Download Creating in-vitro phantoms of blood vessels to support the testing

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts

Medical imaging wikipedia , lookup

Transcript
Creating in-vitro phantoms of blood vessels to support the testing
and validation of new ultrasonic blood flow imaging techniques
Guillaume Pattyn
Supervisors: Prof. dr. ir. Abigaïl Swillens, Prof. dr. ir. Patrick Segers
Master's dissertation submitted in order to obtain the academic degree of Master of
Science in de industriële wetenschappen: elektromechanica
Department of Electronics and Information Systems
Chair: Prof. dr. ir. Rik Van de Walle
Faculty of Engineering and Architecture
Academic year 2015-2016
i
Creating in-vitro phantoms of blood vessels to support the testing
and validation of new ultrasonic blood flow imaging techniques
Guillaume Pattyn
Supervisors: Prof. dr. ir. Abigaïl Swillens, Prof. dr. ir. Patrick Segers
Master's dissertation submitted in order to obtain the academic degree of Master of
Science in de industriële wetenschappen: elektromechanica
Department of Electronics and Information Systems
Chair: Prof. dr. ir. Rik Van de Walle
Faculty of Engineering and Architecture
Academic year 2015-2016
ii
Permission
The author(s) gives (give) permission to make this master dissertation available for consultation
and to copy parts of this master dissertation for personal use. In the case of any other use, the
copyright terms have to be respected, in particular with regard to the obligation to state expressly
the source when quoting results from this master dissertation."
1-6-2016
iii
Preface
As a preface is often to no interest for most of the readers -if they even exist- I’ve shortly listed the
people who helped me during the hatching of this thesis. Many thanks to Jurgen, Viviana, Daniela,
Frank, Stefan, Patrick for the things they know they’ve helped me with. I also would like to thank
prof. Swillens and prof. Segers for sincerely helping when problems occurred.
The choice for a certain subject is partially based on the people you meet the moment you enter a
new environment. It is unfortunate for Joris that he decided to change tracks, although it remains
strange how everyone unconsciously follows the standard road. I wish him all the best, wherever on
the globe he may be cycling.
Lastly, I want to thank my family for their unconditional support.
iv
Extended Abstract
Creating in-vitro phantoms of blood vessels to support the testing and
validation of new ultrasonic blood flow imaging techniques
Guillaume Pattyn
Abstract: Ultrasound has proven to be a versatile technique, and its applications could be broadened
by focussing on the visualization of perfusion throughout the microcirculatory system. This implies
the use of high-frequency imaging (>15 MHz) which augments the spatial resolution needed to
detect small capillaries. The main goal of this thesis is to quantify the accuracy of flow estimations
when using high frequencies, with variation in depth, vessel size, flow speed and frequency itself. To
do so, an experimental set-up has been made, based on the possibilities described in the literature.
The first part elaborates on the attempted methods to produce such a phantom, describing what
went wrong and which methods did prove to work. The second part consists of conducting
experiments on the developed phantom. Theoretical flow speeds were compared to the measured
ones when using a Visualsonics (16-32 MHz) echography machine, and a conventional GE Vivid 7 (5-9
MHz) apparatus. The results showed a high accuracy for spectral PW as well as Colour Doppler
measurements. Further testing should be done using perfused ex-vivo tissue to investigate complex
fractional structures. The third part comprises the production of a 3D model of the Circle of Willis. As
this region is crucial to the perfusion of the brain, the model could simulate problematic flow inputs,
and the results could be validated using 3D printed version of the geometry in an in-vitro setting.
Keywords: high-frequency ultrasound, microcirculation, flow phantoms, Circle of Willis
Introduction
Part I: Phantom construction
The ability to look inside the human body, without
the need for invasive surgery has always been of
great value to physicians. Many imaging
techniques rely on the use of radioactivity, contrast
agents or strong magnetic fields, making it perilous
to utilize when severe health issues are present.
One exception is ultrasound imaging which is until
present considered being safe. It is also convenient
to use because the machine can be brought to the
patient, who doesn’t need to be sedated. But, the
fairly low imaging resolution hampers the
possibility to visualize microcirculation. To tackle
this restriction, the frequency should be altered to
higher levels (>15 MHz), which also leads to a
decrease in penetration depth. This thesis focusses
on the development of a flow phantom which
mimics the microcirculation to a relevant precision.
Afterwards, this phantom is used to check the
accuracy of speed examinations by comparing the
measured speed and the theoretical values.
Materials and methods
For the production of flow phantoms, two
methods have been tested. The first method
involved the use of several very thin wires (10-300
µm). Tube connectors were inserted through the
opposing walls of a plastic container (96*96*62
mm) to leave an opening for the wires to be pulled
through. The TMM consisted at first of a
Gelatine/Agar/water mixture (resp. 8-3-89 wt.%)
which is prepared by mixing the ingredients at 90
°C for the agar, and 65 °C for the gelatine, while
continuously stirring. When the mixture was
cooled down to somewhere around 40 °C, and all
the gas bubbles had risen to the surface, it was
poured into the plastic container. The box was put
in the fridge for an overnight to fully solidify the
TMM. Then the wires were pulled out of the TMM,
hopefully resulting in hollow vessels in the TMM.
The same procedure has been done using PVA
(polyvinyl-alcohol, 99+ hydrolysed). A mixture of
v
PVA and water (10/90 wt.%) has been prepared
using the same procedure as for the Gelatine/Agar,
but the cooling stage consisted of putting the box
into the freezer for 12h. Afterwards, it could be
removed to let it thaw. The second construction
was based on the use of hollow polyimide tubes,
kindly donated by Microlumen® (100,155,165,178
µm). They were embedded in PVA which went
through two freeze-thaw cycles. Hence optimal
acoustic properties (vsound = 1535 cm/s) of the
TMM could be achieved [2].
Results
To fully understand what would happen when
extracting the wires, the GE Vivid 7 echography
machine was used to monitor the procedure
(linear array transducer at 5.5 MHz, B-Mode). The
PVA proved to be too flexible to keep the small
vessels open. They immediately collapsed under
the weight of the TMM lying above. Gelatine/Agar
was rigid enough to keep the vessels down to 120
µm open, although artefacts appeared on the
image. A syringe pump (Alaris Asena, pressure
alarm set to maximum level = 1000 mmHg) was
used to infuse the vessels, but the pressure rose
until the alarm value was reached. The second
design which uses the hollow tubes proved to be
successful. Flow (pure water) could easily be
generated.
array probes (MS250 13-24 MHz, MS550S 32-56
MHz), and a GE Vivid 7 (linear array 5-9 MHz). The
blood mimicking fluid was made using a mixture of
nylon (Orgasol Arkema Caresse R) particles, with
an average diameter of 5µm. Using a calibrated
syringe pump, low flow speeds between 1-10 cm/s
were generated inside the tubes.
Results
The measurements showed high accuracy of speed
estimation in the 155 and 178 µm vessels. The
proportional error on the estimation converged to
zero when nearing 10 cm/s. The results at 16 MHz
proved to be more consistent, whereas at 32 MHz
occasional over and underestimation could be
observed. The 100µm tube was problematic to
infuse due to constant blockage, although the few
successful measurements indicate that a stable
set-up should provide accurate results. The
precision of the colour Doppler was also analysed,
and it proved to be less precise than spectral PW
measurements. However, no aberration of factor
1.1-3.5 could be observed as described in the
literature [3]. At last, the Vivid 7 was used to check
the influence of imaging depth. An increase in
underestimation could be observed at lower
depths, probably resulting from a weaker echo
signal.
Discussion
Discussion
The vessel extraction method is by far the most
adaptable way to create phantoms. But the vessel
wall is prone to rupture, which causes premature
blockages. Also, there is no control over the
roughness of the interior of the vessels, which
could have a drastic effect on the flow pattern. The
hollow tubes proved to be successful. No excessive
reflections arising from the polyimide tubes could
be observed.
Part II: Experiments
Materials and methods
The experiments were conducted using a flow
phantom made as mentioned above (hollow tubes,
PVA 2 ftc.). The experiments were done using a
Visualsonics Vevo 2100 equipped with two linear
The measurements have shown that down to
100µm, no problems could be observed when
visualizing simple straight vessel systems. The
velocity estimates proved that no challenges will
arise at this level.
Conclusion
To fully investigate the problem, further
measurements should be done on ex-vivo tissue. A
microscope should be used to provide the ground
truth while monitoring with the ultrasound
machine. Hence, flow through complex structures
comparable to the real body can be replicated,
which is necessary to be conclusive on the ability
of estimating perfusion at micro scale.
ii
List of figures ...........................................................................................................................................iii
List of abbreviations .................................................................................................................................v
Chapter 1.
1.1
Background knowledge ................................................................................................... 2
Ultrasound technology ............................................................................................................ 2
1.1.1
Sonar................................................................................................................................ 2
1.1.2
Ultrasound waves ............................................................................................................ 3
1.2
Echography .............................................................................................................................. 4
1.2.1
The Probe ........................................................................................................................ 4
1.2.2
Imaging modes ................................................................................................................ 5
1.2.3
Flow imaging.................................................................................................................... 6
1.2.4
Safety concerns ............................................................................................................... 8
1.3
Microcirculation ...................................................................................................................... 9
1.3.1
The human circulatory system ........................................................................................ 9
1.3.2
Special phenomena ......................................................................................................... 9
Chapter 2.
Literature review ........................................................................................................... 11
2.1
Aim......................................................................................................................................... 11
2.2
Flow phantoms ...................................................................................................................... 12
2.2.1
Types.............................................................................................................................. 12
2.2.2
Tissue Mimicking Material............................................................................................. 17
2.2.3
BMF................................................................................................................................ 19
2.2.4
Pump.............................................................................................................................. 19
2.3
Chosen set-up(s) .................................................................................................................... 20
Chapter 3.
3.1
Production of the set-up ............................................................................................... 21
Set-up .................................................................................................................................... 21
3.1.1
Preparation of the TMM................................................................................................ 21
3.1.2
Blood mimicking fluid .................................................................................................... 22
3.2
Wire extraction phantom ...................................................................................................... 22
3.3
Hollow tube phantom ........................................................................................................... 23
3.4
Discussion .............................................................................................................................. 24
Chapter 4.
Experiments ................................................................................................................... 26
4.1
Pressure test procedure ........................................................................................................ 26
4.2
PW flow speed measurements.............................................................................................. 27
4.2.1
Visual Sonics .................................................................................................................. 27
i
4.2.2
4.3
GE .................................................................................................................................. 31
Colour flow speed measurements ........................................................................................ 33
4.3.1
Visualsonics ................................................................................................................... 33
4.4
Flow profile monitoring ......................................................................................................... 34
4.5
Discussion .............................................................................................................................. 36
Chapter 5.
5.1
Extra: CFD ...................................................................................................................... 37
Introduction ........................................................................................................................... 37
5.1.1
Anatomical information ................................................................................................ 37
5.1.2
From images to model................................................................................................... 38
5.1.3
Real model ..................................................................................................................... 39
ii
List of figures
Figure 1 Active sonar relies on the emission and reception of a sound wave (Source: http://www.actus.info) ..................................................................................................................................................... 2
Figure 2 (L) longitudinal wave with zones of compression and rarefaction (R) Transverse wave moving
up and down perpendicular to direction of wave [5] ............................................................................. 3
Figure 3 Reflected wave (ain=aout) and transmitted wave (ain ≠ aout) (Source:
http://www.azooptics.com) .................................................................................................................... 4
Figure 4 Influence of transducer frequency on penetration depth [7] ................................................... 4
Figure 5 Schematic representation of a transducer (Source: http://www.nano-tera.ch) ...................... 5
Figure 6 (L) A-mode imaging: a peak is observed when a change in acoustic impedance is met [8] (M)
B-mode image of a hollow vessel embedded in TMM. The reflective wall is represented by an
elevated brightness (R) M-mode image of a moving mitral valve .......................................................... 6
Figure 7(L) Continuous wave Doppler used to monitor high flow speeds at the aortic valve [9] , (R)in A
the beam and vessel are almost aligned, providing a decent frequency shift, in B and C the angle is
smaller, resulting in a weaker signal and a bigger error (D) same is in A but inverted resulting in a
negative speeds (Source: https://sonoworld.com) ................................................................................. 7
Figure 8 (L) Overview of the human circulatory system (Source: https://www.haikudeck.com) (R)
Diameter and flow speeds of different types of vessels ......................................................................... 9
Figure 9 Formation of rouleaux [16] ..................................................................................................... 10
Figure 10 the use of clutter belts to evaluate the influence of wall motion on the Doppler shift. (L)
wall and vessel are mimicked by a belt (R) vessel replicated by a hollow tube and wall motion
provided by the belt [17] ....................................................................................................................... 13
Figure 11: (L) ex vivo artery made by LifeTecGroup (R) Comparison between the burst pressure of the
same artery in and ex vivo (after preservation method) [18] ............................................................... 14
Figure 12 (L) .STL model (R) printed version of the circle of Willis ....................................................... 15
Figure 13 Scaffold made using two-photon polymerisation (2PP) [23] ................................................ 15
Figure 14 Solid ground curing process (Cubital), where the mask is translucent light-induced
polymerisation will occur [21] ............................................................................................................... 16
Figure 15 Vertebra submerged in a TMM based on gelatine/agar [29] ............................................... 18
Figure 16: Influence of the amount of freeze-thaw cycles on (A) the attenuation (B) the speed of
sound [2] The red line indicates the value of tissue, an optimum is found at 2 ftc ([30] ..................... 18
Figure 17 (A) Tensioned wires and connectors fitted through the wall, ready to pour the TMM (B) box
from A filled with G/A, wires extracted and ready to perfuse .............................................................. 22
Figure 18 Rupture due to elevated pressure, initiated at the tube connector ..................................... 23
Figure 19 Final model before pouring the PVA ..................................................................................... 24
Figure 20 B-mode image of 1 longitudinal and 6 cross sections of 200µm silicon tubes. Distinct
lumens can hardly be observed, and high reflections could impede precise measurements .............. 25
Figure 21 Variation of the pressure in time after setting a certain flow rate to the system.
Convergence takes more than 20 minutes ........................................................................................... 26
Figure 22 Pressure test in stabilized conditions. No pulsatility can be detected. ................................. 27
Figure 23 Post-processing of a sonogram with the Vevo software package. The mean value can be
traced automatically.............................................................................................................................. 28
Figure 24 Measured speeds vs. speed of infusion of a 178µm vessel .................................................. 29
iii
Figure 25 Proportional error on speed measurements, both trend lines convergence to zero. .......... 30
Figure 26 Speed measurements within a 155 µm tube ........................................................................ 30
Figure 27 Proportional error on the speed measurement when infusing a 155 µm tube. At both
frequencies, a convergence towards zero can be observed for higher flow speeds. ........................... 30
Figure 28 Flow speed measured through a 100 µm ID tube, a large spread in results can be observed
due to clots shutting off the vessel ....................................................................................................... 31
Figure 29 Sonogram being processed using the Asklepios Matlab tool................................................ 32
Figure 30 Measured vs. True flow speed at 1.5cm (red) and 3cm (blue) ............................................. 32
Figure 31 Colour Doppler speed estimation in a 178 µm tube ............................................................. 33
Figure 32 Comparison between the flow profiles (speed vs. pixels) for different frequencies and
different tube diameters ....................................................................................................................... 35
Figure 33 The colour Doppler image taken at (L) 16 MHz confines the flow quite precisely within the
tube. At (R) 32 MHz a reflection of the tube can be observed. ............................................................ 35
Figure 34 Anatomical picture of the circle of Willis [1] ......................................................................... 37
Figure 35 (L) before trimming the excess branches (R) after trimming and smoothing ....................... 38
Figure 36 (up) CFD simulation of flow through the circle of Willis (down) zoomed image showing the
individual vectors .................................................................................................................................. 39
iv
List of abbreviations
SONAR
Sound navigation and ranging
US
Ultrasound
COW
Circle of Willis
PRF
Pulse repetition frequency
B-mode
Brightness mode
CW/PW
Continuous and pulse wave
LVRej
Low-velocity rejection
PZT
Lead (Plumbum) Zirconite Titanate
OD
Outer diameter
PVA
Polyvinyl-alcohol
G/A
Gelatine agar
ID
Inner diameter
TMM
Tissue mimicking material
BMF
Blood mimicking fluid
CFD
Computational fluid dynamics
2PP
Two-photon polymerisation
SLA
Stereolithography
ftc
Freeze-thaw cycles
CFI
Color flow imaging
v
Introduction
War can never be accepted, but it has proven to accelerate research on technologies which can be
used to the military’s advantage. At the end of the interbellum, the U.S. feared the possibility of the
Germans being able to produce a nuclear weapon. In the shadows, leading scientists of the 20th
century theorized the first nuclear bomb, hoping to produce one before the enemy did. Although the
devastating capacities of such a weapon are enormous, the fundaments of today’s nuclear energy
can be traced to this project. The same applies to the use of ultrasound for object detection.
Researchers made little progress until, during the First World War, the threat of submarine attacks
became plausible. Huge investments were made, quickly resulting in devices such as the hydrophone
and SONAR. In 1953, ultrasound made its entrance in the field of medical imaging. J.J. Wild and J.
Reid visualized a 7mm thoracic tumour using their machine: “the echographe”. Since that moment,
ultrasound evolved continuously. Features such as Doppler, colour Doppler, and 3D were
implemented, enlarging the scope of application to a major extent. In contrast to other techniques
such as MRI and CT, ultrasound is considered to be safe, and the examination can be done without
moving or sedating the patient.
The aim of this thesis is to study the feasibility of estimating blood flow and perfusion at a
microcircular scale, using new ultrasound techniques. This could be of great use for preterm born
babies who tend to suffer from cerebral lesions and need constant monitoring in the early stages of
life.
To validate these new techniques, flow phantoms mimicking the human microcirculation had to be
designed. In the first part of this text, the fundamentals of ultrasound imaging and the human
microcirculation are explained. Next, a literature review elaborates the possibilities and restrictions
of flow phantoms. The second part envelopes the experimental part. Some disused setups will be
discussed, explaining why they failed and to which extent they could be used. Other setups, which
did work, were used to compare measured flow speeds with the actual speed. Based on these
results, conclusions about the feasibility of perfusion imaging are made. The last part consists of
computational models related to brain flow imaging. A CAD model of the Circle of Willis has been
developed using MRI data and the Mimics software. The geometry was optimized and meshed to
perform CFD calculations using Ansys Fluent.
1
Chapter 1. Background knowledge
It should be stated that this introduction is by no means an exhaustive explanation of physics, nor a
guide to the human anatomy. However, it should provide the reader with enough information to fully
understand the challenges, techniques and solutions described in this thesis.
1.1 Ultrasound technology
The term ultrasound is derived from the Latin words ‘ultra’ and ‘sonus’, which can be translated as
‘beyond the sound’. This refers to the frequency surpassing the human hearing range which is
situated between 20-20000 Hz. In this section, the basic principles of sound waves will be elaborated
using SONAR as an example.
1.1.1 Sonar
Sonar (Sound navigation and ranging) was used by the American army since the 1930s for
underwater object detection. As submarine attacks became a considerable threat, the Navy needed
an instrument to scan the sea. To do so, they mounted a sender to the hull of their ships which
creates a pulse (ping, Figure 1) of acoustic energy propagating through the water. When this wave
reaches an object, an important amount of energy will be reflected. This reflection will be detected,
thus discovering a nearby enemy.
Figure 1 Active sonar relies on the emission and reception of a sound wave (Source: http://www.act-us.info)
Knowing the bulk modulus K and the density ρ of (saline) water, the wave speed c can be calculated.
(1)
√
The sender, which often also acts as a receiver, acquires the reflected wave. The interval between
the emission and reception is the amount of time (t) which the wave needs to cover twice the
distance between sender and object. Hence - knowing the wave speed - the distance x between the
ship and submarine can be calculated.
(2)
Despite being a century old technology, more complex forms of SONAR a still being used for e.g.
scanning the ocean floor and creating 3D plots of the seabed. Furthermore, the intensity of the
received signal provides enough information to determine the composition of the ocean floor [4].
2
1.1.2
Ultrasound waves
1.1.2.1 Propagation
When swimming in the sea, your body goes up and down following the movement of the water. This
motion can be described as a transverse wave (Figure 4), which moves perpendicular to the direction
of the energy transfer. A longitudinal wave is an alternation of high-pressure zones (compression)
followed by low-pressure zones (rarefaction), propagating along the direction of energy transfer.
Figure 2 (L) longitudinal wave with zones of compression and rarefaction (R) Transverse wave moving up and down
perpendicular to direction of wave [5]
The transverse part of a sound wave is known as a shear wave, and is the basis of ultrasound
elastography. This fairly young technique maps the stiffness of tissue, making it possible to distinct
cancerous from normal tissue [5]. However, in the scope this thesis, shear waves will be neglected
because US flow imaging is based on the longitudinal wave.
The longitudinal part can be mathematically described as [6]:
(
with
)
(3)
being the pressure amplitude
(4)
From equation (4) we can see that the longitudinal wave is a cosine function varying with the
wavenumber (
), angular frequency (
), place (x) and time (t). The pressure amplitude
depends on the amplitude of displacement (A), the density (ρ), the frequency and the wave speed
(c).
1.1.2.2 Scattering
The submarine (mentioned above) can be seen because it reflects some of the energy emitted by the
sender. The driving force behind reflections is the change in density and compressibility (bulk
modulus) of the medium through which the waves propagate. When the wave hits the metal wall
(larger than the wavelength), it reaches a boundary between two densities. Some of the energy will
bounce back, leaving the surface with the same angle as the incident wave (law of reflection). This is
called specular (mirror-like, Figure 3) reflection. Another portion of the energy will be transmitted
through the new medium. The ratio of transmission and reflection depends on the difference in
acoustic impedance of the media. This acoustic impedance Z is defined as the product of the density
times the wave speed in the medium.
3
Figure 3 Reflected wave (ain=aout) and transmitted wave (ain ≠ aout) (Source: http://www.azooptics.com)
The transmitted wave will pass through a material with different bulk ratio and density, resulting in
an altered speed and direction of propagation. This is known as refraction. With respect to
echography, a third phenomenon has to be addressed, which occurs when a wave hits an object
smaller than the wavelength. Instead of being reflected, the energy is scattered in every direction.
This is known as diffuse scattering.
1.1.2.3 Attenuation
Lossless energy transportation is impossible, and the same applies to the propagation of a sound
wave. The longitudinal wave will cause vibrations, which will partly be converted to heat. The longer
the distance covered by the wave, the more energy it will have lost, thus weakening the amplitude of
the received signal (Figure 4). The following formula describes the interaction between the depth z,
the frequency f of the wave and the coefficient of attenuation α of the medium.
( )
(5)
This thesis focusses on the visualisation of microcirculation using higher frequencies. Looking at
equation 5 and Figure 4, it is clear that the influence of attenuation will have to be examined,
especially when monitoring deeper structures.
Figure 4 Influence of transducer frequency on penetration depth [7]
1.2 Echography
1.2.1 The Probe
At the heart of the probe lie a series of piezoelectric crystals. These crystals have the special ability to
deform when an electrical voltage is applied, and to generate an electrical signal due to external
4
forces. The cause of this effect can be found in the lattice structure of such a crystal, which consists
of small electrical charges oriented in such a way that electrostatic equilibrium is achieved. Due to
external forces, negatively and positively charged zones will be induced, generating a voltage.
Mutatis mutandis, when a voltage is applied the lattice will deform to restore the equilibrium. This
deformation can produce a wave. Hence, an electrical signal can be used to generate an acoustic
wave and the reflection can be detected because it deforms the crystal. Nowadays, the most
common material used as piezoelectric crystals is Lead Zirconate Titanate (PZT).
C
A
B
Figure 5 Schematic representation of a transducer (Source: http://www.nano-tera.ch)
A specific name has been assigned to every direction in order to avoid any ambiguity. Axis A (Figure
5) is the axial direction, parallel to the trajectory of the beam. Axis B and C are respectively known as
the lateral direction and elevation.
1.2.2
Imaging modes
1.2.2.1 Grayscale imaging
A-mode
The amplitude of the reflected wave is plotted in function of the depth. This is used for e.g.
echoencephalography to detect lesions inside the brain.
B-mode
Brightness mode (or 2D mode) produces two-dimensional images with the brightness varying with
the amplitude of the echo.
M-mode
Motion mode is a special form of B-mode which can be compared with a photo finish. A single scan
line (Figure 6, red line) is plotted through time, hence showing the movements of the observed
structure.
5
Scanline
Motion on the line in
time
Figure 6 (L) A-mode imaging: a peak is observed when a change in acoustic impedance is met [8] (M) B-mode image of a
hollow vessel embedded in TMM. The reflective wall is represented by an elevated brightness (R) M-mode image of a
moving mitral valve
1.2.3 Flow imaging
Flow imaging is powerful and safe, if practiced within its limits. Otherwise, these methods can be
prone to misleading results.
Doppler Effect
When sitting on a train, speeding towards the ringing bell of a level crossing, the frequency of the
observed sound will change. The pitch will be high when getting near to the crossing, and low once
you passed the bell. A simple explanation for the Doppler effect, named after an Austrian physicist
who was observed this when looking at the stellar motion, is the following:
Suppose that at regular interval gnomes (i.e. positive amplitude peaks) are jumping off a moving
truck. The speed of the truck vt is constant, and the gnomes are always running at vg towards the
observer. The truck approaches the observer and gnomes are sprinting towards him. Because the
gnomes are faster than the truck, the distance between the gnome and truck will increase at a rate vd
equal to their difference in speed.
(6)
Once the truck passes the observer, gnomes are still running towards him, thus moving in the
opposite direction of the truck. Hence, the distance between the truck and the gnome will increase
at a rate which is the sum of both speeds.
(7)
The gnomes, representing the positive peak of a sound wave, will be closer together when going
towards the observer, resulting in a higher frequency. When moving away, the opposite effect will
occur.
(8)
6
Continuous wave Doppler
Blood flow can be measured applying the Doppler Effect to ultrasound waves (equation 9). One
crystal of the probe continuously emits a wave, of which the frequency is known, towards the
vessels. When a change in acoustic impedance is met, the wave will partly be reflected and received
by another crystal, which is constantly listening. If the reflection origins from moving particles, such
as red blood cells or vibrating muscles, there will be a change between the emitted and received
frequency. This can be used to calculate the speed at which the particles were moving. A
disadvantage of using continuous wave Doppler is the inability to determine the speed at one
specified depth, as illustrated in Figure 7. The received data will always consist of all the
backscattered information on the scan line. But to its advantage, CW is not bound to the restrictions
of the Nyquist theorem (discussed below), making high speed measurements possible. This is why it
is still being used to assess high flow speeds at cardiovascular level.
(9)
An important variable is the angle between the emitted wave and the vessel because this technique
measures the difference in speed relative to the direction of the beam. If the beam and the vessel
are not aligned, a factor cos(θ) is needed to calculate real speed of the blood flow in the direction of
the vessel.
Figure 7(L) Continuous wave Doppler used to monitor high flow speeds at the aortic valve [9] , (R)in A the beam and
vessel are almost aligned, providing a decent frequency shift, in B and C the angle is smaller, resulting in a weaker signal
and a bigger error (D) same is in A but inverted resulting in a negative speeds (Source: https://sonoworld.com)
Pulsed wave imaging
Pulsed wave (PW) flow imaging has often been –incorrectly- denoted as Doppler imaging. Each
crystal of the transducer acts successively as a sender and receiver, firing short bursts of acoustic
energy towards the tissue. Because the speed of sound through the human body is known, the
moment of receiving a pulse arising from a certain depth can be calculated. Hence, the flow speed
within a volume (i.e. sample volume) at a specific depth can be measured. This technique is called
“time-gating”.
The Doppler shift has to be sampled at least twice as frequently as the highest frequency shift
measured in the volume. The frequency at which the pulses are emitted is known as the pulse
7
repetition frequency (PRF). Numerous examples of PW measurements can be found in chapter 4.
Further information about the physics behind PW imaging and ultrasound in general can be found in
the doctoral thesis of prof. Swillens. [10]
1.2.4 Safety concerns
Ultrasound is used for imaging unborn babies because is it safe. Or isn’t it?
As mentioned in 1.1.2, this technique relies on the propagation of a sound wave through the body.
Travelling waves transport energy, and the way this acoustic energy dissipates or is converted to
other forms of energy is crucial to the safety of ultrasound. There are two major physical effects
which should be considered.
First is the thermal effect of US. A wave propagating through tissue loses its energy, and this loss is
converted to a certain extent into heat. The heating effect depends on the heat absorption
coefficient of the tissue, and proves to be maximal in bone structures. However, it is daunting to
predict the overall rise in temperature, partly due to the many variables (power, focus etc.). The
heating effect of ultrasound examinations on brain tissue has been studied and showed an increase
in surface temperature of 4,3 °C after 120s. Due to the dense vascular system deeper inside the
brain, the increase in temperature was limited to 1,2 °C. [11].
Second is the possibility of cavitation. This is the formation of a gas bubble due to a decrease in
pressure arising from a rarefaction zone. The explosion of such a bubble can cause sudden high
pressures, which may be hazardous.[12] A recent study has shown to possibility the use cavitation as
an alternative to the conventional lithotripsy treatment. After injection microbubbles, they could be
detonated using ultrasound waves. These explosions could fracture kidney stones [13]. In the scope
of ultrasound imaging, cavitation is not desirable.
Older large-scale studies have found no correlation between echography imaging during pregnancy
and later health problems when the child grows up. However, the power output of the newer
generation of ultrasound machines is higher, which implies the need of further testing. When using
higher frequencies, the possible effects should also be examined. But, higher frequencies should not
be confused with higher power outputs.
8
1.3 Microcirculation
The use of Doppler echography has been used extensively to visualize major arteries, quantify blood
flow and to detect plaques. This can be done because the signals arising from the blood flow are
noticeably differing from tissue and muscular noise. When observing microcirculation, these
differences will decrease, leading to possible ambiguity.
1.3.1 The human circulatory system
Our body consists of a several billion to trillion cells which can be seen as small factories. Each cell
must receive enough raw materials to maintain its homeostasis, and similarly get rid of all its waste
products. Blood can be seen as the medium of transportation, and it is propelled by the movement of
the heart. An overview of the circulatory system can be seen in Figure 8.
Aorta
Common
carotid artery
Capillaries
Diameter
2-3 cm
0.55-0.78
cm
5-300 µm
Flow speed
> 40 cm/s
15-25 cm/s
< 3 cm/s
(non-pulsatile)
Figure 8 (L) Overview of the human circulatory system (Source: https://www.haikudeck.com) (R) Diameter and flow
speeds of different types of vessels
Deoxygenated blood (blue) arrives at the right atrium, passes through the right ventricle and enters
the pulmonary artery. When blood reaches the lungs, CO2 is released and O2 is added. The
oxygenated blood (red) flows back to the heart, closing the loop known as the pulmonary circulation.
Then, the aortic artery leaves the heart, carrying oxygen to the organs. At this level, the flow is highly
pulsatile, following the contraction and relaxation of the cardiac muscles.
The aorta branches into smaller arteries and arterioles. As the heart is more remote, the pulsatile
factor decreases, as well as the flow speed. The smallest roads in the human circulation are the
capillaries. These tiny vessels communicate with the surrounding tissue, exchanging nutriments
through the thin wall.
At macrocircular scale, Doppler echography provides us with much information concerning the flow
speed, vortices, valve malfunctions etc. Visualization of microcirculation could give us feedback on
the perfusion of e.g. the brain tissue.
1.3.2 Special phenomena
A well-known example of non-Newtonian behaviour of fluids is the “water-maize” experiment. While
it is impossible to walk on water, the water/maize mixture will become “solid” when a force is
9
exerted on it. This is a result of an increasing viscosity due to a soaring shear stress. Materials
behaving this way are known as shear-thickening or dilatant.
Blood behaves in the opposite way, thus getting thinner (less viscous) when shear rate increases[14].
This phenomenon was first observed by the Swedish scientists Fahraeus en Lindqvist. They forced
blood through a tube with a diameter of 250µm and observed the flow pattern. What they saw was
the formation of a RBC free layer near the vessel wall. Because of this thin layer of plasma the
“oriented” blood will have lower flow resistance than homogeneous blood. The net effect will be an
overall decrease in viscosity. There are two causes which can favour a cell free layer[15]:
 Silberberg effect: neutrally beyond particles in a suspension tend to be pushed towards the
centre of a vessel as a results of several hydrodynamic interactions
 Statistically, the density of RBC will be higher at the centre of the vessel
Another shear thinning factor is related to the formation of red blood cell stacks, also known as
“rouleaux” (Figure 9). This effect is often related to several pathologies such as sickle cell disease and
malaria, but it also occurs naturally as a reversible aggregation process. When shear rates increase,
these “rouleaux” tend to dissolve, leading to a less viscous fluid. [16]
Figure 9 Formation of rouleaux [16]
10
Chapter 2. Literature review
2.1 Aim
The ability to visualize perfusion at a microcircular scale could be a major asset to the already
indispensable ultrasound technique. It would broaden its use and could be of particular interest to
brain flow imaging. However, the methods which are being used to improve the sensibility and
resolution will give birth to some new problems (as discussed below), which will have to be examined
thoroughly.
Capillaries and arterioles are unmistakably smaller than arteries, and necessitate higher spatial
resolution and backscattered power to be observed. This can be achieved by enhancing the imaging
frequency from 5-8 MHz, which is the current standard for flow measurements, to 15MHz and
higher. Sadly, the increase of frequency comes at the cost of an exponential decrease in penetration
depth (cf. 1.1.2.3). The trade-off between depth and resolution has to be quantified before any
prediction can be made on the accuracy of perfusion estimates.
The challenges aren’t limited to physical problems. Signal processing will also need to be reevaluated. The amount of data received from the backscattering is enormous and has to be filtered.
Arterial flow differs significantly from tissue motion and vibrations induced by the operator’s hand,
making a velocity based rejection of clutter signals possible. If the same rejection methods were to
be applied when observing low flow rates in capillaries, the data of interest would be rejected
because the speed of scatterers and clutter signals overlap. Hence, new filters have to be designed.
Furthermore, the blood flow in capillaries can’t be translated as a downscaling of the flow in e.g. the
aortic artery. As mentioned in 1.3.2(“special phenomena”), red blood cells will form stacks which can
cause moderate reflections at one moment and almost none in the next frame.
To address these problems, a multiphysics model has been designed. One facet of this model
focusses on the simulation of ultrasound echoes using a virtual transducer (Matlab). The other facet
calculates the flow pattern within the structure, depending on boundary conditions applied to the
model (ANSYS Fluent). Combining these two, scatterers can be moved along the flow lines (derived
from the CFD), and their echoes can be visualized. However, computational models always suffer
from (over)simplification of complex fluid behaviour, as well as geometrical imperfection. And to be
conclusive, their accuracy has to be validated against a realistic set-up.
Hence, the need for a flow phantom arises.
This literature review attempts to list the possibilities for making a model mimicking the properties
microcirculation to a relevant precision.
11
2.2 Flow phantoms
In its simplest form, a flow phantom is a hollow geometry through which flow is sent. In contrast to
in-vivo or ex-vivo (i.e. on real tissue) measurements, in-vitro (i.e. model) gives the researcher full
control over the flow speed, pressure, geometry etc.
The way such a phantom will be made strongly depends on the system which needs to be replicated,
and what needs to be measured. Our phantom should meet the following criteria:
-
Mimic capillaries and arterioles
-
Diameter between 5-300 µm
-
Low flow speeds 0.1-10 cm/s
-
US compatible -> echogenic material
-
Low budget
-
Different set-ups possible
When going through the published literature, it became clear that numerous methods exist to
produce phantoms for macro scale. However, can these be translated to microcirculation?
Depending on the design, blood and tissue will need to be mimicked using a blood mimicking fluid
(BMF) and tissue mimicking material (TMM). The next section will cover the different types of flow
phantoms which have been made until present, and how TMM and BMF can be incorporated to
mimic an in-vivo setting.
2.2.1 Types
String and belt phantoms
When thinking of flow phantoms, the necessity of using a fluid is the first element that comes to the
mind. However, a moving solid can create a similar echo (to a certain extent, as discussed below).As
the name suggests, a string or belt is submerged in a fluid and connected to an electric motor. The
string is covered with scattering particles to provide a strong echo.
The wave speed through the fluid must be near 1540 m/s (speed in human tissue) to minimize errors.
These phantoms (fig 1) have been successfully used to assess the influence of wall and tissue motion
- known as clutter - on the Doppler shift.[17]
12
Figure 10 the use of clutter belts to evaluate the influence of wall motion on the Doppler shift. (L) wall and vessel are
mimicked by a belt (R) vessel replicated by a hollow tube and wall motion provided by the belt [17]
This kind of phantom fails to replicate the parabolic flow pattern which can be observed in-vivo
because the inner structure of the belt moves at the same speed as the outer regions.
+
A good option when solely quantifying flow speeds
-
No fluid dynamics to analyse
-
No control over scatterer concentration in time
Ex-vivo material
It is possible to use excised (Figure 11) arteries instead of trying to replicate them. This can be of
utmost importance when the elastic properties of the tissue have a big influence. However, it is a
very cumbersome procedure. Donor tissues have to be found and treated for storage. Otherwise, the
material’s properties will deteriorate fast, making it almost impossible to do exhaustive
measurements. Even if the tissue is kept at optimal conditions, the elastic properties could change.
Figure 2 shows the difference in burst pressure of an epigastric vein, in vivo and ex vivo (fused invivo). A significant difference in burst pressure can be seen. [18]
Muscle
Vessel
Blood
Density
(kg/m³)
1041
1066 ± 10
1060
Velocity
( m/s)
1595 ± 20
1616 ± 23
1587 ± 3
Attenuation
(dB cm-1 MHz-n)
1.47
1.026
0.17
Table 1 : In vivo characteristics of different types of human tissues [19]
Reperfusion of capillary ex-vivo networks has been done before, and flow could easily be observed.
[20] However, it is daunting to assess which volume is passing through each individual capillary when
infusing at artery level. But, the combination of assessing the speed using ultrasound and an optical
microscope to watch the particles travel through the system could be the ground truth solution.
13
+
Good elastic properties
+
Perfect geometry and attenuation
-
Biological degradation after short time
-
Poor control on speed in individual capillaries
Figure 11: (L) ex vivo artery made by LifeTecGroup (R) Comparison between the burst pressure of the same artery in and
ex vivo (after preservation method) [18]
Tubing
Hollow tubes can be used to replicate vessels, but the design will be limited to straight lines without
branching [3]. To be representative as a microcirculation mimic, the tubes must have an internal
diameter (ID) between 50-300µm. Moreover, they must be made from a material which has good
echogenic properties.
Common silicon or PTFE tubing can rarely be found with an ID smaller than 0.4 mm, which is at least
the double of what is needed. Smaller dimensions can be found when looking at Tygon-based tubes,
but the ID still hovers around 0.25 mm. The thinnest tubes are used for gas chromatography, with
ID’s as low as 0.06 mm. But, they come at a very high cost, with the uncertainty if they are echogenic.
To save expenses, different tube manufacturing companies will be addressed to ask if samples can be
donated for research purposes.
+
Widely available and economic
+
Several set-ups are easily produced
-
No complex geometries with continuous cavity
3D printing
Rapid prototyping has been booming since the beginning of the 21st century. New machines and
methods made it possible to print in a variety of materials. A 3D model, based on CT or MRI images,
can be made easily using mimics software. The generated STL file (Figure 12) can be post-processed
and directly send to the printer, resulting in an almost perfect duplicate of the region of interest.
14
Figure 12 (L) .STL model (R) printed version of the circle of Willis
Stereolithography (SLA) and fused deposition modelling (FDM) are two common technologies within
the rapid prototyping industry. They are used for making prototypes at early stages of the developing
process. Hence, faults and design flaws can be detected by visually inspecting a real model. Sadly, the
precision required for making a microcirculation phantom is beyond the capabilities of these
technologies.[21]
Two-photon polymerisation (2PP) is a state of the art technology, which could be used to create
objects to nanometre precision. It uses very powerful lasers (up to 200 GW) to induce polymerisation
in photosensitive materials. In contrast to normal polymerization, 2PP uses near infrared light which
can travel through the photosensitive material. This makes printing in 3D possible, instead of using a
layer-by-layer approach. Because 2PP is still in the experimental phase, this method can’t be applied
to this thesis. [22, 23]
+
Perfect geometry and different materials
-
Technological limits to precision
Figure 13 Scaffold made using two-photon polymerisation (2PP) [23]
15
Dialysis cartridges
Cartridges have been used as capillary flow phantoms to evaluate the behaviour of contrast agents at
low flow speeds. A cartridge is a bundle of several thousands of hollow tubes with an internal
diameter in the range of 200 µm. It could be used to evaluate averaged flow speeds in a dense
network. But because the slightest dimensional error can cause higher or lower flow resistance, it
can’t be used as a reliable indicator of flow speeds in individual vessels (cf. reperfusion of tissue). [24]
+
Dense network of fibres and easy to attach to circuit
-
Poor control of flow speed
Microfluidic systems
This technique is used to make tiny vessels, down to 50-100 µm. To produce the vessels,
photopolymers are used which cure when exposed to light (Figure 14). Starting from a 3D CAD
model, a mask is manufactured for each layer of the curing process. Hence, where the mask is
transparent, polymerization can take place. The resulting product which has the positive geometry is
known as the “master”. This can be used to create the negative geometry in PDMS
(poly(dimethylsiloxane)).[25] [26]
Figure 14 Solid ground curing process (Cubital), where the mask is translucent light-induced polymerisation will occur
[21]
These models have many applications in the field of medicine and biochemistry.
+
Ultra high precision
+
Perfect geometry
-
Expensive
16
2.2.2 Tissue Mimicking Material
The system has to be embedded in a tissue-like material, referred to as the Tissue Mimicking
Material (TMM). To estimate perfusion, the attenuation coefficient and the speed of sound through
the TMM will be important. The average speed of sound in human tissue is 1540 m/s, and this value
is also set in the echography machine. When differing significantly from this value, the resulting
image will have a dimensional distortion. (cf. equation 2)
The attenuation in the TMM is an important variable when monitoring deeper structures. If it is too
low, the received echo from the phantom will be higher than would be the case with in vivo material.
These results could be rescaled using equation 5, if the attenuation coefficient of the material is
known. However, it is preferred to start with a material which has acoustic properties close to human
tissue.
Table 2 Characteristics of TMM candidates. The velocity should be near 1540 m/s to replicate the properties of in vivo
tissue [19]
Gelatine/agar
Gelatine and agar are ingredients which are used to make gel-like structures for nutritional purposes.
Because a mixture of both materials with water has good acoustic properties, it has been used for
several tissue mimic applications, of which some will be discussed. Gelatine/agar was used as a wall
mimic, to make carotid flow phantoms for strain imaging research.[27] The fabrication started with a
metal rod which was centred in a cylindrical mould. Then, the liquid gelatine/agar (resp. 8%/3%) mix
was poured around the rod. After putting it in the fridge for 12h, the rod was extracted. Hence, a
hollow carotid remained with a wall thickness of 5mm.
This mixture can also be used as a tissue surrounding ex vivo or artificial structures such as vertebra
(Figure 15) [27] or catheters[3]. A third possibility is to let the TMM act as wall and tissue mimic [28].
These phantoms are known as “wall-less”.
17
Figure 15 Vertebra submerged in a TMM based on gelatine/agar [29]
Gelatine melts at 60 °C and agar at 93 °C, and they are both easily available. This makes them
accessible to use. But, biological degradation can occur fast if no anti-bacterial agents are added. Jia
Wei Li et al. reported the use of chlorhexidine as an anti-bacterial agent.
PVA
Polyvinyl-alcohol (PVA) is a material with unique properties. When dissolved in water, it forms a gel
with varying degree of cross-linkage (crystallinity) depending on the number of freeze-thaw cycles
(ftc). As can be seen in fig. 5, the amount of ftc will alter the attenuation and speed of sound. When
comparing the acoustic properties of human tissue, as listed in Table 1, with Figure 17, it can be seen
that two freeze-thaw cycles give the TMM an optimal speed of sound[2]. No predictions can be made
on the attenuation because the frequency used during the measurements will be far greater than the
values mentioned in Figure 17.
As mentioned above, gelatine agar can be used to mimic different structures. The same design can
be made using PVA. However, PVA is much more tedious to process due to its difficulty to dissolve in
water, and the time consuming freeze-thaw cycles
A
B
Figure 16: Influence of the amount of freeze-thaw cycles on (A) the attenuation (B) the speed of sound [2] The red line
indicates the value of tissue, an optimum is found at 2 ftc ([30]
18
2.2.3 BMF
Doppler flow imaging in vivo relies on the detection of moving red blood cells. To replicate these
echoes in artificial blood, particles have to be added to the fluid. Most BMF’s are water based,
however, oil based emulsions have also been reported [31]. Real blood can also be used, but as with
ex vivo material, there are many complications. Blood will only be representative for in vivo blood
when it is heated to 37°C. The red blood cells are easily damaged, and there is always a biohazard
risk (i.e. infection by pathogens in the blood). Hence, it is preferred to use alternatives. The BMF
needed for this thesis should meet following criteria:
-
Particle size: smaller than 30 µm
-
Density: neutrally buoyant
-
Concentration: high enough to be normally distributed (central limit theorem)
-
Viscosity: around 4 mPa.s
When observing flow patterns or wall shear stress, the viscosity of the BMF must be similar to real
blood viscosity. This value can be altered by adding glycerol to the mixture.[32]
Table 3 Properties of the most common particles used as scatterers in a BMF preparation [19]
Nylon particles are ideal due to their low dimensions and appropriate density. But, as with the other
particles, they have to be filtered. Otherwise, contaminants or bigger particles could block or distort
the flow.
Table 2 omits starch as a BMF candidate. It has a density of 1.5 g/cm³ and the particle size varies
between 2 µm (rice starch) to 100µm (potato starch) [33].
2.2.4 Pump
When accidentally cutting in your finger blood won’t gush out, but a drop of blood will steadily grow.
This is because the heart is remote compared to the capillary system, and the attenuating effect of
arteries, resulting in almost no pulsatility in the blood flow.
19
The pump propelling the fluid has to be able to deliver very low and linear flow rates with high
precision. This can be achieved by placing a reservoir at a certain height, and measuring the mass of
the outlet flow in a measured time frame. A second possibility is using a syringe pump. This device is
designed to administrate a precise flow rate of medicinal drugs. A third option is by using a roller
pump designed for handling small volumes with high precision. Camfermann et al. report that high
friction between plunger and wall when using a plastic syringe can be problematic. Instead moving
linearly, the pressure in the syringe increases until the friction can be overcome, resulting in a
flow/pressure peak. Hence, a glass syringe is preferred.[3]
2.3 Chosen set-up(s)
The feasibility of making wall-less flow phantoms with micrometre range lumens will be checked
because solid wires can be found in the almost every dimension. Hence, models could be made fast
and cheap, which is useful when tiny changes have to be made. If unsuccessful, tubing will be used.
These have the disadvantage of being a lot more expensive and a lot less adaptive in the scope of
available diameters.
As for the TMM, both Gelatine/agar and PVA will be tested. The preparation of PVA is timeconsuming, but the acoustic properties are far superior to these of Gelatine/agar. PVA is also less
prone to biological degradation. However, a homogenous solution of G/A is easier to make due to
the low melting point of both components.
The initial BMF will consist of starch particles because it is safe to use. When the set-up is in its final
stage, ready for the decisive experiments, the starch will be replaced by orgasol (nylon) particles. Not
only are they smaller, but the consistency of particle size is superior compared to starch.
The propelling apparatus will be either a syringe pump or a roller pump. This way, the displaced
volume, as well as the flow speed can be determined with high precision.
20
Chapter 3. Production of the set-up
The literature provided some techniques to make flow phantoms. Because many of these methods
are applied for the production of larger vessels, it remains uncertain if downscaling to a
microcirculation level will be possible. Hence, the experimental part can be seen as a feasibility
“study” within the feasibility study of monitoring perfusion.
3.1 Set-up
In this section, every component of the flow phantom will be discussed. Due to failure, many designs
have been tested and changes have been made accordingly. However, they all rely on the same TMM
and BMF. This is why these will firstly be explained in general. Afterwards, every flow phantom
design will be covered separately, in detail.
3.1.1 Preparation of the TMM
Ultrasound waves can’t travel through air. This is why some material is needed between the flow
vessel and the transducer. It is preferred that the acoustic properties of this material match those of
the human tissue. Hence, dimensional distortion of the image and excessive attenuation of the
signals can be avoided. The literature review revealed two valuable substances to use as a TMM,
being PVA and a mixture of Gelatine and Agar. The following paragraph embodies the preparation of
these materials.
PVA
The optimal acoustic characteristics of polyvinyl-alcohol are achieved when (Table 1) using a 90/10
(wt. %) mixture of H2O and PVA-grains (Aldrich chemistry, 99%+ hydrolysed). The mixture was slowly
heated to 95 °C for at least 40 minutes while continuously stirring to avoid any burning at the bottom
of the glass beaker. During this step, a lot of water evaporates, which will need to be compensated
for by adding extra water until the initial concentration is restored. Once the solution got translucent,
meaning the PVA was fully dissolved, it was left to cool to ambient temperature. During the cooling
stage, the container must be covered to prevent excessive evaporation to occur, which could alter
the initial concentration of PVA. After waiting at least 30 minutes, a skin will form on the cooled
solution. This needs to be removed before pouring the mixture into the box.
When no more gas bubbles could be seen inside the phantom, it was put in the freezer for at least
12h. After letting it thaw, the wires were removed thus forming a cavity inside of the TMM.
Note: if the dimensions must remain unaltered, it is preferable to use a mixture of glycol and H2O
(resp. 40/60 wt.%) instead of only using H2O. The addition of glycol lowers the melting point which
results in less distortion during the freezing stage.
Preparing the Gelatine/Agar
A similar procedure was used to prepare the Gelatine/Agar TMM. A mixture of 356 g water and 12 g
agar (Agar Agar, Vahiné) was heated to 90 °C until a homogeneous melt was obtained. Afterwards,
the temperature was lowered to 60 °C before adding 32 g of gelatine (Dr Oetker). Once the gelatine
was fully dissolved, it was cooled to ambient temperature using the same procedure as described in
the PVA section. However, no freezing stage was involved. The sample was put in a fridge overnight
before using it. A summary of the composition of both PVA and G/A can be found in Table 4.
21
PVA phantom
PVA
Water
Glycol
Wt. %
10
90 (54*)
(36*)
Gelatine Agar
Gelatine
Agar
Water
Wt. %
8
3
89
Table 4 Composition the used TMM, * values when using a Glycol/Water mixture
3.1.2 Blood mimicking fluid
If water were to be sent through the vessel, only a few echoes would be generated by the
echography machine. To amplify the backscattered power little particles, known as scatterers, have
to be added to the fluid.
Potato starch was used during the initial testing. These ovoid particles have a diameter ranging from
30 to 100 µm, and a density of 1.55 g/mL. Using a magnetic stirring bar, starch can easily be
homogeneously dispersed in water. However, due its relatively high density compared to water, the
particles tend to sink after a few minutes. This is no problem when using a closed circuit with high
flow rates gushing through the vessels, because little turbulence will propel the particles back into
suspension. However, when a syringe pump was used to administer very low flow rates (< 10 mL/h),
the particles quickly sank to the bottom of the syringe. This caused fade out of the echoes after some
minutes of measuring, because the scatterers remained in the syringe. Furthermore, individual
potato starch particles are too big to be used in capillary vessels (<200µm), not mentioning the fact
that they tend to agglomerate and form bigger clumps.
For making the final measurements, orgasol nylon particles were used (Orgasol (R) Caresse). The
density of orgasol is 200-500 kg/m³, which is a lot less dense than water, hence the need for a
wetting procedure as described in [32].
3.2 Wire extraction phantom
Knowing how the BMF and TMM can be made, the design of the phantom itself should be addressed.
The first technique that will be explained relies on the extraction of wires which are embedded inside
the TMM. After letting it solidify, the wires would be removed, thus leaving vessel-like cavities inside
the phantom. The advantages of this technique are the lower cost of solid wires compared to hollow
tubes, and the broad spectrum of available dimensions. But this design, simple as it may look, proved
to be daunting to produce.
Tube connector
Tube
connector
Figure 17 (A) Tensioned wires and connectors fitted through the wall, ready to pour the TMM (B) box from A filled with
G/A, wires extracted and ready to perfuse
22
A simple plastic box (Figure 17, 96x96x64mm) was used as a TMM container. Small holes were made
in the opposing walls using a needle (diameter of 0.81mm and 0.56mm), making it possible to embed
the wires within the sample volume. To provide a flow inlet, a tube connector was inserted in the
hole. Then the wires were tensioned, and the edges between the connectors and the box were glued
to assure no TMM could leak. However, the holes of the connectors would still allow the TMM to
leak out of the box. An attempt was made to seal the aperture with glue or silicone, sadly resulting in
clotted vessels due to remaining residue when pulling the wires out. The most efficient solution was
forcing a toothpick through the connector, making care to not break the tensioned wires.
The wires were removed while observing with the GE vivid 7 ultrasound. It was clear that the cavities
of the 52µm and 180µm collapsed immediately after the extraction, in PVA as well as in G/A. The two
other vessels measuring 300µm and 700µm remained seemingly intact. However, after one minute
of perfusion, the 300µm also closed. To find the cause, the phantom was cut open along the
direction of the closed vessel. No visible problems could be seen. Although, when manually purging
one of the connectors, a small amount of TMM came out of the cavity which could have led to the
blockage. During an attempt to expulse the clot through the vessel with higher pressure (by setting
the syringe pump to its maximal level = 1000 mmHg), the TMM ruptured (Figure 18).
Figure 18 Rupture due to elevated pressure, initiated at the tube connector
3.3 Hollow tube phantom
The use of hollow tubes eliminates problems such as wall roughness, TMM-tear and vessel collapse.
But, they can’t be found in a variety of dimensions as solid wire. Furthermore, they come at a high
cost. Luckily, 20 samples of polyimide tubes were donated by Microlumen®.
The geometry of phantom should meet the challenges faced when observing complex and small
vascular networks. In contrast to carotid imaging, there is no clear distinction between the strong
signal arising from the artery, and the negligible signals arising from smaller arterioles and capillaries.
Hence, following influences should be examined at different flow rates:
-
Depth
Angle
Tube diameter
23
-
Crossing patterns
A box (similar to 3.1.3) was used to make a phantom with a depth of 4 cm. The tubes were put
through holes (made with a syringe). After tensioning the tubes, the aperture between the vessel
and the box were filled using glue. The ends of the tubes were inserted in a larger silicon vessel (ID =
200µm), because no needle could be found thin enough to fit inside the tubes lumen. The silicon
tube can be infused using an ultra-fine stent needle.
Figure 19 Final model before pouring the PVA
Figure 19 shows the final phantom before the TMM (PVA , 2 ftc) was inserted. It basically consists of
5 parallel tubes (100,155,165,178,200µm) inserted under an angle of 25 degrees. It should provide
enough space to measure at depths ranging from 5-30 mm.
The perfusion of this system proved to be straightforward, as will be discussed in chapter 4.
3.4 Discussion
The problems of making flow phantoms for microcirculation lie in the attempt to reconcile both the
simplicity of the design (“low-cost”) and the high-precision needed at such low dimensions.
The wire extraction method proves to be useful when mimicking structures at the level of arterioles
or bigger (>300µm). When going smaller, the structural properties such as rigidity and strength of the
TMM (both PVA and Gelatine/Agar) are insufficient to generate a reliable set-up resulting in blocked
or collapsed vessels. The biggest problem proved to be the opening of the tube connector. To
prevent the TMM from leaking, it has to be shut when pouring the liquid PVA or G/A inside the
container. But once the tissue mimic has solidified, it has to be reopened, to make perfusion
possible. It became clear that the substance used to seal the aperture (glue, wax etc.) is prone to
granulate and cause obstructions when perfusion is started.
24
Hollow tubes can be embedded in a TMM and purged to simulate flow through smaller vessels.
However capillary tubing comes at a high cost and the range of dimensions is limited. Depending on
the material that is used, high reflections arising from the vessel wall could be detrimental for the
measurements. This could clearly be observed when embedding a close-packed 3*3 matrix of 200µm
silicon tubes. The reflections of the superficial tubes could be noticed through the image down to the
lowest tubes. Nonetheless, flow through vessels down to 100µm can be replicated, making a
parametrical study possible.
Reflections
Figure 20 B-mode image of 1 longitudinal and 6 cross sections of 200µm silicon tubes. Distinct lumens can hardly be
observed, and high reflections could impede precise measurements
25
Chapter 4. Experiments
In this chapter, the experiments will be discussed. As with the build of the phantom, many
complications occurred in the early stages of testing, which created the necessity of doing some
extra tests. The first section describes the response of the set-up to changing the infusion speed.
Next, the accuracy of the flow speed measurements, at different depths and through various tubes,
will be tested. By recording the CFI and PW data, a comparison between the two can be made.
4.1 Pressure test procedure
During the early stages of testing, it became clear that any manipulation (e.g. increasing the speed)
on the flow had a long term effect on the measurements. This paragraph describes a pressure test
which has been done to quantify the time needed for the flow to stabilize (after being altered).
Firstly, the basics of a syringe pump will be discussed.
A syringe pump is the perfect machine to deliver a fluid at a precise flow rate. As it was produced to
administer medicinal drugs, it works with high precision. The driving element is a stepper motor,
which turns a ball screw mechanism to achieve “pseudo-linear” movement of the plunger. The word
“stepper” refers to the way this electrical motor works. In contrast to other electrical motors, where
a slightly varying torque is applied to the rotor at almost any instance, the rotor of a stepper motor
jumps from one angle to another. Depending on its design and the speed at which it turns, this
movement can look quite smooth. But, some oscillation of the fluid’s speed and pressure can’t be
excluded. Hence, a test was done to investigate if a pulsatile factor could be observed. The second
reason to conduct the test is to quantify the time needed for the pressure to stabilize.
The syringe pump (Alaris GH carefusion, calibrated) was connected to a 200µm hollow tube,
embedded in the final phantom. A Y-shaped connector was inserted between the phantom inlet and
the syringe, to connect a pressure measurement tool. Figure 21 show the results of the test, but it
should be mentioned that connecting the system generated in initial pressure, leading to an offset
from the origin at the start of the test.
Variation of pressure in time
Pressure (mmHg)
60
40
20
0
0
200
400
600
800
1000
1200
1400
Time (s)
Figure 21 Variation of the pressure in time after setting a certain flow rate to the system. Convergence takes more than
20 minutes
26
Pressure (mmHg)
Stabilized pressure in time
40
20
0
0
20
40
60
80
100
Time (s)
Figure 22 Pressure test in stabilized conditions. No pulsatility can be detected.
The time needed for the pressure to stabilize is substantial ( > 1200s) because of the relatively large
dimension (radius of 10mm, longer than 10 cm) of the branch connecting the pressure measurement
tool to the circuit, compared with the flow phantom vessel, and the low flow rate (4mL/h). An
approximation of the time needed for the pressure to stabilize is done using volume proportions of
the capillary vessel and the “measurement circuit”.
(
)
(
)
It can be assumed that constant pressure results in a constant flow speed (when keeping the flow
resistance constant). This simplified calculation shows that the pressure should converge to the final
value when waiting for at least 30s. Figure 22 shows the absence of any variation in pressure once it
has reached its final value.
4.2 PW flow speed measurements
The goal of the experiment described in this section is to see if the frequency, flow speed and vessel
diameter have an influence on the accuracy of pulsed wave speed estimation. The first tests were
performed with the Visualsonics high-frequency echography machine. Afterwards, a similar
procedure has been performed on a “traditional” machine.
4.2.1
Visual Sonics
4.2.1.1 Testing procedure
Figure 19 shows the phantom that has been used for the measurements. It basically consists of five
parallel tubes (internal diameter of 100,155,165,178 and 200µm) embedded under a moderate angle
(around 25°) inside PVA (2ftc). The syringe pump was used to infuse the BMF at flow rates ranging
27
from 1 to 10 mL/h. The scattering function was provided by Orgasol particles, which have been
filtered extensively with a 40µm filter to avoid clotting. The echography machine used is a
Visualsonics Vevo 2100, equipped with the MS250 (13-24 MHz) and MS550S (32-56 MHz) linear array
transducers. Sadly, due to technical limitations of the transducer, flow measurements with beam
angle steering could only be made at one frequency for each transducer, being 16 and 32 MHz.
Before testing, all the air bubbles remaining in the system had to be removed. To do so, the phantom
(the specific tube) was perfused at high flow rates (>15 mL/h), for at least 5 minutes. Afterwards, the
machine was paused until the flow ceased. At this stage, the syringe pump could be set at any
desired flow rate to start testing.
In section 4.1 we addressed the danger of measuring too fast after changing the flow variables. To
prevent collecting data from a set-up in a transient state, the recordings were done after waiting at
least two minutes after every manipulation. And, to avoid any signal arising from movements when
measuring, the probe was fixed inside a mount.
The flow speeds were calculated using following formula:
(with Q being the flow rate, A the surface of the tubes lumen, and
the flow speed)
4.2.1.2 Data processing
The Visualsonics machine has its own proprietary post-processing software package (Figure 23). This
can be used for measuring distances or angles, or changing the contrast and brightness of the image.
But, it can also be used to calculate the mean speed based on the sonogram, which proved to be very
useful for this test. However, it should be stated that the identification of this mean value is based on
a sensitivity scale1, which has to be determined intuitively.
Sample
volume
Ultrasound
beam with
angle
adjustment
Automatically
tracked mean
Figure 23 Post-processing of a sonogram with the Vevo software package. The mean value can be traced automatically.
1
By lowering the sensitivity scale, the software will look for clear echoes only. If the sonogram is discontinuous or has
ambiguous regions, they will be neglected resulting in a local zero speed estimation. A higher sensitivity value will also take
the faintest echoes into account, and cancel the obscure regions by interpolating the strong echoes confining this zone.
28
4.2.1.3 Test results
Figure 24 shows the results of PW speed estimations on a 178µm ID tube at 32 MHz (blue) and 16
MHz (red), and the theoretical value (black line). The speeds were calculated using the Vevo software
mentioned above, with an optimal sensitivity setting chosen for each recording. An overestimation of
the speeds can be observed at 32 MHz as well as at 16 MHz. Both the trend lines have a R² value
close to one. This means that the difference between the theoretical value and the measured values
won’t be a consequence of design flaws leading to inconsistent results. The overestimation could be
caused by a poorly chosen angle correction, because -as mentioned before- the measured speed
has to be corrected to the real speed
using the angle between the US beam and
the flow direction. This calculation is based on following equation:
( )
The angle correction can be changed in steps of 5 degrees which implies that if the angle would be
set to 65°, instead of the real 67°, an error of 10% will be generated. Note: the difference between
using 60° and 65° for results in an aberration of 18,3 %, whereas the difference between 40° and
45° is only 8.3%. This is why the angle should always be minimized, preferably below 60° when
doing measurements in-vivo and in-vitro [34].
Measured vs true flow speed, 178 µm
Measured speed (cm/s)
10
R² = 0,9926
8
R² = 0,9992
6
4
2
0
0
2
4
6
infusion speed (cm/s)
Linear (16 MHz)
Linear (std)
8
10
Linear (32 MHz)
Figure 24 Measured speeds vs. speed of infusion of a 178µm vessel
Figure 25 shows the trend lines of the proportional error calculated by:
A convergence towards the theoretical value can be observed at higher speeds for both frequencies.
But, because many variables can cause a certain offset in the results (e.g. angle correction), it can be
stated that the consistency between the measurements is crucial, rather than the absolute value of
the estimation. At 16 MHz, the results are more accurate than at 32 MHz, and this also applies to the
155µm tube.
29
Proportional error on speed measurement
error (%)
120
90
60
30
0
0
2
4
6
8
infusion speed (cm/s)
Log. (16 MHz)
Expon. (32 MHz)
Figure 25 Proportional error on speed measurements, both trend lines convergence to zero.
Figure 26 shows almost perfect estimations at 32MHz, whereas at 16 MHz there is a small offset of
0.4481 cm/s (
). Again, a convergence towards zero can be observed
with higher infusion rates (Figure 27).
Measured vs true flow speed, 155 µm
Measured speed (cm/s)
10
R² = 0,9947
8
R² = 0,9894
6
4
2
0
0
2
4
6
8
10
infusion speed (cm/s)
Linear (std)
Linear (16 MHz)
Linear (32 MHz)
Figure 26 Speed measurements within a 155 µm tube
Proportional error on speed measurement
(155µm)
error (%)
100
0
0
2
4
6
8
10
-100
-200
infusion speed (cm/s)
16 MHz
32 MHz
Figure 27 Proportional error on the speed measurement when infusing a 155 µm tube. At both frequencies, a
convergence towards zero can be observed for higher flow speeds.
30
Figure 28 shows the results acquired from the 100 µm tube. Although the results look decent, the
results can be denoted as untrustworthy. The first reason is the very frequent obstructions faced
while doing the test. Narrow as the lumen is, only a few large BMF particles have to coagulate to
create a clump big enough to close the vessel. Second, the system responded slowly to changing the
settings of the pump. To visualize these problems, the logarithmic trend line has been plotted,
showing a good fit between the continuous measurements done from 4-10 mL/h. Furthermore, the
resistance of the vessel was so high that even when observing stable laminar flow, the syringe pump
went into alarm (> 1000 mmHg at highest setting). This could possibly be solved by filtering the BMF
in the 8-20 µm range, and by keeping the tube as short as possible. Nonetheless, the sonograms of
these measurements showed no hiatuses or any apparent lack of backscattered information
necessary to track the mean value.
Measured speed (cm/s)
Measured vs true flow speed, 100 µm
12
R² = 0,9463
10
8
6
4
2
0
-2 0
2
4
6
8
10
12
infusion speed (cm/s)
32 MHz
Linear (32 MHz)
Linear (std)
Figure 28 Flow speed measured through a 100 µm ID tube, a large spread in results can be observed due to clots shutting
off the vessel
To be conclusive on the influence of diminishing diameter sizes, even thinner tubes should be
acquired to perform the same experiments on. However the results are showing no complications on
the level of speed measurements in a simple setup.
4.2.2 GE
The influence of varying depth could not be measured with the Visualsonics because the transducers
are optimised for imaging superficial (5-15 mm) structures in rodents. The imaging depth is also
limited due to the high frequency and consequently high attenuation. Hence, a similar experiment as
mentioned above has been done using the GE Vivid 7, equipped with a linear array transducer (5-9
MHz), but with a certain variation of depth.
4.2.2.1 Data processing
In contrast to the Visualsonics, there was no software package available for post-processing data
acquired with the Vivid 7. A Matlab tool (Asklepios) that had been written in the context ultrasound
research was used to estimate the velocity on a sonogram. The working of this program is based on
estimating where the white zone of the sonogram (i.e. “stronger” echoes) stops (Figure 29). The
distance between this endpoint and the baseline can be multiplied with the scale (contained in the
dicom header tags), which results in a single speed estimation. Repeating this procedure for every
point of the contour will provide a good approximation of the maximum speed. However, the true
31
averaged speed lies within the centre of the white band, and can’t be approximated by using the
peak values. This is why, for this experiment, the lower contour was added to the upper contour, and
divided by two.
Upper contour
(white dots)
Averaged speed
Lower contour
Figure 29 Sonogram being processed using the Asklepios Matlab tool
Figure 30 shows an underestimation of flow speeds when observing deeper structures. This can be
expected owing to the fact that the attenuation will increase exponentially when going deeper inside
the TMM. As the intensity of the backscattered signal drops, the upper and lower contours are likely
to drop because the echoes won’t be detected unambiguously. Changing the sensitivity could
address this problem because weaker signals won’t be neglected. But, this will generate many
erroneous interpretations, making the measurement unreliable.
Measured vs. true flow speed at 2 depths
10
Measured speed (cm/s)
R² = 0,9887
8
6
R² = 0,9586
4
2
0
0
2
4
6
8
10
infusion speed (cm/s)
Figure 30 Measured vs. True flow speed at 1.5cm (red) and 3cm (blue)
32
4.3 Colour flow speed measurements
4.3.1 Visualsonics
The extraction of flow speeds starting from colour Doppler images is quite cumbersome. There was to my knowledge- no software which could directly relate images to flow speeds. The procedure for
extracting the data is explained below.
4.3.1.1 Data processing
Studies saved on the Visualsonics are stored in a proprietary file type. These can be read by the Vevo
software package, and they can be exported to .dicom files. This is a standard file type developed
specifically for “digital imaging and communications in medicine” (DICOM). Once the images are
converted to .dicom, they can be read into Matlab.
The first part of a dicom file consists of several attributes which refer to the imaging settings
(frequency, flow scale etc.) and some general information (time and date, patient’s name etc.). The
second part contains the pixel data, stored in an m*n*3 matrix, respectively being the number of
rows and columns, and the red, green and blue values of each pixel. When looking closely at a colour
flow image, it is clear that the intensity of the flow (and colour) is directly related to the green value
(for arterial = red and venous = blue flow). A piece of Matlab code was written which relates the
green values of every pixel to the colour scale. The GE machine has a linear scale which makes a
linear interpolation quite precise. The Visualsonics has a discontinuous parabolic scale, thus making
linear interpolation imprecise. This is why the measurements below have been calculated manually
by linking the green value to the parabolic scale designated by the machine. The true speed can be
calculated using following formula:
( )
with
being the average green value of a cross section divided by the maximum green value (=
255), and s being the flow speed scale factor.
Theoretical speed vs Color doppler estimation
Measured speed (cm/s)
12
R² = 0,9935
10
8
6
4
2
0
0
2
32MHz
4
6
8
10
12
Theoretical speed (cm/s)
Linear (std)
Linear (32MHz)
Figure 31 Colour Doppler speed estimation in a 178 µm tube
33
As can be seen in Figure 31, the results are surprisingly accurate even at low speeds. This
corresponds to the results obtained in a previous study done by Carolus J.P.M. Teirlinck and Peter R.
Hoskins d [35], who noticed an overestimation of only a factor 1.0 to 1.25 at 2.5 MHz. However, a
more recent study showed there was an overestimation with a factor 1.5-3 [3]. These results can’t be
confirmed by the measurements done in this thesis. Lower flow speeds were not measured because
the lower boundaries of the flow speed scale were already reached, resulting in almost no colour
data (cf. [36]).
4.4 Flow profile monitoring
The collected CFI data could be used to estimate the flow profile within the tube. If this proves to be
accurate, problematic flow patterns within the smallest arterioles could be analysed. Ideally, the flow
pattern should be parabolic. However, the information the transducer acquires is based on the
momentary distribution of scatterers (in vivo = red blood cells) which could deviate from the
parabolic pattern. To draw the flow profile of one single cross-sectional line, the average of 15
frames has been calculated using the same methods as described in 4.3. The results are shown in
Figure 32. The profiles based on the 16 MHz images approach the parabolic shape although a minor
plateau can be seen, which arises from an artefact in the image. The 32 MHz profiles display a
different pattern differing a lot from the parabolic form. The explanation for this phenomenon can be
seen in Figure 33. At 32 MHz, the reflections are high, leading to a falsely observed flow outside of
the vessel. This problem doesn’t exist at 16 MHz, or for the 100 µm tube.
Having accurate flow estimates within a small sample volume (cf. 4.2.1.3) doesn’t necessarily prove
that erroneous flow patterns can be detected. But, combined with the knowledge that the measured
distribution of speeds across the tube is corresponds to the theoretical profile; more information
could be gathered about the flow pattern of the blood. This could be useful for e.g. making surface
plots of the blood flow through of narrow site.
34
32 MHz
100 µm
16 MHz
32 MHz
reflection
155µm
178 µm
Figure 32 Comparison between the flow profiles (speed vs. pixels) for different frequencies and different tube diameters
Reflection
Figure 33 The colour Doppler image taken at (L) 16 MHz confines the flow quite precisely within the tube. At (R) 32 MHz a
reflection of the tube can be observed.
35
4.5 Discussion
Because this discussion embodies several aspects of the measurements, they will be addressed
separately.
Set-up
The set-up showed to be reliable up to 155µm. The flow was easily controllable without generating
resistance which exceeded the syringe pumps maximal value. The scatterers remained in suspension
throughout the entire measurement procedure and provided an echo strong enough to be detected.
It was feared that the polyimide tubing could cause strong reflections, but they demonstrated a
decent echogenicity. When testing the 100µm tube, several problems emerged. The filtering of the
orgasol with a 70µm filter was insufficient to prevent blockage of the vessel. This caused periodical
pressure build-ups, followed by high flow speeds until the system stabilized. Furthermore, the
resistance generated by this short tube was high enough to push the syringe pump into alarm. This
made continuous measurements arduous.
Measurements
Both the measurements at 32 MHz and 16 MHz showed to be accurate for estimating “high” and low
(1 cm/s) flow speeds through vessels down to 155µm. It is expected that the same trend will be
observed when downscaling to 100µm, but the phantom reached its boundaries making it impossible
to produce conclusive data at this level. It should be noted that the sonogram gets ambiguous when
observing speeds lower than 3 cm/s due to two reasons. The first being the weak signal received
from such slow fluid movement consequently producing a vague sonogram. Second, it is all but
optimal to estimate values which are closely packed around the baseline, even when the scale is
minimized.
Optimal set-up
It will be extremely hard to incorporate the additional difficulties such as smaller vessels (<100µm),
muscle movement en complex patterns into a model as produced. Nonetheless, these challenges will
be decisive to the question If perfusion at microcircular scale is feasible or not. A better solution
would be the perfusion of an embalmed ex-vivo structure. Such a set-up comprises the fractional
nature of a complex network of crossing vessels. As stated in the literature review, it can be difficult
the asses which volume is passing through each capillary, but this can be solved using a microscope
to track the moving particles while performing US measurements. Hence, a comparison can be made
between the ultrasound imaging and the ground truth.
36
Chapter 5. Extra: CFD
5.1 Introduction
Preterm born babies are prone to develop cerebral strokes within the first weeks after birth. There
are many techniques to detect abnormalities, which could be an indication of cerebral lesions. But,
they all impose a high burden on the patient. Most of these imaging techniques rely on the use of
contrast agents or radioactivity, and often the patient has to be transported to the machine and
sedated to acquire decent images. It won’t be surprising to state that the toll of similar examinations
will be too high for preterm born babies. If high-frequency ultrasound proves to estimate brain
perfusion accurately, it could be used as a bed-side tool to monitor the brain circulation of preterm
born babies on a regular basis. Hence, any indication of an upcoming problem could be detected at
an early stage which would minimize the possibly detrimental outcome of e.g. a stroke.
The experiments which have been conducted on the parametrical model (cf. chapter 4) are not
sufficient to state that high-frequency ultrasound is ready for clinical use. Before doing so, exhaustive
testing should be done on an in-vivo mimicking model, which could replicate the flow with high
precision. The difference between this set-up and the parametrical model would be the complexity
of the geometry and flow pattern. If possible, low frequency (5-36 Hz) clutter noise should be
introduced to replicate muscle vibrations throughout the tissue [37]. The geometry of choice is the
Circle of Willis (CoW) because it is of utmost importance to the perfusion of the cerebral vascular
system.
In this chapter, the importance of the CoW will be briefly explained by showing the basic anatomy.
Next, the basics of the model acquisition and meshing procedure will be covered, as well as its
possible uses to further research.
5.1.1 Anatomical information
The human brain is the most crucial part of the body because it is
the centralized computer powering every mechanical, physiological
and chemical process necessary to live. It is fuelled by a complex
circulatory system with a built-in safety feature, known as the Circle
of Willis. If one of the major arteries supplying the brain with
oxygenated blood becomes narrow or gets blocked, the other
arteries can still provide enough blood to avoid severe health issues
(e.g. ischemia) by using alternative pathways. This set of linkages
between vessels (anastomosis) is known as the Circle of Willis. A
peculiar aspect of the CoW is its design, which strongly varies from
person to person. Only 20-25% of the population possesses a fully
develop system.[38]
Figure 34 Anatomical picture of the circle of Willis [1]
37
5.1.2 From images to model
The production of a 3D model can be a cumbersome procedure involving many different challenges.
In this section a short overview of the basic steps will be given. Exhaustive guides can be found
online.
An MRI-TOF (=time of flight) dataset was used as the basis of the model. Such an examination
consists of making many thin slices, each representing a cross-sectional image of the head. Based on
several phenomena which will not be addressed in the scope of this thesis, the internal structures
can be clearly distinguished in each image. When loading this data into the Materialise Mimics
software, the Circle of Willis can be delimited in each slice. Doing this for many images, an
approximation of the regions of interest its contour can be calculated. This results in a rough model
with many artefacts and defects. Further cleaning can be done by either erasing some pixels on each
slice, or by using smoothing and triangle reduction algorithms implemented in the Magics ® postprocessing software. The first is preferred because solving the problem at its roots proves to be timesaving and better results will be achieved afterwards. For finalizing, the post-processing software can
be used. To prevent the model from being too complicated as can be seen in Figure 35, it is beneficial
for the CFD results to eliminate the smallest vessels and to concentrate on the larger arteries.
The next step involves the generation of a mesh using ICEM. Depending on the application, the mesh
can be made out of 1D, 2D (thinshell) or 3D elements. In the context of fluid simulations, a 3D mesh
will be required.
Figure 35 (L) before trimming the excess branches (R) after trimming and smoothing
To start the CFD calculations, there should be enough data on the flow speeds through the afferent
vessels. The outlet flow can be set to “velocity outlet” and based on the proportional surface areas;
approximations of the volume distribution through each separate efferent vessel can be made. Test
solutions, based on parameters found in the literature [39] showed good convergence behaviour,
with very low residuals. Figure 36 shows a solution on which the vectors are depicted, and their
magnitude is related to a colourscale. By making such a visual representation of flow data, problems
can easily be seen and communicated to non-technical staff.
38
Figure 36 (up) CFD simulation of flow through the circle of Willis (down) zoomed image showing the individual vectors
By combining the design possibilities of Magics with the CFD power of Fluent, many interesting
pathologies could be created and simulated, on general models and on patient specific data.
5.1.3 Real model
The .STL file shown in Figure 35 has been 3D printed; the result can be seen in Figure 12.
Unfortunately, the geometry is too small and too complex for standard stereolithographic printing
methods, leading to several failed attempts to remove the wax core.
The benefit of having a real model is the ability to validate problematic and complex flow patterns
calculated using CFD software, and vice versa to check if ultrasound measurements can accurately
visualize them.
39
List of references
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
Circle of Willis. [cited 2016 14-3].
K J M Surry, H.J.B.A., A Fenster,T M Peters, Poly(vinyl alcohol) cryogel phantoms for use in
ultrasound and MR imaging. Phys. Med. Biol, 2004. 49: p. 5529-5546.
Fleur A. Camfferman1, G.M.E.-G., Jhuresy E. La Roche2, Nico de Jong3,4, Willem van ’t
Leven3, Hendrik J.Vos3,4, Martin D.Verweij 4, Kazem Nasserinejad 5, Filip Cools1, Paul
Govaert 1,2 and Jeroen Dudink 2,6*, Calibrating Doppler imaging of preterm intracerebral
circulation using a microvessel flow phantom. Frontiers in human neuroscience, 2015. 8.
Embley, B. Sea Floor Mapping. 2001.
J.-L. Gennisson, T.D., M. Fink, M. Tanter, Ultrasound elastography: Principles and techniques.
Diagnostic and Interventional Imaging, 2013. 94: p. 487-495.
Giancoli, Mechanica en Thermodynamica. Vierde editie ed.
CM, O., Principles of echocardiographic image acquisition and Doppler analysis. Textbook of
Clinical Ecocardiography. 2000.
Venables, H. Types of Ultrasound - A-mode. Ultrasound: Physics and Basic Equipment Settings
2012.
Spectral doppler. [cited 2016 2/5].
Swillens, A., A multiphysics model for improving the ultrasonic assessment of
large arteries. 2010, Universiteit Gent.
11.
MARGOT M. HORDER, S.B.B., † GILBERT J. VELLA,‡ and a.A.K.W.W. MARSHALL J. EDWARDS,
IN VIVO HEATING OF THE GUINEA-PIG FETAL BRAIN BY PULSED ULTRASOUND AND
ESTIMATES OF THERMAL INDEX. Ultrasound in Med. & Biol, 1998. 24: p. 1467–1474.
12.
Diane Dalecki , C.C., Hemorrhage in murine fetuses exposed to
pulsed ultrasound. Ultrasound in Med. & Biol, 1999. 25: p. 1139–1144.
13.
Æ, S.Y.Æ.T.I.Æ.A.I. and R.O.Æ.S.T.Æ.Y. Matsumoto, High intensity focused ultrasound
lithotripsy with cavitating
microbubbles. Med Biol Eng Comput, 2009. 47: p. 851–860.
14.
T. Bodnár a, b., ⇑, A.S.c. , and M.P.d. , On the shear-thinning and viscoelastic effects of blood
flow
under various flow rates. Applied Mathematics and Computation, 2011. 217: p. 5055-5067.
15.
Cell-free marginal layer model. [cited 2016 6-4].
16.
a, C.W., P.S.a. , and S.S.b. , Aggregation of Red Blood Cells: From Rouleaux to Clot Formation.
Comptes Rendues en Physique, 2013. 14.
17.
Rickey, R.D.F.A., A velocity evaluation phantom for colour flow and pulsed Doppler
instruments. Ultrasound Med Biol, 1992. 18: p. 479–494.
18.
Cezo, J.D.R., Mark E; Kramer, Eric A; Schoen, Jonathan A; Ferguson, Virginia L; Taylor,
Kenneth D, Tissue storage ex vivo significantly increases vascular fusion. Surg Endosc, 2015.
25.
19.
Hoskins, P.R., SIMULATION AND VALIDATION OF ARTERIAL ULTRASOUND. Ultrasound in Med.
& Biol, 2008. 34: p. pp. 693–717.
20.
Willaert, W., Lifelike Vascular Reperfusion of a Thiel-Embalmed Pig Model and Evaluation as a
Surgical Training Tool. 2016.
21.
Cardon, L., Product development additive manufacturing. 2015.
22.
Rosei, A.K.F., Nanoelectronics and Photonics From Atoms to Materials, Devices, and
Architectures. 2008: Springer-Verlag New York.
23.
Andreas Ostendorf, B.N.C. Two photon polymerization, a new approach to micromachining.
[cited 2016 27 november].
40
24.
25.
26.
27.
28.
29.
Veltmann, C., et al., ON THE DESIGN OF A CAPILLARY FLOW PHANTOM FOR THE. Ultrasound
in Med. & Biol, 2002: p. 625-634.
J. Cooper McDonald, D.C.D., Janelle R. Anderson, Daniel T.Chiu, Hongkai Wu,Olivier J.A.
Schueller, George M. Whitesides, Fabrication of microfluidic systems in
poly(dimethylsiloxane). Electrophoresis 2000. 21: p. 27-40.
Spence, D.M., Automation and Microfluidic Assays: In Vitro Models of the Mammalian
Microcirculation. Journal of the Association for Laboratory Automation. , 2005. 10.
Hendrik H. G. Hansen*, R.G.P.L., and Chris L. de Korte, Noninvasive Carotid Strain Imaging
Using Angular Compounding at Large Beam Steered Angles: Validation in Vessel Phantoms.
TRANSACTIONS ON MEDICAL IMAGING, 2009. 28.
Rickey DW, P.P., Christopher D, Fenster A, A wall-less vessel phantom for Doppler ultrasound
studies. Ultrasound Med Biol, 1995. 21: p. 1163–1176.
Jia Wei Li, M., Manoj K. Karmakar, MD, Xiang Li, PhD, Wing Hong Kwok, FANZCA, Warwick
Dean Ngan Kee, MD, Gelatin-Agar Lumbosacral Spine Phantom
A Simple Model for Learning the Basic Skills Required to Perform
Real-time Sonographically Guided Central Neuraxial Blocks. J Ultrasound Med, 2011. 30: p. 263-272.
30.
V. Pazosa, R.M., J.C. Tardif, Polyvinyl alcohol cryogel: Optimizing the parameters of cryogenic
treatment using hyperelastic models. Journal of the mechanical behavior of biomedical
materials, 2009. 2: p. 542-549.
31.
Law YF, J.K., Routh HF, Cobbold RSC, On the design and evaluation of a steady flow model for
Doppler ultrasound studies. Ultrasound in Med. & Biol, 1989. 15: p. 505-516.
32.
CP, O., Towards an ideal blood analogue for Doppler ultrasound phantoms. Phys. Med. Biol,
1991. 36: p. 1433-1442.
33.
https://en.wikipedia.org/wiki/Starch. Starch. [cited 2016 23 April].
34.
K. Logason 1, T.B.r., M.-L. Jonsson1, A. Bostro¨m1, H. G. Ha° rdemark2 and S. Karacagil1, The
Importance of Doppler Angle of Insonation on Differentiation Between 50–69% and 70–99%
Carotid Artery Stenosis. Eur J Vasc Endovasc Surg, 2001. 21: p. 311–313.
35.
Carolus J.P.M. Teirlinck, R.A.B.a., Christian Kollmann b, Jaap Lubbers ‘, and P.F.e. Peter R.
Hoskins d, Knud-Erik Fredfeldt f, Ulrich G. Schaarschmidt g, Development of an example flow
test object and comparison of five of
these test objects, constructed in various laboratories Ultrasonics, 1998. 36: p. 653-660.
36.
XIAOCHEN XU, L.S., JONATHAN M. CANNATA, JESSE T. YEN, and K. KIRK SHUNG, HIGHFREQUENCY ULTRASOUND DOPPLER SYSTEM FOR
BIOMEDICAL APPLICATIONS WITH A 30-MHZ LINEAR ARRAY. Ultrasound in Med. & Biol, 2008. 34: p.
638–646.
37.
Andreas Heimdal, H.T., Ultrasound doppler measurements of low velocity blood flow:
limitations due to clutter signals from vibrating muscles ieee transactions on ultrasonics,
ferroelectrics, and frequency control, 1997. 44: p. 873-881.
38.
Circle of Willis. [cited 2016 25-5]; http://radiopaedia.org/articles/circle-of-willis].
39.
Morris, P.F.P.D.P.M.S.S.N.H.L., An In Vitro Assessment of the
Cerebral Hemodynamics
Through Three Patient Specific
Circle of Willis Geometries. Journal of Biomechanical Engineering, 2014. 136.
41