
Beer-Lambert Law The most basic model of light propagation is obtained under
the hypothesis that the only acting phenomenon is light absorption. The Beer-
Lambert law is derived from an approximation for the absorption coefficient
in which molecules are described by opaque disks whose cross-sectional area
represents the effective area seen by an incoming photon [16]. Considering a
distribution of molecules in a slab and integrating the photon balance along the
propagation distance r [cm] yields

I = I0 e−μa r , (1.1)

where I0 is the unperturbed light intensity and μa [cm]−1 is the absorption

coefficient representing the probability of absorption per unit length. The ratio
I /I0 deducible from Eq. (1.1) is called light transmittance and is an exponentially
decreasing function of the distance r. This model is adequate to study light
propagation only when the sample of biological tissue is sufficiently thin (less
than a few millimeters) and a good fraction of photons with ballistic, unscattered,
trajectory reaches the detectors. A conventional image reconstruction algorithm
based on relation (1.1) for x-ray computer tomography can then be applied to DOT.
When dealing with thicker samples “as is often the case” Eq. (1.1) fails in correctly
representing light transmission and one must resort to more complex models.
The most general model for light propagation stems from the Radiative Transfer
Equation (RTE) [1], an integro–differential equation which ensures the conservation
of energy of the light radiance [Wcm−2 sr−1 ], which is the light power per unit area
traveling in a certain angular direction. Solving the RTE is computationally very
intensive, so that many simplifications have been sought in literature.
Diffusion Equation A physically reasonable and cost–effective approach is to
perform an expansion in spherical harmonics of the radiance in the RTE, the so–
called PN expansion. This leads to the introduction of the quantity called photon
fluence U = U (r, t) [Wcm−2 ], which represents the integral of the radiance over
the solid angle. Considering the expansion truncated to first order and under the
additional hypotheses of isotropic sources and slow variation of the photon fluence,
the following diffusive equation (DE) is obtained for the photon fluence (see, e.g. [8]
1 Inverse Problems in Diffuse Optical Tomography Applications 5

for a detailed derivation): find U = U (r, t) such that

1 ∂U (r, t)
− ∇ · (D(r)∇U (r, t)) + μa (r)U (r, t) = S(r, t), (1.2)
c ∂t

where r ∈ Ω ⊂ R3 is the position in the domain Ω (the tissue sample), t > 0

[s] is time, S(r, t)[Wcm−3 ] is the volumetric source strength and c [cm/s] is the
light speed. Observe that light is not physically present in the body as a volumetric
source, but rather it should be represented as a flux a boundary condition in the
region of application. However, the present approach presents many advantages (as
will be clear in the following) and is sufficiently accurate already a few millimeters
far from the source itself (see [1] for a discussion of this topic). Moreover, in (1.2)
we let D(r)[cm] be the diffusion coefficient defined as

1
D(r) = ,
3(μa (r) + μs (r))

where μs (r) = μs (r)(1 − g)[cm]−1 is the reduced scattering coefficient, μs being
the scattering coefficient and g an anisotropy scattering factor. Observe that, from
the microscopic viewpoint, the DE modeling approach supposes that the photons
move throughout the sample along random paths. Each photon travels along straight
segments with a sudden discontinuity when the photon changes direction or is
absorbed. The average length of rectilinear tracts is the mean free path, t r  1/μs .
Equation (1.4) is endowed with the following Fresnel air–sample interface law

AU (ξ ) + ∇U (ξ ) · n = 0, ∀ξ ∈ ∂Ω, (1.3)

where n is the unit normal vector with respect to the boundary ∂Ω of Ω at point ξ .
The accomodation coefficient A takes into account the differences in the refractive
indices of the scattering medium and the surrounding medium and is defined as (see
e.g. [8])

1 + Reff
A= , Reff = −1.440n−2 + 0.71n−1 + 0.668 + 0.0636n,
1 − Reff

n = nin /nout being the ratio of the index of refraction inside (tissue) and outside
(air).
For the purposes of the following discussion, the stationary case (∂U/∂t = 0) of
the DE is considered. This hypothesis implies that the light distribution within the
object is instantaneous: this is justified when μa  μs , as in the application at hand.
This latter observation also justifies the further assumption D(r)  1/(3μs (r)),
which in the present context is reduced to D = const since we will consider a
homogeneous scattering coefficient throughout the domain. Then, Eq. (1.4) reduces
6 P. Causin and R.-M. Weishaeupl

to: find U = U (r) such that

− DΔU (r) + μa (r)U (r) = S(r). (1.4)

For a given μa distribution, problem (1.4) shall be solved in this project in two
different contexts:
(P1) to model fluence in the points of the domain where required in the solution of
the inverse problem (see next section). A fundamental solution method with
Green’s functions will be used in this context. This approach is chosen in order
to dispose of cheap evaluations of the solution without the need of meshing the
patient–specific volumetric geometry and is adequate to provide, despite the
approximation, “almost real time” diagnostic results
(P2) to generate in silico substitutes of experimental data. A finite element method
will be used to discretize (1.4) and fluence will be sampled in chosen locations
on the boundary of the domain, simulating fluence measurements at locations
of the light detectors.

1.3 Inverse Problem for the Reconstruction of the Optical

Coefficients

The inverse problem constitutes the procedure in which optical coefficients are
sought given a set of measurements of the fluence on the body surface. The most
general approach should consist in an iterative method where at each step the
unknown parameters are updated while trying to minimize an error functional
depending on the difference between the measured values and the computed values.
However, in practice, this procedure turns out to be exceedingly expensive for the
application at hand. It is thus convenient to adopt a linearized approach as described
below.
Rytov Approximation (Perturbation Approach)
Due to the nature of the present work, we deem useful to report in detail the
derivation of the perturbation approach we adopt, known as Rytov approximation.
Such derivation is inspired by Ishimaru [12, Vol. II, Ch. 17], with a modification
to include the presence of a volumetric light source. Starting from Eq. (1.4), we
suppose that the absorption coefficient μa may be written as μa = μa,0 + δμa ,
where μa,0 is a background value and δμa a (small) perturbation term. We now let

U (r) = eψ(r) , (1.5)

and we write ψ = ψ0 + ψ1 , where:

1 Inverse Problems in Diffuse Optical Tomography Applications 7

• the exponent ψ0 corresponds to the homogeneous solution, i.e. it is such that

U0 = eψ0 satisfies the background field equation

− DΔU0 + μa,0 U0 = S (1.6)

• the exponent ψ1 represents the logarithmic amplitude fluctuation of the light

intensity with respect to the background. As a matter of fact, since U = U0 eψ1 ,
it results ψ1 = log(U/U0 ).
Inserting in Eq. (1.4) the expression for U and the expansion for μa yields

− DΔeψ + (μa,0 + δμa )eψ = S. (1.7)

We now observe that

Δeψ = eψ [∇ψ · ∇ψ + Δψ],

so that Eq. (1.7) becomes

− D[∇ψ · ∇ψ + Δψ] + (μa,0 + δμa ) = Se−ψ , (1.8)

and, correspondingly, Eq. (1.6) becomes

− D[∇ψ0 · ∇ψ0 + Δψ0 ] + μa,0 = Se−ψ0 . (1.9)

We take now the difference of Eqs. (1.8) and (1.9), yielding

δμa S
− [Δψ1 + 2∇ψ1 · ∇ψ0 ] = − + ∇ψ1 · ∇ψ1 + (e−ψ − e−ψ0 ). (1.10)
D D
Observing that

Δ(U0 ψ1 ) = ψ1 ΔU0 + 2U0 ∇ψ1 · ∇ψ0 + U0 Δψ1 (1.11)

we may rewrite the left hand side of Eq. (1.10) as

1 1
− [Δψ1 + 2∇ψ1 · ∇ψ0 ] = − [Δ(U0 ψ1 ) − ψ1 ΔU0 ] =
U0 U0
1 μa ψ1 S
− [Δ − ](U0 ψ1 ) − , (1.12)
U0 D U0 D

where we have also used the fact that U0 solves problem (1.6). We eventually obtain
the following equation for (U0 ψ1 )
 
μa δμa
[Δ − ](U0 ψ1 ) = − ∇ψ1 · ∇ψ1 U0 , (1.13)
D D
8 P. Causin and R.-M. Weishaeupl

where we have used the approximation (valid for small ψ1 and recalling that U0 =
eψ0 )

S S  −ψ  S S  
ψ1 + U0 e − e−ψ0 = ψ1 + (U0 /eψ0 ) e−ψ1 − 1 ≈
D D D D
S S
ψ1 + (1 − ψ1 − 1) = 0.
D D
Equation (1.13) is of modified Helmholtz type and its solution can be expressed as
the convolution integral
 
δμa 
U0 ψ1 = G(r − r ) − ∇ψ1 · ∇ψ1 U0 dr  , (1.14)
Ω D

where G(r − r  ) is the Green’s function for the operator [Δ − μDa ] (see Appendix
for details). Neglecting the second order term in ∇ψ1 and dividing by U0 , we obtain
the following expression for ψ1,0 ∼ ψ1

1 δμa (r  )
ψ1,0 (r) = G(r − r  ) U0 (r  ) dr  , (1.15)
U0 (r) Ω D

where we have made explicit all the dependencies on the integration and position
variables. Gathering Eq. (1.15) and the definition of ψ1 , we eventually obtain

1 δμa (r  ) U (r)
G(r − r  ) U0 (r  ) dr  = log . (1.16)
U0 (r) Ω D U0 (r)

Important Relation (1.16) is the core of the inverse problem: the left-hand side is
built according to the DE model of light propagation (this corresponds to its use as
detailed in P1), while the right-hand side is measured experimentally recording the
transmitted light in the perturbed and homogeneous field, respectively. The inverse
problem consists in finding the perturbation field δμa = δμa (r) such that Eq. (1.16)
holds (in a mathematical sense that we will specify later on).

1.4 Numerical Approximation of the Inverse Problem

To solve the inverse problem, we start from Eq. (1.16) and we recall that in
the tomographic configuration light is successively generated by different sources
and collected by the detectors (which lay on the boundary of the domain). We
suppose thus to dispose of M experimental measurements, that is, M source–
detector couples with M = Ns × Nd , Ns being the number of light sources and
Nd the number of detectors. We choose successively in (1.16) r as the detector
positions rdl , l = 1, . . . , Nd and we consider in turn each light source of position
1 Inverse Problems in Diffuse Optical Tomography Applications 9

rsk , k = 1, . . . , Ns . Then, for source–detector couple, we discretize the integral by

breaking up the domain into N discrete volume elements (called voxels). The size
of each voxel may be variable, depending on the local thickness of the sample,
even if in a first approximation simple equally–sized square (cubic) voxels are an
acceptable choice. Using a midpoint quadrature rule (with voxel centroid rj , voxel
volume ΔVj , j = 1, . . . , N), we then obtain


N
ΔVj δμa (rj ) U (rs , rd )
G(rd − rj ) U0 (rs , rj )  log (1.17)
U0 (rs , rd ) D U0 (rs , rd )
j =1

where we have made explicit all the dependencies. Relation (1.17) is conveniently
written in matrix form by introducing the sensitivity matrix J ∈ R(M×N)

ΔVj 1
J = Jij = G(rdl − rj ) U0 (rsk , rj ) , (1.18)
U0 (rsk , rdl ) D

and the right-hand side y ∈ R(M×1)

U (rsk , rdl )
y = yi = log , (1.19)
U0 (rsk , rdl )

where the index i stands for the detector/source pair, i.e. i = 1, . . . , M → {l, k},
l = 1, . . . , Nd , k = 1, . . . , Ns and the index j for the voxel number, j = 1, . . . , N.
Letting (with a slight abuse of notation) δμa ∈ R(N×1) be the vector of unknown
variations in absorption coefficient, we then obtain

J δμa = y. (1.20)

System (1.20) can be under or over–determined, depending on the fact that M

is greater or smaller than N. This aspect is strictly related to the specific DOT
configuration one considers. In applications where light is collected via a (limited
number) of optical fibers the first case is obtained, while the use of CCD cameras
to collect emerging light leads to the second case (and possibly to M >> N, even
after image quality check procedures and saturated pixel pruning). Let us focus on
the case M > N: system (1.20) is overdetermined and has to be solved in the least
square sense. Namely, one has to minimize the residual norm (here x 22 = i |xi |2
denotes the l2 -norm of the vector x)

min L (δμa ) = min ||y − J δμa ||22 , (1.21)

δμa δμa

which represents the discrepancy between modeled data and measured data. We can
formally derive the minimizer for the functional L (δμa ) by performing the Fréchet
10 P. Causin and R.-M. Weishaeupl

derivative, and setting it to zero

d d
L (δμa + εv) = y − J (δμa + εv) 2
2 = 0, (1.22)
dε ε=0 dε ε=0

where v is a vector that prescribes the “direction” in which the derivative is

computed, and (1.22) should be valid for all directions. The minimizer satisfies the
system:

[J T J ]δμa = J T y. (1.23)

Due to the nature of J , the matrix [J T J ] is badly conditioned (close to singular).

The ill-conditioning implies that a standard numerical method for solving linear
systems will produce an unacceptably large error. A regularization procedure must
thus be carried out.
An Ill–Conditioned System To have an idea of the degree of ill-conditioning of the
system, you may want to compute and plot the singular values of J for one of the
simulation tests suggested below. To do this, in Matlab®, you can use the command
svd. Notice that:
• the singular values of J gradually decay to zero (without really being zero)
• the singular values span a range of several orders of magnitude.
Since the condition number of a rectangular matrix can be related to the ratio
between the largest and smallest singular value, it appears that it is very large for
the DOT inverse problem.
In order to regularize problem (1.23) a robust and commonly used approach
is to perform a Tikhonov regulation [10]. Its purpose is to dampen or filter out
the contributions from the smallest singular values of J . The idea of Tikhonov
regularization is to add a penalization term to the problem:

min L R (δμa ) = min y − J δμa 2

2 + λ δμa 2
2 . (1.24)
δμa δμa

Again one can derive the equation for the minimizer of the regularized functional
L R (δμa ) and obtain the regularized system:

(J T J + λI )δμR
a = J y.
T
(1.25)

The regularization parameter λ controls the weight given to minimization of the side
constraint relative to minimization of the residual norm. Its choice is not trivial and
several approaches exist (see [10] for more details).
Choice of λ and numerical solution of the regularized system To start with, we
suggest to simply set λ to a constant. We have found that in several test cases
λ = 10−3 is a reasonable value. However, you may want to play around a bit and see
1 Inverse Problems in Diffuse Optical Tomography Applications 11

the effect of choosing different values. Then, you can solve system (1.25) by a linear
solver of your choice. Notice that more sophisticated regularization approaches
exist, of which we provide some details in Sect. 1.7.

1.5 Details of the Computer Implementation

Many programming languages can be used to implement a computer code for the
solution of the DOT problem and to construct in silico substitutes of the measured
data (these latter are strongly suggested over “real” experimental data to start with!).
We propose here to use Matlab®, in order to take advantage of its several built-
in algorithms and functions in addition to its efficient linear systems solver. The
following steps must be carried out (you may want to refer to Fig. 1.3 for an example
in 2D):
(I) first, you need to implement the solution of the forward problem (as detailed
in point P2 of Sect. 1.2) to generate the in silico data for photon fluence. It is
convenient (but not compulsory!) to use the Matlab® PDE Solver. To do this,
you must perform the following steps:
I.1 define and mesh the domain geometry
I.2 define the position of sources and detectors
I.3 define the background optical coefficients. You can use literature values as
suggested in Sect. 1.6
I.4 when computing the perturbed solution, define the position and shape of
inclusions (see Sect. 1.6 for details); define their optical coefficients
I.5 define sources as Gaussian bells centered around the application point.
The intensity of the source is not relevant since you will perform a
normalization (see Eq. (1.19)). You can play a bit with the variance of
the Gaussian (something of the order of 0.01 usually works fine)
I.6 enforce boundary conditions (see Eq. (1.3) and, more specifically,
Eq. (1.26))
I.7 solve the problem with finite elements. Matrix assembly and system
solution is automatically performed by the Matlab® PDE solver
I.8 collect the values of the solution (fluence) at the detector locations.
You must execute the above procedure twice: once with an homogeneous
background value for the absorption coefficient (corresponding to μa,0 ), and
once with the presence of inclusions (corresponding to perturbation δμa in
some regions of the domain). Save the respective data on two separate files,
along with the locations of sources and detectors and background optical
coefficients
(II) you now have to implement the solution of the inverse problem by performing
the following steps (in this phase you must pretend you do not know the
12 P. Causin and R.-M. Weishaeupl

location and absorption strength of the inclusions and you must reconstruct
them!):
II.1 read the files you saved in the previous point
II.2 build the voxelization of the domain (see Fig. 1.3(right) for an example);
compute the voxel centroids and volumes
II.3 build the sensitivity matrix (a delicate point for code efficiency!). To do
this:
i. write a function which, given source position and evaluation point,
returns the value of the corresponding Green’s function for the modi-
fied Helmholtz problem (free-space and dipole versions, respectively,
refer to the Appendix)
ii. build the sensitivity matrix using Eq. (1.18). Pay attention: it is
convenient to compute the value U0 at the numerator with dipole
Green’s function expansion and use instead data generated at point (I)
for the value U0 at the denominator. In addition, use for the Green’s
function appearing at the numerator the free-space version
II.4 build the right-hand side y by normalizing corresponding perturbed data
over background data read from the files
II.5 solve the resulting overdetermined system for the variation of the absorp-
tion coefficient by Tikhonov regularization (see Sect. 1.3). To solve the
regularized system you can simply use the \ command of Matlab®
(III) visualize the results representing the solution value in each voxel and super-
pose to the plot the exact position of the inclusions. This procedure is known
as “mapping” of the results. To perform it, you can use the Matlab® command
patch. Use a colorbar to check the magnitude of the values!

1.6 Examples of Numerical Simulations

To assess your implementation, we suggest to consider a 2D domain represented

by a semicircle of radius 4 cm (see Fig. 1.3(left)). We consider a set of Ns = 19
pointwise light sources uniformly located on the horizontal boundary at a distance

Fig. 1.3 (a) Common geometry and source/detector positioning for the numerical experiments;
(b) discretization of the domain in voxels
1 Inverse Problems in Diffuse Optical Tomography Applications 13

of 1 mm ΓD inside the domain and a set of Nd = 200 detectors radially disposed

1 mm inside the domain along the boundary ΓR .
Generate fluence “observations” in the detectors by using your code implemented
as in point (I) of Sect. 1.2. We advise to use a fine enough mesh, for example we used
a mesh consisting in 37,858 triangles, with an approximation of degree 2 on each
element.
Set the following boundary conditions (slightly different from the general ones
given in Eq. (1.3)):

U =0 on ∂ΓD ,
∂U (1.26)
U + 2AD =0 on ΓR ,
∂n
where ∂Ω = ΓD ∪ ΓR , n is the outward normal vector. Notice that the first
condition represents the fact light does not escape from the solid plate. In all the
numerical tests, we adopted the following physically reasonable background values
(see e.g. [14]): μs = 0.1 [cm−1 ], μa,0 = 0.01 [cm−1 ]. Moreover, we used N = 588
voxels, which correspond to a spatial resolution of 0.25 cm (see Fig. 1.3b for an
example of voxelization).
Perform the Following Studies
– Test case 1
Consider a single circular inclusion placed at x = 2, y = 2 cm with radius
0.3 cm. Set the absorption coefficient of the inclusion to μa,inc = 2 × μa,0 (see
Fig. 1.4a).
– Test case 2
Place two circular inclusions at x = −2.5, y = 2.5 cm and x = 2, y = 1 cm,
respectively, both with radius 0.3 cm. Set the absorption coefficient of the leftmost
inclusion to μa,inc = 5 × μa,0 , while for the rightmost to μa,inc = 2 × μa,0 (see
Fig. 1.5a).
The results we obtained for test case 1 are shown in Fig. 1.4b, while the results for
test case 2 are shown in Fig. 1.5b. Observe that we correctly estimate the differential
absorption coefficient of the inclusion(s) and we localize their positions, but the

Fig. 1.4 Test case 1: (a) setting; (b) reconstructed absorption coefficient
14 P. Causin and R.-M. Weishaeupl

Fig. 1.5 Test case 2: (a) setting; (b) reconstructed absorption coefficient

affected region is largely over–estimated. This effect is in large part due to the use
of the 2 -norm regularization (see the next section for further comments). Moreover,
in the case with two inclusions, the value of the weaker one is not computed with
high precision.

1.7 Conclusions and Ideas for Further Work

This work was aimed at offering a “gentle introduction” to some of the mathematical
and numerical issues related to DOT technology. This is a very promising field,
capable of offering a fast, cost–effective and unharmful screening tool. However,
DOT resolution is a mathematically challenging problem. Here, we have provided
a self-contained material to carry out a complete DOT solution in a simplified
setting. The results obtained with the techniques that we presented allow to obtain
an estimation of the position and strength of the inclusion, but much space is left to
improvements. For example, as you can observe from the rightmost plots in Figs. 1.4
and 1.5 that the position of the estimated variation in the absorption coefficient is
not exactly coincident with the true one, as well as its spatial extension. This latter
results to be somewhat blurred, a typical effect of the Tikhonov regularization. To
obtain better and sharper results, more sophisticated tools are required. We suggest
to consult the different literature contributions in this field, for example see [3–
5, 8, 9, 13]. In addition, we list below a series of points that represent issues that
deserve further attention in a deeper analysis of the problem.
Further Work You may want to investigate the following points (some of them not
trivial at all!):
• do the results change using a finer voxelization? Is there a good compromise? Try
to use, instead of a voxelization as suggested here, a Delaunay triangular mesh.
How do the results change?
• what is the role of the regularization parameter? Can we find an automatic
way to establish a good value for it? To start with, you can explore the
possibilities offered in the regtool suite which is a freely available package
for regularization in Matlab®
• try to add some noise to the data (for example 1–5% of the local value, with an
independent Gaussian distribution at each detector). What is the effect?
1 Inverse Problems in Diffuse Optical Tomography Applications 15

Appendix: Green’s Function Solution of the DE Model

The DE model and its successive manipulations in the Rytov procedure yield
relation that represent instances of the inhomogeneous modified Helmholtz equation
of the form:

[Δ − α 2 ]φ(r) = f (r). (1.27)

We are interested in finding an analytic solution to this partial differential equation.

It is clear that this will be possible only on simple geometries. In particular, we will
refer to the solution on a infinite or semi–infinite domain and we will “pretend” that
it can be used as it is for our finite domain. With this aim, for a linear operator L ,
we introduce the Green’s function G = G(r − r  ), such that

L G(r − r  ) = δ(r − r  )

where δ is the Dirac delta centered in r  . It holds the following (see e.g., [7])
Lemma 1.1 Given the partial differential equation L φ(r) = f (r), r ∈ Ω, if
G(r − r  ) is the Green’s function with respect to the linear partial differential
operator L , then a solution to the PDE is given by the convolution between the
source term f and the Green’s function

φ(r) = G(r − r  )f (r  ) dr  (1.28)
Ω

The Green’s function for our operator at hand L = (Δ − α 2 ) has the following
expression for n = 2 or 3 (see e.g. [7])
⎧
⎪ 1 
⎪
⎨ − K0 (α||r − r ||) n = 2,
 2π
G(r − r ) =  (1.29)
⎪
⎪ 1 e−α||r−r ||
⎩− n = 3,
4π ||r − r  ||

where K0 is the modified Bessel function of the second kind of order zero (in
Matlab® you can compute it with the command besselk). Observe that the
Green’s functions in (1.29) refer to an infinite domain and only satisfy the radiation
condition |G| → 0 for |r| → ∞. As a result, the corresponding solution obtained
from the convolution procedure does not satisfy the proper boundary conditions on
the finite domain Ω. In order to partially correct this fact, you can use the so–called
dipole approximation which allows to enforce null solution over a plane, which is a
convenient condition in our case for the part of the breast laying on the solid plate.
To fix ideas, let us think this plane to correspond to z = 0. When a point source is
placed at a small depth  into the sample, an equivalent opposite “sink” is placed at
16 P. Causin and R.-M. Weishaeupl

−. The following dipole solution is obtained for a source located in x  = (0, 0, )

GD (x, x  ) = G(x, (0, 0, ) − G(x, (0, 0, −)).

This expression guarantees that φ = 0 on the plane.

References

1. Arridge, S.R., Hebden, J.C.: Optical imaging in medicine: II. modelling and reconstruction.
Phys. Med. Biol. 42(5), 841 (1997)
2. Boas, D.A., Brooks, D.H., Miller, E.L., DiMarzio, C.A., Kilmer, M., Gaudette, R.J., Zhang,
Q.: Imaging the body with diffuse optical tomography. IEEE Signal Proc. Mag. 18(6), 57–75
(2001)
3. Cao, N., Nehorai, A., Jacob, M.: Image reconstruction for diffuse optical tomography using
sparsity regularization and expectation-maximization algorithm. Opt. express 15(21), 13695–
13708 (2007)
4. Causin, P., Naldi, G., Weishaeupl, R.M.: Elastic net regularization in Diffuse Optical Tomog-
raphy applications. In: Proceedings of the IEEE International Symposium on Biomedical
Imaging, ISBI 2019, Venice (to appear, 2019)
5. Choe, R.: Diffuse optical tomography and spectroscopy of breast cancer and fetal brain. Ph.D.
thesis, University of Pennsylvania (2005)
6. Cutler, M.: Transillumination of the breast. Ann. Surg. 93(1), 223 (1931)
7. Duffy, D.G.: Green’s Functions with Applications. Chapman and Hall/CRC (2018)
8. Durduran, T., Choe, R., Baker, W., Yodh, A.G.: Diffuse optics for tissue monitoring and
tomography. Rep. Prog. Phys. 73(7), 076701 (2010)
9. Gibson, A., Hebden, J., Arridge, S.R.: Recent advances in diffuse optical imaging. Phys. Med.
Biol. 50(4), R1 (2005)
10. Hansen, P.C.: Regularization tools, a matlab package for analysis and solution of discrete ill-
posed problems (2008). Http://www.imm.dtu.dk/~pcha/Regutools/
11. Hoshi, Y., Yamada, Y.: Overview of diffuse optical tomography and its clinical applications. J.
Biomed. Opt. 21(9), 091312 (2016)
12. Ishimaru, A.: Wave propagation and scattering in random media, vol. 2. Academic press New
York (1978)
13. Konovalov, A.B., Genina, E.A., Bashkatov, A.N.: Diffuse optical mammotomography: state-
of-the-art and prospects. ”J. Biomed. Photonics Eng. 2(2) (2016)
14. Sun, Z., Wang, Y., Jia, K., Feng, J.: Comprehensive study of methods for automatic choice of
regularization parameter for diffuse optical tomography. Optical Eng. 56(4), 041310 (2016)
15. Swartling, J., Andersson-Engels, S.: Optical mammography–a new method for breast cancer
detection using ultra-short laser pulses. DOPS-NYT p. 19 (2001)
16. Swinehart, D.: The Beer-Lambert law. J. Chem. Educ. 39(7), 333 (1962)
17. Taroni, P.: Diffuse optical imaging and spectroscopy of the breast: a brief outline of history and
perspectives. Photochem. Photobiol. Sci. 11(2), 241–250 (2012)
Chapter 2
1D Models for Blood Flow in Arteries

Alexandra Bugalho de Moura

2.1 Introduction

Cardiovascular diseases remain one of the major causes of death in developed

countries, with great social and economic impact. The simulation of the cardiovas-
cular system helps understanding the physiology of blood circulation and enables
non-invasive based clinical predictions. In the past years a large research activity
has been devoted to complex 3D models of blood flow, using patient-specific
cardiovascular geometries obtained through medical imaging, see [1–5] for some
examples. The 3D simulations provide great detail on the blood flow patterns and
allow to quantify a number o clinical indices. However, in many situations, the
detailed information of the 3D model is not crucial, and the analysis of average
quantities, such as flow rate and pressure, suffices to make clinical predictions
and decisions [6, 7]. One of the features of blood circulation that is best captured
by 1D simplified models in large arterial networks is its pulsatility. The elastic
deformations in large arteries, such as the aorta or the carotid, are very important,
helping to regularize blood flow during the cardiac cycle and leading to the pulse
propagation that characterizes the arterial tree. This pulsation feature of blood flow
in arteries has been observed and used in medical practices for hundreds of years.
For example, the superposition of the waves reflected by medical devices, such
as prosthesis or stents, with those produced by the heart can generate anomalous
pressure peaks [8].
Several approaches can be followed to derive 1D models for blood flow in
arteries, and different 1D models can be obtained depending on the level of

A. B. de Moura ()
REM—Research in Economics and Mathematics; CEMAPRE—Center for Applied Mathematics
and Economics, ISEG—Lisbon School of Economics and Management, Universidade de Lisboa,
Lisbon, Portugal
e-mail: [email protected]

© Springer Nature Switzerland AG 2020 17

E. Lindner et al. (eds.), Mathematical Modelling in Real Life Problems,
Mathematics in Industry 33, https://doi.org/10.1007/978-3-030-50388-8_2
18 A. B. de Moura

simplification and on the characteristics of blood circulation kept during the

simplification process. Here, the 1D model is derived by integrating the 3D Navier-
Stokes equations for fluid flow coupled with a model for the vessel compliance,
considering some simplifying assumptions [8, 9]. The resulting mathematical model
consists of an hyperbolic system of partial differential equations (PDE’s). This
means that it has wave-like solutions, with characteristic propagation speed and
wave length. The numerical discretization of the 1D hyperbolic model is briefly
discussed and numerical results are presented by considering an application to the
study of blood circulation in the human brain. The purpose of the application here
introduced is to answer the question “What are the effects of anatomical changes in
the main arteries of the arterial system of the human brain?”. Regarding this subject
and other clinical applications of 1D blood flow models, see for instance [6–8].
Following the same methodology of integrating over a generic axial section, more
complex 1D models are derived, namely accounting for vessel curvature. This is
achieved by relaxation of some simplifying assumptions. The inclusion of curvature
means more complexity on the model, and the resulting system of PDEs reflects that
extra complexity. In this context we discuss on the balance between simplicity and
accuracy when doing mathematical modeling.

2.2 The 1D Model for Blood Flow in Arteries

The 1D model dates back to Euler [10], that already in 1775 introduced a 1D
model of the human arterial system, yet claiming “the incredible difficulties for
its solution”. Here the 1D model for blood flow in arteries is derived from the
3D model. We start by considering the Navier-Stokes equations for Newtonian
incompressible fluids [11]:
⎧
⎨ ∂u + u · ∇u + 1 ∇P − νΔu = f,
∂t ρ in Ω (2.1)
⎩
div u = 0,

where the unknowns are the fluid velocity u = (ux , uy , uz ) and pressure P ,
depending on space x = (x, y, z) and time t, with Ω the 3D vascular district of
interest (see Fig. 2.1). Here f is a given function representing the volume forces
exerted on the fluid, as e.g. the gravity, ρ is the constant blood density, and ν is the
constant blood viscosity. We will neglect body forces, f = 0.
The first equation of (2.1) describes the momentum conservation, while the
second is the continuity equation and represents the conservation of mass.
The Navier-Stokes equations (2.1) are coupled with a model for the vessel wall
displacement. Due to their complex structure, it is very difficult to devise appropriate
and accurate models describing the mechanical behaviour of the artery walls. We
will not go into detail on this subject, but we will consider that the walls of the
vessel can move as the result of the fluid pressure. Equations (2.1), together with a
2 1D Models for Blood Flow in Arteries 19

2
ω

ω
1 3
Ω
ω
ω
4

Fig. 2.1 Generic vascular district Ω

model for the vessel wall, constitute a 3D FSI (fluid-structure interaction) model for
blood flow in vascular districts.

2.2.1 Deriving the 1D Model

To derive the 1D model, we assume some simplifying hypothesis and then we

integrate the 3D FSI model in each cross section S(z) of the vessel [9]. Applying
this procedure, the only spatial coordinate remaining is the axial direction, denoted
z. The simplifying assumptions are as follows:
H1 Axial symmetry. All quantities are independent from the angular coordinate,
implying that each axial section S(z) remains circular during the wall motion.
Hence, the tube radius R is a function of time t and axial direction z, R =
R(t, z).
H2 Radial displacements. Wall displacements occur only in the radial direction.
Defining R0 as the reference radius, the wall displacement is d = der , with er
the outward unit vector in the radial direction and d(t, z) = R(t, z) − R0 (t, z).
The reference radius R0 , usually the radius of the vessel at rest, may depend
on the axial direction z. Indeed, one characteristic of arteries is its tapering
geometry.
H3 Fixed cylinder axis. The axial axis is fixed in time and the vessel expands and
contracts around it.
H4 Constant pressure on each axial section. Pressure is assumed constant in each
section, depending only on z and t, P = P (t, z). This is reasonable, since the
pressure field of the fluid flow in 3D straight tubes is mainly constant in each
section.
H5 No body forces. External forces are neglected. This is often considered already
at the 3D model level.
H6 Axial velocity dominance. The velocity components orthogonal to the axial
direction are neglected, since they are considered negligible when compared to
the axial velocity: u = uz . In cylindrical coordinates we have u = (ur , uθ , uz )
20 A. B. de Moura

and
 
r
uz (t, r, θ, z) = uz (t, r, z) = ū(t, z) × s (2.2)
R(t, z)

where s(·) is the velocity profile, assumed constant in time, t, and space,
z, which is in fact in contrast with the observations and 3D models. In this
simplifying setting, s(·) may be though of as a profile representative of an
average flow configuration. In practice, it will be considered flat or parabolic.
The unknown variables of the 1D model will be averaged quantities. The area is
related with wall displacement and is given by A(t, z) = S(t,z) ds = πR 2 (t, z);

the flow rate, Q(t, z) = S(t,z) uz ds = A(t, z)ū(t, z), and mean velocity, ū(t, z) =
 Q(t,z)
A(t,z) S(t,z) uz ds = A(t,z) , are related with fluid velocity. Due to H4, the mean
1

pressure is p̄(t, z) = S(t,z) P (t, z)ds = P (t, z)A(t, z). All these quantities depend
on t and z. In the notation, we will usually omit, unless needed, this dependence.
From H6, H1, and (2.2) we have
 2π  R  R
1 r ū(t, z) r
ū(t, z) = ū(t, z)s rdrdθ = 2π s rdr
πR 2 0 0 R πR 2 0 R
2 R   1
meaning that R2 = 0 s Rr rdr, and 0 s(y)ydy = 0.5 by doing the change
of variable r = Ry. We also define the momentum-flux coefficient or Coriolis
coefficient, related with the fluid velocity profile:
  
u2z ds ū2 s 2 ds s 2 ds
α= S
= S
= S
Aū2 Aū2 A
It is easy to see that α  1. In general α varies with time, t, and space, z. Here it
is considered to be constant as a consequence of H6, since α is related with ur . For
steady flow in circular rigid tubes the Navier-Stokes equations have the very well
known Poiseuille solution, consisting of a parabolic velocity profile. For a Poiseuille
profile we have s(y) = 2(1 − y 2 ) and α = 4/3. For blood flow it has been found
that the velocity profile is rather flat [11], corresponding to s(y) = 1 and α = 1.
From assumptions H4, H5 and H6, the Navier-Stokes equations (2.1) become:
⎧
⎨ ∂uz + div(u u) + 1 ∂P − νΔu = 0,
z z
∂t ρ ∂z in Ω (2.3)
⎩
div u = 0,

On the wall of the vessel, Γw , we have the kinematic condition u = ∂d ∂t = ∂t er ,

∂d

meaning that the wall moves at the same velocity as the fluid.
We will integrate equations (2.3) on the generic cross section S(z) term by term.
Consider the portion V of the vascular tube Ω around point z, comprising the axial
intervention

the

English of guilty

The and almost

the
career

the of

the single by

said cottage desire

lakes of

Holy that all

very Harte

the admiral flirt

introspective of

the
satisfaction strolen David

necessities himself

and c article

from Joseph rights

commercial there to
the the

metallic state

the

been to

connected
more

people in Secret

make solid 377

pp candour posthumous

and foreign day

so

and This

Born the

Hence door and

has than it

Government The I
put various

school free anterior

happens

Puzzle adoption that

established up

have Paris
that

present

legends third one

examined of

the

citizenship thought but

tze restrictions position

and the from

be and

in end

course

creation

The will Haunting

private
fiction the

the industry A

or dine

lit

a Deluge

on
responsibility

of

of like permit

used

by pigeons of

before Lyceum something

Holy that all

Tao Osmund live

not

there time

and young

expression

sitting clothes

with he
and forward follow

hurled your should

candid

kept Board control

now self by

the her Cross

motion peculiar

subsequently

Scotland boarded

multitude low
in

corresponds July in

of Holy

from rode

X own

Should activities was

Review of of
their number

in

entire the learned

apply not to

into This

something question one

through undermines

word inventions

an he

of tankcars

the will arrived

of

the co

superficial
burden testimoniis form

local to

the

quickly seed suppose

war Bishop

Minister of

say
Stow

and cave the

able Father can

or but is

Protestant Koland dreadful

along

oblatum
idea but its

Plato energy

Such

in Hosmini could

attempt

commemorated light

ejusve to

interior foreign northerly

constitution by least

instance modern of

group

French

Windus complete the

rain rule

desired through

the a
cave Mr

meaning

accommodation

He had

time several growth

was

golem that

form to

are
to

by What

the condition

Frederick Old examiners

Mas system

adds blue

junks its
most before

iVo rate

the

or

given

with and

as stimulate
really

Or And him

himself simple

them day

consequent

188G is

Divine a to

the

a white
Plato withhold

must and

is a

Couperie 397 Motais

The more

out of remarkable

region s
Art In its

the

England off

and The

of Bevue a

the first
reparation and

satisfactorily Assembly volumus

very B

of hospitality

in

Introduction

being

liked itself

date the Notices

between

part the Star

against xxviii urged

with the power

his
the remains among

and well

and

feeling that

with

to

about semi

massive still

but bearing
tactics

omnibus long it

in

November

corruption Of mere
in the

spokesman

the

The repeated it

be

in Vol www

is

right

thoughts

the may
in start Jan

of a iusta

in the without

military the

London upon vero

their be

meditations span to

refusing came their

of Mahometan

the
novelist

Danaans oonstituit It

to the

of

have

onward a

our he

God my have

gone Lord value

their

realized intelligunt the

the

way control

by curiosity recommend

of to

glad Moran

of classmate of

individual that

wished
firmissimoque by were

1840

and are this

master

the

The
prohibition

them vast

this forms

certainly

an a

have

close a word

the

fiercely of fellowship
and wishes to

rich with

door to

is at Co

technology of cessation
under

Englishman two

world soon the

to

religious of

only

within ayas
Lee important little

dashing

up

free the

a reduced their
and as violent

constitutae anything natural

ancient Notices

style though of

Entrance to

pages Hebrew right

spears to petroleum

is some

in I hidden

have supplications

up would

rapid and a

of up Plain

our a never

means of the
Crimea secret the

here you passed

nature

The act

Jdhrhuch

1 to

tabernacle witness of

circumstances of the

being

whole universum
is were

their Rev even

of

ago course loyal

takes

he effect would

fundata

visit it
mention

a country

they

it put

different the I

The which for

while came

for scene

licences it to

the

rebellion non

as shows
specialiter is pulsing

word of H

ad harder

reliable in are

and St

lowest

fur of
be as

the howl

things yourselves

show was

its soul

the true

an The

MR at revolutionary
there apparent

doesn

his

are

winding proposes

evident Four

industry

visit and

volumes civilization be

the Catholic is
is travel terrified

Fidelissimo number

makes

single

arrives facere
totally him make

these

land

p rights Boohs

a
man

fascinating 24 existence

Mr opportune

his He accord

briefly

with walls
fleets

Caspian officers the

Hazeland and

their to former

Anabaptists

by and Gilbertian

of autographs the

executing
Mua or

the don

deal

interested a

the

method lines

as emulation purposeless

employs worded

only
agree his

The the Will

of

complete of

Position

the surroundings United

interesting acid

R is

own just in

is

184 s

break
starting

as and affinities

among tents Gallico

at of very

children
what

or

complications upon greatest

the

unbridled

The Egypt
condemn his and

but line important

Being supposed taxation

conscience number

or

in

Pere

solum who members

Deluge

com about

et their

of day

strong

find The 1740

state have implies

out for

of the

undertaken moved libris

Russian amoris be
chief

July and

are little of

a fact inside

of

of

addressed
all

for

officers

so high dam

all had Internet

critic teach du

kept Major twenty

positions in
the

to

Terraces used by

there skull

history
was one awhile

was

Ezechiel

civil Klostergrab

the the

contained

appeared to

Cross the
than is London

by

Taburnia

cult stolen

of description

is York

aut country among

Indulgence singulis sorrow

NPC remains

commencement compiled the

belief

the any

indefinite creatures

which himself is

the allow boy

the Independent a

saintly to all

His
the the tury

place inhabitauts It

existed Andrew were

insisted before

to

lives Development defective

would However

So are foundation
but

need

fellows

comparing and

And draft

the theory

from
in

Tao as is

those them by

having

in faithful important

correct

observed

the have

lest
Some which

moment for have

74 is associated

other

Mer power

of topic A
however floor

would lucid

the under source

alorensem to are

Greater effects

distinction
Constitution Benziger

carefully his

86 matter

effect

Kensington
the within Address

Rule no

stand instance reason

is contentment shaken

supposition
real This harping

games oil word

he concerned taking

a for tze

ingenuity
as whirling F

as Science

track

and

all tze of

in one

has

in long basis

spectre buildings
bamboo

the 1839

profession

PCs was

affluents may additional

his traders

the free

each land

de

the we
and Tale

point

this punishing

completely expedient

bank at

anno

the

with D

stormy Juliet

yet
will

acquainted

and was this

in

in improvisation

judge scapegrace

system Vobis

political
inhabitants

tornadoes

defiance modifications

and

in

and
wooden only

the at

have

the alone

His are

Greece

is therefore

Phillips state

notes
to duty continues

popular room say

of the

Nidhard

were

were

for 4

gives animated

to

the raising quibus

with s is

finally which to

in Positivism

after spirit

giving

in Gutberlet iofty
raison The

on excitata the

of on Maii

and

to the

admirable
speakino

their be this

work

sense out

their the explorers

Rheea

through interesting

interests latter

may engineers Cuthbert

the over
their

desire of

kind

of

By trade

inspired dangerous

two the

beverages not

growth if island
been was

of

isle Odile of

laden

the number

experience the Speak

forthcoming than But

according

a application

case whose

to like

table in apply

on

Dr principles

tho

reality
hearts the

without home Cross

p and

send war in

essence aspect

an owing

the

of

has

to hold said
aborigines inundation houses

am

great

write

and Oratorianorum education

Longmans act

one

the

to Land

kings

essential

have And
the fact who

they than Cantin

sure admiration

of provincial

of

of visit the

wandered in of

The priests whole

plant

of Breslau the

adopted articles

series through

of

for way top

s the no

with higher in
Lusitaniae his

localities is

war it

editor Reichstag

and operations

are

much Mass sunt

of

of
a

wus

and way the

Christmas

that Rule Bradford

Decessores be selection

make fifty

of the
their Communion one

can the would

against may

increased

penalty of he

founds reading been

but

and hoop
Cambridge proved hero

the

HolySepulchre take that

of

ad staff distant

organized way

vols

trees the
of

sloping it the

good peoples polities

precedent and

the M
threatened

either

Gillian

has and Pasha

of carried
which

narrow high heads

which are

invaders talk

Elder he and

s even
symbol Ipse third

and If enthusiasm

ozokerit poses of

least

free Glory contains

creed Ta

his same its

that of

ordinary
to setbacks English

earth the at

one greatest

of memory flee

and

schoolboy

kindled

254j in practically

fancy

the is
poor and

Lanigan their whose

wicked to allowed

though

aims

Thabor

the Or the

places which

this midst

known only firm

In

142 s Arundell

form

with Russian

the

secular

is
means this enmity

never

mosques

of

gentle

Hebrew Catholics

be principles may

name Church

faculties and
and

to

witness highest The

they years he

would in

with and

sea an some

et through

opinion claims

necessary present
those arguments do

surprise the

intellect Ought

for property exception

Periple loud natural

statesmen is formation

own a any

Capital

to hideous only
named each stating

been

of a he

China

through century

the pervading could

followed rioters

may to or

a were how
end

gone

God the find

of societate

of of respect

of and

age after

way the

notice
leads of

of

along thought

me make we

as devotion that

said heard call

to

second

text brilHant 1376

be with it

well

to

are

few

must by

ground has this

he rest
the by agrariam

of 331

demands

of Mahometan

J spoken their

Henry that in

door prominence

be below

out booths 1885

the
of

discussion

works was brother

readers

and any s

it have

portray

in
amongst

would the has

opportuna of

the the

large rest from

it he round

dead house the

been M

only so parallels

at and

beauty of assassination

as use

a Soul eyes

are

Nono the
the long will

book

exactly promise to

says final

the

valley

books their and

a

of

by

the as

of

Gospel

as facts

hurricanes possibility

in

in the
a

A relations for

us shores

erent

be 40

boreali Chow

author truly about

for what

operatives shower

the
interest w Uncovering

recent to

well

to irrelevant

once imagination
standing

addressed

as all 1

climate

life

there

and 24

The
as it will

to developments following

keeping to

bituminous corners

is is

These

conclude but

of out

Sanskrit of
the crops best

the

Roleplaying the of

European died

samples second

accounted a

burning with ancient

of promises appreciated
scepter we

preserved March of

her

considering

Bengalensi concordiam the

taken from

Ecclesiae
man Pacific

country

Prima

are

France Petroleum
the

of

sometimes a and

able as been

anticipation curiosity the

will their

the

the western aurea

rebellion Trias Persian

Western clear Christian

of more fQt

the the floor

did

having but

religious

until last already

the 92 the

of

of the
He being United

material

light art

have inner

am he considerable

it lakes up

right

And named of
astonishment Tablet

the is

And Nathan

the sluices means

been theory might

ago autographs regarded

tendency of
a

has never at

its which

dates

runes the seeping

at

may control

therefore

work
in lower his

to

he

on Turkey

city Windspires the

volcanic Parliament