information to userscollectionscanada.gc.ca/obj/s4/f2/dsk1/tape11/pqdd_0005/mq44044.pdfwillingnes to...
TRANSCRIPT
INFORMATION TO USERS
This manuscript has been mpmôuœâ from the rnicrdilm master. UMI films the
text directly from lhe original or copy wbmitted. Thus, some thesis and disserkation wpies are in typewritsr fsce, whiie othem may be from any type of
cornputer printer.
The qualify of thir nprodudkn k d.p.nd.irt upon th. qurlity of the copy
submitteâ. Brdcen or MMnct prinf co(orsd or poor qwlity i I I u ~ s and photographs, print Mcredlhriough, substandard marghs, and irnpmper alignmerrt
can adversely affect mpdwîhn.
In the untikely event that üm author did not semd UMI a complete manuscript and
there are missing pages, thse wi# be llofed. Also, if unauttwked copyright
mateflaf had to be removeci, a note will indicate the dektim.
Oversize materials (eg., W. dMngs, chartr) are reproduced by setcüming
the original, beginning at the upper lefthand corner and continuing f m left to nght in equal sections with small overîaps.
Photographs induded in Ihs original manudpt haw bem reproduded
xemgraphically in this -y. Higher quality 6' x W bbdr and white photographie
prints are availabie for any photographs or illustrations -ring in this copy for
an additional charge. Contad UMt dirscHy to d e r .
Ml & H d l Infomatbn and Leaming 300 North Zeeb R W , Ann Arbor, MI 481-1346 USA
Modeling and Analysis of Ultrasound Backscattering by Red Blood Cell Aggregates with
a System-Based Approach
Beng-Ghee Teh Department of Electrical Engineering
McGill University, Montreal
March, 1998
A thesis submitted to the Faculty of Graduate Studies and Research in partial fuüiliment of the requirements for the degree of Master of Engineering.
O Beng-Ghee Teh, 1998
National Library 1*1 of Canada BiMiotheque nationale du Canada
Acquisitions and Acquisitions et Bibliographie Services services bibliographiques 395 Wellington Street 395, rue WetUingîon Ottawa ON K1A O N 4 OttawaON K1AON4 Canada cmada
The author has granted a non- exclusive licence allowing the National Library of Canada to reproduce, tom, distribute or seii copies of this thesis in microfom, paper or electronic formats.
The author retains ownership of the copyright in this thesis. Neither the thesis nor substantial extracts from it may be printed or othenilise reproduced without the author's permission.
L'auteur a accordé une licence non exclusive permettant à la Bibliothèque nationale du Canada de reproduire, prêter, distribuer ou vendre des copies de cette thèse sous la forme de microfiche/nlm, de reproduction sur papier ou sur format électronique.
L'auteur conserve la propriété du droit d'auteur qui protège cette thèse. Ni la thèse ni des extraits substantiels de celle-ci ne doivent être imprimés ou autrement reproduits sans son autorisation.
McGill University
Faculty of Graduate Studies and Research
Modeiing and Analysis of Ultrasound Backscattering by Red Blood Ceii
Aggngates witb a System-Based Approaeh
Thesis presented by:
Beng-Ghee Teh
and evaluated by:
Dr. Guy Cloutier supervisor
Dr. Howard C. Lee CO-supervisor
Dr. Richard S. C . Cobbold extemal examiner
Thesis accepted on April20,1998
The fear ofGod is the beginnzng of knowledge ...
Proverbs Ir 7
Abstract
The present study concemi the modeling and analysis of ultrasound bacb t te r ing
by red blood ce11 aggregates, which under pathological conditions play a signincant role in
the rheology of blood within human vessels. A theoretical model based on the convolution
between a tissue matrix and a point spread funciion, representing respectively the RBC
aggregates and the characteristics of the ultrasound system, was used to examine the
influence of the scatterer shape and size toward the backscattered power. Both scatterers
in the fom of clumps of red biood ceil aggregates and rouleaux were modeled. The
sirnulated results were used to postulate potential scatterer shape and sue teading to the
"black hole" phenomenon, a hypoechoic zone observed in ultrasonic B-mode images. It is
concluded that the ultrasound backscattered power does not always increase with the size
of the aggregates, especially when they are no longer Rayleigh scatterers. New potential
causes of the "black hole" phenomenon werr also proposed based on the model, in
addition to the proposais suggested by earlier researchers.
Résumé
Ce mémoire porte sur la modélisation et l'analyse du signal ultrasonore rétrodiffisé
par des agrégats de globules rouges qui, dans des conditions pathologiques, jouent un rôle
important dans la rhéologie du sang dans les vaisseaux sanguins. Un modèle théorique,
basé sur la convolution d'une matrice tissulaire servant à décrire les propriétés des
agrégats et d'une matrice représentant les propriétés du système d'ultrason, a été utilisé
pour examiner l'effet de la forme et de la taille des d i b u i s sur la puissance ultrasonore
rétrodi ffisée. Des difhseurs ayant la fonne d'agrégats sphériques de globules rouges et la
forme de rouleaux ont été modelis&. Les simulatioas avaient principaiement pour but
l'étude de l'effet de la fonne et de la taille des agrégats sur les caractéristiques d'un
phénomène acoustique nommé "trou noir". Ce phénomène observé sur des images
échographiques en mode B de vaisseaux sanguins simulés est caractérisé par des variations
de l'intensité de l'écho à l'intérieur de l'image et la présence d'une zone de faible intensité
au centre du tube. Les résultats montrent, lorsque la relation entre la taille des agrégats et
la longueur d'onde du faisceau d'ultrason ne suit plus la loi de Rayleigh, que la puissance
rétrodiffusée n'augmente pas nécessairement avec la taille de ces derniers. De nouvelles
hypothèses portant sur la genèse du ''trou noir" sont proposées aimi que la suggestion
faite par d'autres chercheurs à ce sujet au cours des dernières années.
Acknowledgment
1 would like to express my utmost gratitude to God for His grace and mercy, and
also for pmviding such a manrelous opp tun i ty to pursue post-graduate studies in this
laboratory. 1 would also like to give th& to Him for the constant source of strength,
encouragement and peace 1 fimi in His Word during this period of tirne- Without Him, this
degree would not have been possible.
1 would Iike to thank my thesis supervisor Prof. Guy Cloutier for his technical
leadership, patience, and encouragement throughout this entire period. His guidance
played a pivotal role in the sumssful completion of this thesis. 1 thadc Guy for providing
me with the opportunity to explore different technical areas, and for always k i n g there to
help and support. 1 would also like to thank my other thesis supewisor, Prof. Howard C.
Lee for his guidance, assistance and encouragement, especiall y du ring the beginning of my
M.Eng. program. 1 wish to thank both of my thesis supervisors for their careful revision of
this thesis. Without them, this thesis would not be in its current form today. I would also
li ke to thank Prof. Ooi for k i n g my CO-supervisor with Prof. Cloutier during the last stage
of my M.Eng. program.
1 am indebted to Prof. Michel Bertrand at the Institut de Génie Biomédical (IGB)
of the École Polytechnique de Montréal for always k i n g there to prwide technical
assistance and to share his insight. 1 also wish to thank Michel for providing the access to
the workstations as well as the cornputer programs developed at IGB.
1 would like to express my th& to Jocelyn Durand, Louis Allard, Luc
Dandeneau, and Richard Cimon at the Laboratory of Biomedical Engineering of the
Institut de recherches cliniques de Montréal (IRCM) for their patience and help in the
building of the flow loop mode1 as well as the interface circuit for data acquisition. Their
willingnes to go the extra mile to help was invaluable to the successful completion of the
experimental setup that was complementary to this thesis.
1 wish to take this opporhmity to expnss my gratitude to Prof. Lxniis-Gilles
Durand, Francine Durand, Dr. Zhenyu Guo, Marjan Yazdanpanah, Dr. Raja Bedi,
Danmin Chen, Xuan Zhang, Dany Leong Kon, Zhao Qin, Dr. Xiaoduan Weng, Isabelle
Fontaine, Dr. Herkole Sava, Damien Garcia, Frederic Sakr and Dr. Philippe Pibarot for
their many insightful discussioos, their friendship, and for making this laboratory at IRCM
a stimulating environment for reseaxch and advance 1eamiDg.
1 am also indebted to the Medical Research Council of Canada, the Whitaker
Foundation of USA, and the Heart and Stroke Foundation of Quebec for providing the
research funds throughout the entire course of my studies.
Finally, many t h h are due to my pastors, ail my Enends at the Christian
fellowships at McGill as well as the Taiwanese Presbyterian Church of Montreal, and my
family for their constant support, prayers and encouragement.
Contents
Abbreviated Lbt of Symbob œoœœœooooœœœœoœœwoœœoœœœoaœœoooooœoœoœœoœœœooooooaœœooœœoooooœooœoœoœœœœwœooœo-aœoœooo~ii
List of F i g ~ m ~ooo~œœaoœœœœœœ~œœoœooooœœœœœwœooœœœoooœoooœooooœoœœoœoœoœœoœœooooœoooœoœooœooœœoœœœœœœoœœooowœœooœoœœœœm~iii
Lbt OC Tables oœooœœœoœoœooœooœœœœœœ.œœœœœoœœoœoœeoeooœoœœooooooaœœœoaoooaooaoœoaoœœooooœooœooœœoœœooœœ.œoooœmœœœoœœmœœoœœœoo~fi
Chapter 1: Bickgt~uid and Obje~tive~ œ~aœo~œœ~oœœœoooœoœœoooœœœoœoooœoœoo~œœœoooœooooœoœœooooaœoooooo 1
1.1 The Rationale Behind the Study of RBC Aggregation ............................................ 1
1.2 Methods Roposed for the Study of RBC Aggregation .......................................... 5
1.2.1 Microscopie Obsenration ................................................................................ -5
......................................................... 1.2.2 Erythrocyte Sedimentation Rate (ESR) 6
.............................................................. 1.2.3 Direct Observation Under Shear Flow 6
................................................ 1.2.4 Laser Light Reflection Method ............... .... 7
................................................................. 1.2.5 Ultrasound Backscatteriog Methods 8
................................................................................................... 1.3 Proposed Study 10
1.3.1 A Brief Review on the Hypoechogenic Zone Discovery ................................. 10 1.3.2 Summary .................................................................................................. 14
1.4 Objectives ............................................................................................................ 15 Chapter 2: Tbeory, Implementation and Methodology ................................O............ 16
2.1 Scattering of Ultrasound by Blood ...................... .. ...................................... 1 6
..................................................... 2.1.1 RBCS as Scattering Targets ........... . 1 6
........................................................................................ 2.1.2 Rayleigh Scattering 17
2.1.3 The Modeling of Ultrasound Bacbcattering by RBCs .................................... 19 .................................... 2.1.3.1 The Particle Approach ...................... .. .... ......... 1 9
........................................ 2.1.3.2 The Continuum Approach
2.1.3.3 The Hybnd Appmach .................. .. ..................................................... -22
2.2 The Modeling of Ultrasound Backscattering by RBCs with the Bystem-
........................................................................ Based Approach ................ ....... 22
2.3 Implementation ..................................................................................................... 29
2.4 Methodology ........................................................................................................ 36
....................................................... 2.4.1 Verification of the Validity of the Mode1 3 8
2.4.1.1 Isotropy / Anisotropy Due to the Structure of the Scatterers .................... 38
....................... ............ 2.4.1 2 Power vs . Scatterer Volume Relationship ...... -42
2.4.2 The Effects of the Scatterer Structure and Size on the Signal Power .............. 43
2.4.3 The Potential Causes of the "Black Hole" Phenomenon ................................. 44
2.5 Summary ........................ ... ............................................................................... 44
Chapter 3: Resulîs oooœoooœooooœœoooœœooooœoooœooooooooœooœoooooooœoooooœooœoooooooooooœoooooœoooooooœooootooooooo.oooœo-œ45
3.1 Verification of the Validity of the System-Based Model ..................................... -45
.......................... 3.1.1 Isotropy / Anisotropy Due to the Structure of the Scatterers 45
3.1.2 Power vs . Scatterer Volume Relationship ....................................................... 47
3.1.3 Summary ....................................................................................... ............. 48
3.2 The Effects of the Scatterer Size and Structure on the Power ............................... 49
................................................. 3.3 The "Black Hole" Phenomenon: Potential Causes 52
3.4 Summary ........................................................................................................... 6 6
Chapter 4: D ~ ~ ~ ~ ~ b ~ ~ o ~ a o o o . o o ~ o o o o ~ o o œ o o o o o o o o o o o o œ o o œ o o o o o œ o o o o o o o o o o œ o o o o œ o o o o o o œ o œ o o o o œ œ o o o o o o o o o o 67
......................................................................................... 4.1 Analysis of the Results -67
................. 4.1.1 Power vs . Volume Relationship €rom a System-Based Perspective 67
4.1.1.1 Anisotropic cylindrical scatterers ......................... .... ........................... 67
..................... 4.1.1.2 Effects of the insonififation angle for anisotropic scatterers 70
.................................................................... 4.1.1.3 Isotropie spherical scatterers 71
4.1.1.4 Other considerations ................................................................................ 72
4.1.1.5 Simulation of a more realistic tissue image .......................... .... ..... 7 3
4.1.2 Most Probable Causes of the "Black Hole" Phenornenon ............................ ...75 4.2 The Strengths and Limitations of the Mode1 ......................................................... 79
4.3 Summary .......................... .. ........................................................................... 8 0
Chnpter 5: Conelusbn ................................................................................................. 82
References .................................................................................................................... 84
Abbreviated List of Symbols
BSC
ESR
FFT
IFFr
MAI
PSF
i? RBC
RF
ROI
s 10
s 6 0
tA
tF
VOI
backscattering coefncient
erythrocyte sedimentation rate
fast Fourier transform
inverse fast Fourier transfonn
micrmcopic aggregation index
point spread function
correlation coefficient
red blood ceIl
radio frequency
region of interest
the mean kinetic index at 10 seconds
the mean kinetic index at 60 seconds
the primary aggregation time
the final aggregation time
volume of interest
the scattering cross-section
List of Figures
Fig. 1.1 Diluted blood sampIe (20 % hematoctit) nom a patient with coronary
artery disease show ing heavil y networked RBC aggregates (magnified
330 times). Couaesy of Dr. Xiaoduan Weng .......................,.... ..........O............ 2
Fig. 1.2 Diluted blood sample (20 % hematocrit) fiom a normal subject showing
some rouleaux of a few to several RBCs (magnified 375 times) 1531 .................. 3
Fig. 1.3 The vicious cyde triggered by reduced blood flow due to increased RBC
aggregation level. Adapted h m (511. ................... ....... .................... 4
Fig. 1.4 B-mode images of 28 % hematocrit porcine whole blood shwing the
hypoechogenic zone ("black hole") at the center of the vessei. Both
images were obtained at an entrance distance 60 times the diameter of
the tube with a mean velocity at 1.4 cm/s ..................... .. ....................... 1 2
Fig. 2.1 The rotation of the X-Y plane about the z-axis to simulate different
Fig. 2.2
Fig. 2.3
Fig. 2.4
Fig. 2.5
Fig. 2.6
insonification angles. Y represents the direction of propagation of the
ultrasonic waves and X is the axis corresponding to the beamwidth. ................ 32
An illustration of the angle of insonification fiom a physical perspective. ......... 32
A drawing of a small RBC clump and a few rouleaux. ................................... 37
The PSF of the transducer used in the present study (Eq. 2.19). ...................... 40
Sampies of tissue images at an angle of zero degree (Eqs. 2.13 and 2.21). ....... 40
RF images of RBC clump and rouleau mirnics at an angle of zero degree
(Eq. 2.6). ....................... .. .......................................................................... 41
Fig. 2.7 B-mode images of RBC clump and rouleau des (Hilbert
...................................... transformation of Eq. 2.6) at an angle of zero degree 41
Fig. 3.1 Cornparison of the mean backscattered power between simulations of
RBC clumps (60 pm in diameter) and rouleaux (60 pm in leagth). The
error bars for each angle represent one standard deviation obtained from
30 tissue matrices. ......................................................................................... 46
Fig. 3.2
Fig. 3.3
Linear power-volume relationship for isotropie spherical scatteres. The
error bars for each scatterer size represent one standard deviation
obtained nom 30 tissue matrices. The correlation coefficient (I) was
....................... fitted onto the mean backscattered power values. ... ..... ..... 47
Linear power-volume relaîiomhip for anisotropic cy lindrical scatterers.
The error bars for each scatterer size represent one standard deviation
obtained Oom 30 tissue matrices. The correlation coefficient (4 was
fitted oato the mean backscattered power values. .................................... 4 8
Fig. 3.4a Power vs. the diameter of RBC clumps. The data are for 5 different
insonification angles (O0, 22.S0, 4S0, 67.5" and 904. Since no angular
dependence is observed for these angles, al1 data were pooled together
and plotted as one cucve. The error bars for each scatterer size represent
one standard deviation obtained from 30 tissue matrices. ................................. 49
Fig. 3.4b Power vs. RBC rouleau length at 5 different insonification angles. The
Fig. 3.5
Fig. 3.6
Fig. 3.7
Fig. 3.7
Fig. 3.8
Fig. 3.9
error bars for each scatterer size represent one standard deviation
obtained kom 30 tissue matrices .............................. .,l ............................... 50
Power vs. the number of RBCs per aggregate across 5 different
insonification angles. For the RBC clump mimic, data were pooled for
angles of O", 22.S0, 49, 67.5" and 90'. The error bars represent one
standard deviation obtained fkom 30 tissue matrices ......................................... 51
.......... An illustration of the size distribution of RBC rouleaux across the tube. 53
.............................................................. Simulated Wack hole" for Case la. 5 3
........................................................................................ (coat.) ............. .. 54
An illustration of the size distribution of RBC rouleaux across the tube
for Case lb ......... .................................... .................................................. 55
Simulated %lack hole" with the hypo-echogenic ring, appeariag at low
.............................. ............................ angies (O0, 22S0, 45") for Case lb. ... 56
Fig. 3.10 An illustration of the ske distribution of RBC rouleaux across the tube
...... for Case 2. ....................... ... ............,..... 57
........................ ............................... Fig. 3.1 1 Simulateci "black hole" for Case 2. ..... .58
Fig. 3.12 An illustration of the size distribution of RBC rouleaux acr(36s the tube
................................................................................................... for Case 3a 5 9
.............................................................. Fig. 3.13 Simulated "black hole" for Case 3a. 6 0
Fig. 3.14 An illustration of the size distribution of RBC rouleaux across the tube
for Case 36 ............................................................................................. 6 1
Fig. 3.15 Simulated "black hole" with the hypolechogenic ring for Case 3b ..... ............... 61 Fig. 3.16 An illustration of the size distribution of RBC aggregates across the tube
...................................................................................................... for Case 4. 62
Fig. 3.17 Sirnulated %la& hole" for Case 4. ............................................................. -.-63
Fig. 3.18A graphitai illustration of the scatterer structure and mean size
........................... ............................................ arrangement for Case 5. ... 64
...................... Fig. 3.19 Simuiated "black hole" for Case 5. ....................................... 6 5
Fig. 4.1 The transducer PSF (2nd order derivative of H(4y)) in the spatial domain
....................... with the comsponding magnitude spectmm (zoomed version). 6û
Fig. 4.2 Tissue matrix mimicking 20 pm RBC rouleaux at 10 % hematocrit with
the corresponding magnitude spectnim. Note that the DC wmponent has
been removed from the spectnim for better visualization. The RBC
........................................ ........... rouleaux are aligned parallel to the y-axis. .. 6 û
Fig. 4.3 Tissue matrix mimicking RBC clumps of 20 Fm diameter and 10 %
hematocrit with the corresponding magnitude spectmm. Note that the
DC cornponent has been removed from the spectrum for better . . visualization. .................. ..... ........... .. ......... 71
Fig. 4.4 A sample tissue matrix rnimicking RBC rouleaux of 250 pn long with
the corresponding magnitude spectrum. Note that the DC cornponent has
been removed ftom the spectrum for better visuaiization. The RBC
rouleaux are digneci at an average angle of 4S0, with a random
.......................... camponent of I 5". ......................... Fig. 4.5 Aggregation index of normal human RBCs at 45 % hematocrit as a
fundion of the shear rate (adapted fkom [IO]). RBCs were separated
h m the plasma and were suspended in a dextran saline solution. .................... 76
Fig. 4.6 "Black hole" magnitude vs, imnification angle wmputed from the
simulation of Case la The results were expresseci in terrns of mean t one
standard deviation and were averaged over 30 tissue matrices. ... ..................... 78
List of Tables
....... Table 2.1 The sias and concentration of major blood particles (adapted h m [18D 17
.......... Table 2.2 Acoustic properties of blood constituents at 20 O C (adapted nom [49D 27
............................ Table 2.3 The parameters used in the simulation for the present study 36
Chapter 1
Background and Objectives
This thesis is a study of red blood ceIl (RBC) aggregation in human blood by using
ultrasound. In this chapter, the essential background and the objectives will be pcesented.
1.1 The Rationde Behind the Study of RBC Aggregation
The human blood is composeci of plasma and cells. The plasma, in which the cells
are suspended, contains proteins such as dbumin, globulin and fibrinogen; nutrients,
hormones, mineral electrolytes and metabolic end products. There are mainly three types
of cells in blood: leukocytes (the white blood cells), platelets, and erythrocytes (the red
blood cells) which make up over 99 % of the cells in blood [52].
Under nomal physiologicai conditions, the red blood cells (RBCs) may aggregate
into stach called rouleurci; which is a result of the interaction between plasma proteins
and the RBC membrane. The rouleaux may further interact with other rouleaux to form
rouleau networks. Fibrinogen plays a very important part in rouleau cohesion and size
1511, and the RBC aggregation rate is observed to be accelerated with increased
fibrinogen concentration 1321. Aggregates formed in stationary Bow or low flow
conditions will disaggregate at higher flow rates due to the increase in shear forces.
Because RBC aggregation is a reversible process, reducing the shear forces wiii result in
the reaggregation of RBCs.
The aggregability of RBCs, i.e. the tendency of RBCs to form aggregates, has
been shown to play a major role in b l d flow, especially in the microcirculation [3],
where 80 % of the £iow resistance is. It is also known that RBC aggregability is a major
determinant of the viscosity of blood [9]. Under pathological conditions where RBC
aggregability is increased, the blood is more viscous and the adhesive strength between
RBCs forming an aggregate is increased. This c m result in reduced, or even absence of
flow in localized microvasçular regions. Such condition leads to increaseà flow resistance,
the reduction of tissue perfusion, and ischemia, a condition where the tissues are deprived
of the blood supply. In larger vessels, RBC aggregation is believed to be implicated in
thrombus formation 131, where a blood dot formed within a b l d vesse1 perturbs the
flow, may detach to block totally a smaller vessel, and consequently, causes tissue
ischemia Figs. 1.1 and 13 compare a pathological state blood sample to a normal one.
Although the two photographs were taken under static condition, the ciifference between
the two are expected to remain under flowing condition. Reduced blood flow due to the
increased RBC aggregation level may also trigger a vicious cycle [SI] which further
enhances RBC aggregation, as shown in Fig. 13.
Fig. 1.1 Diluted blood sample (20 % hematmritl) m m a patient with coronary artery
disease shawing huvily networked RBC aggregates (magnified 330 times).
Courtesy of Dr. Xiaoduan Weng.
Fig. 1.2 Diluted blood sample (20 % hematocrit) from a normal subject showing some
rouleaux of a few to several RBCs (magnified 375 times) [53].
I Local Acidosis
Fig. 1.3 The vicious cycle triggered by reduced blood flow due to increased RBC
aggregation level. Adapted h m [SI].
RBC aggregation levels, among othec hemorheologic parameters such as the levels
of whole blood and plasma viscosity, are found to be higher in patients with coronary
heart disease than in heaithy subjects [16,23,42]. It is also observed that the RBC
aggregation level provides a good indication of cardiwascular risk [47, as well as
microvascular hemorheological disorders [34]. An enhanœd level of RBC aggregation has
also k e n observed in patients suffering from diseases such as hypertension [46,57],
diabetes mellitus [30], gynacological malignancies [371, and canceis [28]. Thus, the study
of RBC aggregation is very important in clinical hemorheology .
1.2 Methods Proposed for the Study of RBC Aggregatïon
With the exception of the ultrasound baclrscattering method descnbed later, al1 the
methods described below require the withdrawal of blood fiom the patient for the
determination of the RBC aggregability. This may affect the tendency of RBCs to fom
aggregates and consequently the level of aggregation may ciifter fkom that present in vivo
within vessels, Only the erythrocyte sedimentation rate method is used clinically (described
in Section 1.2.2).
1.2.1 Microscopie Observation @]
This method is a static method, and the quantification of RBC aggregation is
achieved by dividing the RBCs into two separate groups, the first k i n g suspendeci in
plasma o r other maaomolecular solutions, and the other in a Knge? solution containing
0.5 % of serum albumin3. The albumin preserves the shape of the RBCs without inducing
aay aggregation.
M e r the RBCs have sedimented to the bottom of a container, photomicrographie
images are taken so that the number of cell units per volume of suspension can be
counted. Note that each ce11 can be a single red blood ce11 or an aggregate of RBCs. Using
the same RBC concentration in both solutions, the average number of red cells in each unit
can be expresseci in ternis of an index called the microswpic aggregation Utdk (MAI):
Number of cells in the Ringer solution MAI =
Num ber of wits in plasma or other macromolecular solutions
A balanctd saline solution uscd in physiologial experimtnfs to provide an isotonic medium for living tissucs. A simple proiein that is bat-coagulabk and watcr-soluble, and prcsent in blood plasma or senim.
Thus, when there is no RBC aggregation, MAI equals to 1. The drawbacb aPsociated
with this method are that the ce11 concentration in the solution must be low (e.g. 1 %) and
only the static condition is reflected, which is not the case in vivo.
1.22 Erythroeyte Sedimentation Rate (ESR) W
As the name suggests, this method relies on the rate of sedimentation of RBCs. A
plot of plasma-cell interface versus time is created, and the maximum sedimentation rate is
obtained f b m the steepest dope of the curve. The ESR should be determined at a Gxed
hematocrit, which provides a convenience in that the ESR can be determined UI witro
simply by using a hematocrit level similar to that in the normal circulation.
ESR reflects the degree of RBC aggregation as larger aggregates sediment €aster.
Again, the weakness of this method is that only the static condition is coosidered, which is
not really relevant physiologically. Moreover, the visc~sity of the plasma, which may differ
between samples of different patients, has an influence on the rate of sedimentation.
1.2*3 Direct Observation Under Shear Flow [8]
Chen [a] pmposed a method to obtain quantitative measures of RBC aggregability.
A computerized image analyzer was developed to acquire images of a single layer of
aggregates flowing in a chamber with controllable flow conditions. The flow chamber was
constructed by sandwichhg a thin layer of metal sheet that was approximately 40 Pm thick
between two transparent plates. A rectangular window of 1 x 20 cm was cut in the metal
sheet to form the flow chamber. Two holes were drilled in one of the transparent plates to
form the inlet and outlet of the fluid. The images of the RBCs in flow were magnified by a
microscope, monitored by a CCD TV carnera and were digitized. The images of the RBC
aggregates were di&rentiated by segmentatiori of the background, and that would give
the projected area of each aggregate. The volume of the aggregate was obtained by
integrating the projedeci are% and the size of the aggregate was obtained by dividiag the
volume by the average volume of a single r d blood c d .
With this system, aggregate size distribution curves at various shear stresses could
be obtained, and this prwided aggregability parametes such as the average or median size
of the aggregate population, the peak aggregate size, as well as the aggregation kinetics.
The shear stress required for complete disaggregation could be known by varying the
shear stress applied.
The problem with this method is that the chamber imposes a constraint on the
aggregate volume and size as only a single layer of aggregate is allowed to pass through.
This effectively forces a three dimensional network of aggiegates into a two dimensional
version, which is not entirely valid in physiologicaf conditions, especidly for large vessels.
The thickaess of 40 pn of the flow chamber may mimic the conditions in the
microcirculation, but the width of 1 cm does not; therefore the aggregate size estimated
and al1 the statistics derived rnay not be relevant to what adually occurs Ui vivo. This is
especially true for RBC aggregates in pathological states where the cross-section of an
aggregate may mach beyond the dimension of the chamber.
1.2.4 Laser Llght Reflection Method [17]
This method relies on the study of the variation in laser intensity refleded by the
scatterers (RBCs). Two series of measurements c m be performed. The first is to apply a
variable shear rate to the blood sample and measure the refleded light intensity at each
shear rate. As the shear rate increases, the aggregate ske decreases, leading to the
decrease in scatterer spacing and the increase of the refleded intensity. A curve showing
the refleded light inteusity versus the shear rate can be obtained for each blood sarnple.
From the curve, parameters on the adhesive strength holding the RBCs together can be
measured.
The second set of measurement gives the aggregation kinetics of the b l d simple,
Le. the tendency of RBCs to form rouleaux at zero shear rate. The blood is s h e d for 10
seconds at 550 S-' to disrupt rouleaux and orient RBCs with the flow. The refleded light
intensity at that stage is recorded. The shear stress is then terminated abmptiy, which
results in the reagpegation of RBQ. The variation of the refleded light intensity is
recorded during the rouleau formation process and a curve showing the reflected Iight
intensity versus time is obtained. From this curve, the primary aggregation time (tA), the
finai aggregation time (tF), and mean kiaetic indices at 10 s (Sio) and 60 s (&O) are
evaiuated. The aggregatioa proass is assumed to be completed 2 min. after the Bow
stoppage. T o date, the laser light refledion method is the most reliable approach for the
measurement of the dynamics of RBC aggregation in vitro. However, as other methods,
the disadvantage is that blood withdrawal is still required.
1.15 Ultrasouad Backscattering Methds
Each of the methods described above has its own advantages, but none of them
permits the study of RBC aggregability in vivo. Oniy noninvasive methods such as
ultrasound has the potential to do so, and in red time.
Several research groups have shown that ultraswnd is sensitive to the presence of
RBC aggregation. Boynaid et al. [SI used ultrasound in an in vitro set up to measure the
backscattering intensity of an RBC saline suspension, which do not produce RBC
aggregation. They made an attempt to relate the mean size of RBC aggregates to the
backscattered intensity [4]. They have also compared the RBC aggregation in eldedy and
young subjects using ul trasonic i nterferometry and backscatteri ng met hods [24], and were
able to observe the difference in RBC aggregation levei between these two groups of
subjects. However, their expenments were not conducted under flowing conditions and
relied on the sedimentation of RBCs.
Kim et aL 1291 demonstrated in an experiment that the ultrasonic baclcpcattered
intensity increased afler the blood sample in an oscillatory tlow was abruptly stopped.
They also observed a rapid decrease in echo intensity when the fluw was resumed. Sirniiar
results were obseived by Yuan and Shung [55] where the ultrasoaic backatter h m
flowing whole blood was fouad to be different fkom that of RBC saline suspensions in that
the former is shear rate dependent. The experiments were conducted with animal blood
ftom different specis, and the rcsults pointed to a species dependent backatter which
could be explained based on the degree of aggregation of the species' blood sample. In
addition, the same group also sbowed that ultrasonic backscatter Eiom flowing whole
blood was dependent on the concentration of fibriwgen when RBC aggregation exists
[54]. Shehada et al. [48] also reported an inverse relatiooship between the ultrasonic
echogenicity and the applied shear rates in an experiment conducted with 28 % hematocrit
porcine whole blood.
AU the studies conducted above were in vitro studies. Recently however, Cloutier
et al. [14] demonstrated that a difference in RBC aggregation levels between veiis and
artenes of normal subjects and patients with hyperlipidemia4 could be observed in vivo
with the use of power Doppler ultrasound. Thus, this study contïrms the possibility of
assessing RBC agpgability in vivo and noninvasively.
Al1 the results prwided by the different research groups above can be summarized
as follows: A correlation exists between the size of RBC aggregates and the ultrasonic
backscattered power. When the blood sample is in motion, the shearing effects of the flow
cause the RBC aggregates to break apart, which lead to smaller aggregates and weaker
backscattered power. Cowersely, when the flow is ceased or reduced, the shearing effeds
decay accordingly and consequently, the RBCs reaggregate, which is refiected by a
stronger backscattered power. Blood samples with M e or no aggregates such as bovine
whole blood or RBCs suspended in saline solutions are independent of the shearing effeds
- -
A condition where an excessive amount of fat or lipids is pxcscnt in the blood.
of the flow. Since fibrinogen auxlerates the RBC aggregation rate [32], it is no surprise
that the fibrinogen concentration affects the backscattered power.
13 Proposed Study
Fmm the above information, it is clear that the ultrasound backscattered power
can be used as a signature to study RBC aggregation. However, although it is h o w n h m
the literature t hat RBC aggregation increases ultrasound backîcattenng, the exact
mechankm by which the power is iaawed is unlmown. The volume of the aggregates
certainly has an effect, but other factors such as the hematocrit, the packing structure of
the aggregates, the variance in the aggregate volume, and the fluctuation of ail these
parametes in time and space can al1 contribute. In the present study, the iduence of the
volume of the scatterers and their structure will be specifically addressed. The contribution
of both factors in explaining what is Lnown as the "bblack hole" phenomenon will also be
i nvestigated.
1.3.1 A Brief Review on the Hypoechogenic Zone Discovery
The observation of blood echogedcity clifferences in ultrasonic B-mode images
was first reported by Sigel et aL [SOI. An ultrasonic scanner was used to scan the
surgicdly exposed inferior vena cava and portal veim of eight anesthetized dogs, and
hypoechogenic zones were observed immediately downstream fkom the entry of the two
tributary veins. Such hypoechogenic zones could be traced back to the tributaries since the
echo in the large vein was strooger. The cause of such hypoechogenic zones was
attnbuted to the absence of RBC aggregation in blood originating from those tributary
veins in which the shear rates were higher.
More recently, Yuan and Shung [56] observed the presence of a hypoechogenic
zone at the center of the flow conduit while imaging porcine whole blood under laminar
fiow. This observation, hiown today as the "black hole", was obtained by collecting data
at a position well beyond the entrance length of the vessel. They reported that such a zone
appeared to be more liiœly to occur at higher hematoait levels. Under similar Qow
conditions, no signïficant echogenicity variation was observed for bovine whole blood,
which does not have a stioag tendency to form aggregata. The authors suggested that the
cause was more likiely to be due to a low level of ultrasonic backscatter rather than
attenuation, and that a very low local hematocrit couid be the reason. They îuxther
proposed that local motion of RBCs in the b l d flow could result in such variations in
hematocrit- Radial migratory behavior of particles in a fluid is a phenornenon Lnowa as the
Segre-Silberberg effect. It was pstulated that as the erythrocytes migrated radiaily h m
the center axis to the wail and vice versa, an equilibrium was reached, resulting in a
maximum ce11 concentration region somewhere between the tube axis and the wall, thus
leading to the hypoechogenic zone in the center of the vessel.
la another study conducted by Mo et al. 1401, porcine whole blood at 28 %
hematocrit was circulated in a flow loop model mimicking a large vessel, and cross-
sectional and horizontal-plane B-mode images were obtained at different entrance lengths
with an ultrasound Iinear array transducer. The flow loop model had shear rates ranging
from O at the tube center to 5.3 s-' at the wall. Again, a hypoechogenic zone was present
at the center of the vessel, but the zone was observed to develop slowly as the entrance
length increased. Near the entrance of the tube, the echogenicity was very low; further
downstream, the echogenicity increased around the tube axis and the hypoechogenic zone
was developed. Experiments were also conducted with porcine RBCs suspended in saiine-
plasma mixtures with proportions ranging from pure plasma to pure saline solution. The
hematocrit was kept constant at 28 % in al1 mixtures. It was noticed that the
hyperechogenic ring around the hypoechogenic central zone decreased in echogenicity as
the proportion of saline increased, leading to a l e s distinct "black hole". From these
observations, the aut hors proposed the following:
Echogenic variations across the tube are related to the degree of RBC aggregation.
The hypoechogenic zone arises h m disaggregated RBCs having imufficient time to
significanly reaggcegate by the time they reached the other end of the vessel.
Aithough high shear rates (> 10 s-') generijliy nsult in disaggregation, the aggregation
level for low shear rates (e.g. 02 S-3 could be greater than that at zero shear rate
[10,15] due to the imreased i n t e d o n among the RBCs. Thus having the shear rate
ranging fkom O to 53 s-' might produce a condition where the degxee of RBC
aggregation is maximum at an intermediate radial location leading to a hyperechogenic
ring.
Fig. 1.4 shows sample B-mode images of the hypoechogenic zone adapted from
[40]. The image on the ieft was obtained by positioning the transduar dong the tube axis
whereas the one on the right was acquired cross-sectionally:
"Black hole"
Fig. 1.4 B-mode images of 28 % hematocrit porcine whole blood showing the
hypoechogenic zone ("black hole") at the center of the vessel. Both images were
obtained at an entrance distance 60 times the diameter of the tube with a mean
velocity at 1.4 cm/s.
Shehada et aL [48] further investigated the "black hoie" phemenoa, and they
proposed that the shear rate was high enough to dismpt the RBC aggregates at the tube
entrance, leaciing to a unifom echogenicity. Ar the tlow was M y developed huiher
downstream, the appearance of a hypoechogenic zone at the center of the tube was
proposed to be caused by shear rates lower than 0.05 s-'. They further suggested that the
hyperechogenic ring surrounding the "black hole" was the result of shear rates favorable
for the formation of RBC aggregates (0.05 to 2 9'. approximately).
Using a Doppler ultrasound system, Qin et aL [43,44] obsenred a hypoechogenic
zone at the center of the tube in experiments conducted with home blood. The entrance
length used in the experiments was sufficiently long to allow rouleau build-up, and the
flow rates were vacied €tom 102 ml/min. to 1250 &min. to obtain a wide range of shear
rate across the tube. The data was wllected at 5" increments kom 40" to 70" where the
angle is between the tube axis and the transducer. That was to confinu their hypothesis
that the orientation of large rouleaux in addition to their size contributes to the "black
hole" phenomenon. Since the refkaction of sound waves at the vessel wall could affect the
Doppler backscattend power as the angle was changed, only relative power was
measured. The ratio between the maximum and minimum power across the vesse1 was
used as a measure of the "black hole" magnitude. It was found that the power drop at the
center of the tube was more pronounced for hyperaggregating blood samples, and that it
was not related to the flow rates. In addition, the magnitude of the "black hole" increased
as the angle was raised, suggesting the psibi i i ty that rouleau orientation contributes to
such phenomenon.
Finally, in a set of experiments conducted by Cloutier and Qin [12], it was
observed that the hypoechogenic zone was oniy observed occasionally in the middle of the
vessel in experiments conducted with porcine blood. Those blood samples exhibiting such
a zone did not show any trend of hyperaggregation. However, the 'Wack hole" was almost
always observed in experiments conducted with home blood, as noted by Qin et al.
[43,44]. H o m blood is known to have a strong tendency to fonn long chairs of rouleaux
[53], however this may not be the case for porcine whole blood.
1.3.2 Summary
To the best of our howledge, Sigel et aL [SOI were the first t o identify
hypoechogenic zones downstream from the entry tributary veins in dogs. Since the
scanneci position was relatively close to the eatrance, they proposed that their observation
was most pmbably due to the lack of RBC aggregation as most of the aggregates got
disrupted at the entrance of the receiving veim where the shear rates were high. The cause
of the phenornenon did not appear to be that simple when the hypoechogenic m n e was
obsewed at positiom far beyond the entrance Iength and at the center of the tube where
the shear rate is minimum [56]. It was then hypothesized that the hypoechogenic zone o r
the "black hole" was due to the Segre-Silberberg effect, a phenornenon where particles in
a fluid cross-migrate radially such that an equilibrium was achieved and resulted in the
hypoechogenic zone.
Using porcine RBCs suspended in saline-plasma mixtures at a fhed hematocrit
with varying amount of saline content, Mo et aL [40] later confirmeci that the echogenic
variations obsewed in B-mode images were related to the degree of RBC aggregation,
thus casting some doubts on the Segre-Silberberg effect proposed by Yuan and Shung
1561. The dependence on the degree of RBC aggregation was also observed by Qin et al.
143,441. Mo et ai. [40] suggested that the hypoechogenic zone arose from the
insufficiency in time for the RBCs to reaggregate as they reach the end of the vessel, and
together with Shehada et al. [48], they proposed that the hypoechogenic zone was caused
by shear rates lower than 0.05 s-l at the center of the tube. In the last two studies, the
hyperechogenic ring around the Wack hole" was postulated to be due to shear rates
favorable for the formation of RBC aggregation (0.05 i1 to 2 s", depending on the flow
rate). Closer to the wall of the tube, the higher shear rate reduced the aggregate sizes and
the echogenicity.
Qin et al. [43,44] postulated the prrsence of organized structure in the
hypoechogenic zone, and proposeci that the shape and organization of the scatterers may
contribute to the %la& hole". Thus, the proposais by Mo and Shehada rnay not be the
only reasons. Aithough the experiments by Qin were conducted at much higher shear rates
( b m 1 s" to 55 d), the blood samples used were home whole blood, charaderiad by
strong intercellular lioks. This may explain why the "black hole" wuld be observed at a
high shear rate.
1.4 Objectives
Qin et al. [43,44] postulated that the shape and structural organization of RBC
rouleaux may play a role in the formation of the "black hole". In the present study, we
propose to study such phenomenon with a theoretical model based on a system approach.
The model provides the conditions of an ideal environment, where parameters ordinarily
beyond human control in an actuai laboratory setting can be managed. Examples of such
parameters are the size, orientation and number of scatterers within a given sample volume
of the blood vessel. Using a model also allows the possibility of performing "experiments"
beyond the capabilities of devices currently available in the laboratory, thus providing the
freedom from the constraints of existing apparatus and hardware. The following are the
objectives of the present thesis:
1. To study the effkds of the scatterer shape and size toward the backscattered power
with a theoretical simulation at various imnification angles.
2. To propose possible scatterer organization, shape and size that may contribute to the
"black hole" phenomenon.
Chapter 2
Theory, Implementation and Methodology
In this chapter, a brief introduction on the scattering of ultrasound by blood is first
presented. This is then followed by a description of the system-based mode1 used in the
present study. The parameters involved in the implementation and the methodology wiil be
presented Iast.
2.1 Scattering of Ultrasound by Blood
2.1.1 RBCs as Scattering Targets
Fluids have the two essential characteristics to support the acoustic wave
propagation: elasticity and inettia, which can also be expressed in terms of compressibility
K and mass density p, respediveiy. Elasticity implies that any deviation of the fiuid
molecules fkom the state of equilibrium wili tend to be corrected in the opposite direction.
Inertia implies that such correction will have the tendency to overshoot the state of
equilibrium and thus, requiring the need for a correction in the opposiie direction.
When an acoustic wave propagates through a homogeneous lossless medium, it
can travel indefinitely, However, in reai biological tissues, neither the compressibility nor
the density is constant throughout the entire media. Rather, they fluctuate €tom one spatial
location to another. An example is blood, where RBCs and other types of cells and
macroproteim are suspended in plasma. It is the differences in the compressibility and
density between these particles and the surrounding plasma that result in the scattered
acoustic waves. In general, other processes do occur when an acoustic wave propagates
through an inhomogeneous medium, and these indude the reflection and reîkaction of the
wave by tissue intertaces, as well as the absorption of the acoustic energy and its
conversion into heat by the medium. The backscattering of ultrasound by blood is largely
due to the RBCs since they are significantly more numerous than the slightly larger
leukocytes, and much larger and more wmerous than the platelets [18] (Table 2.1).
Erythrocytes (RBCs)
Concentrat ion
@articles/mm~
5 x 106
Approximate
dimensions (pm)
% of the total blood
voiume
Table 2.1 The sizes and concentration of major blood particles (adapted h m [18D.
The scattering of ultrasound by RBCs depends on the size of the individual
scattering target, the RBC concentration, and the acoustic properties of the scatterers. In
most commercially available clinical ultrasound instruments, the frequency range of the
acoustic wave is generally between 2 and 30 MHz, which corresponds to wavelengths of
approximately 785 pm and 52 Pm, respectively. Note that such wavelengths are much
Iarger than the greatest dimension of an RBC, which imply Rayleigh scattering.
2.1.2 Rayleigh Scattering
The backscattering coefficient (BSC) is a common parameter used to characterize
ultrasound scattering by blood and tissues. BSC is defined as the powr bacb t t e r ed by a
unit volume of scatterers per unit incident intemity, per unit solid angle in the direction
opposite of the incident wave [49]. For the case when Rayleigh scattering occurs, i-e.
when the incident acoustic wave encounters a particle much smdler than its wavelength, a
portion of the acoustic energy is scattered uniformly in ail directions. Such scattering was
first studied by Lord Rayleigh in 1871, and thus the oame Rayleigh scatterbtg is
commonly used today to refer to such phenomenon 1451.
When an incident wave encounters a group of scatterers in the medium, the
acoustic waves scattered uniformly 6rom each particle will result in mulîiple scattering, a
phenomenon where a scattered wave €tom one particle encounters other particles in the
area and is rescattered again. In theory, multiple scattering may go on indefinitely. In
practice however, multiple scatteri ng can be safely ignored especiail y for weak scatterers
such as RBCs since acoustic waves rescattered by neighboring RBCs are usually so weak
that their effects are negligible. Multiple scattering will not be considered in the
simulations of the present study.
Another commonly used parameter in the characterization of the ultrasound
scatteriag properties of blood is the scattering cross-section a@), which is the ratio of the
total scattered power, S, to the intemity 1 of the incident wave on one single target [33];
Le. a = SM. Under Rayleigh scattering condition, a@) is defined as:
where u(B) is the scattering cross-sectional area of a sikgle scatterer, B is the angular
direction of the scattered wave measured with respect to the direction of the incident
wave, V, is the scatterer volume, h is the wavelength of the incident wave, and K, p, and
K, p, are the compressibility and density of the scatterer and of the surrounding medium,
respeaively. Note that the scatterer is assumed to be spherical in shape in this model.
From Eq. 2.1, one can see that the baclcrcattering cross-section, ah for 0 = 180" is
predicted to be proportional to the square of the particle volume, and to the fourth power
of the signal âequency Cf = c h where c, the speed of sound in blood is appmximately
1570 m/s). In pradice, the b a c b t t e r e d power due to a single scatterer cannot be
measureci. In the following sections, it will be shown that the presence of several scatterers
and their spatial-temporal arrangements affect the backscattered power of the signal.
2.13 The Modeüng of Ultrasound Bockscattering by RBCs
Under normal physiological conditions, RBCs in human blood are densely packed,
and on average, the separation between adjacent cells is l e s than 10 % of the diameter of
a single RBC [49]. Such condition implies that the positions of any pair of RBCs are
neither totally uncomlateâ, nor perfectly correlateà, Le. they are partially conelated. This
essentially means that the acoustic wave scattered from each individual RBC can either
interact comtructively or destructively, making the modeling of ultrasound signals very
difficult. Such unpredictable interaction of the backscattered waves has a direct influence
on the total backattered signal power, and this can be d ina ly litked to the spatial-
temporal arrangements of the RB& at the time of insonification.
Several approaches have been pmposed in the past to model ultrasound
backscattering from RBCs, and they can be roughly categorized into the particle, the
continuum, and the hybrid methods. The following sub-sections were adapted fcom
Chapter 5 of [49].
2.13.1 The Particle Approach
T o the b a t of our howledge, Brody and Meindl [6] and Albright and Hanis [l]
were the first to model ultrasound scattering by the particle approach. The scattering
medium for [6] was assumed to be made up of a collection of identical, independent point
scatterers which scatter isotropically. It was also assumed that the average number of
scatterers per cubic miliiliter at any given time was Poisson distributed. Both research
groups treated the backscattered acoustic waves as the sumat ion of the contribution
€rom each individual scatterer.
Mo and Cobbold [38] developed a more general mode1 by tceating blood as a
suspension of RBC aggregates whose e&divc volume was a random variable, rather than
independent, point-sized paxticles. This approach was used to simulate the random
distribution of RBC aggregate sizes. The size of the agpgates in this model was
considered much smaller than the wavelength. Based on this model, the backscattering
coefficient, BSC, was given by:
where ob is the backscattering cross-section, H is the hematocrit, V, is the mean volume
of the scatterers, and W is the packing factor, which can be perceived as a measure of the
spatial orderliness of RBCs. The more "orderly" the scatterers are, the lower is the value
of W. Thus W is unity when the packhg of the scatterers is wmpletely randorn, and
gradually approacha zero when the correlation between RBCs increares. In terms of the
hematocrit, increasing the number of scatterers in a 6xed volume wiil invariably result in a
greater orderliness in terms of the spatial arrangement of the scatterers. At high hematocrit
levels, one can always find a scatterer that interfere destructively with the backscattered
acoustic waves fkom another scatterer, leading to a reduced BSC. Therefore, at a given
hematocrit, BSC is proportional to the scatterer volume weighted by the packing factor,
and to the fouah power of fkequency (see Eqs. 2.1 and 2.2). At very low hematocrit
levels, W = 1 and BSC is proportional to H. In Eq. 2.1, the values of the density and
compressibility of the medium are fixed, which can only be either the value iaside o r
outside the particle. Most of the particle-based models can be considered as special cases
of Eq. 2.2, w here W is expressed as an explicit hinction of
2J3.2 The Continuum Approach
This approach treats blood as a continuous medium where the backscattered
waves originate fiom local fluctuations in density and compressibility rather than nom
individual particles. This is based on the argument that the spatial cesolution of
conventional ulttasonic transducers is limited, and that individual RBCs and small RBC
aggregates are much smaller than the wavelength such that they ~ n n o t be tesolved
individually by these transducers.
Compared to the particte approach, the continuum approach models the
backscattering of ultrasouad as the summation of al1 contributions fkom the entire acoustic
field rather than tracking the position of each RBC in the sample volume. Thus, the
variations in the density and compfessibility are d o m for the continuum approach,
reflecting the random spatial distribution of the scatterers in the medium. Mo and Cobbold
(Chapter 5 of [49D proposed an expression for the BSC using the continuum approach:
where the subscript m in the equation indicates the average of the conesponding
parameter in the random medium; IL,,, = cJf is the average ultrasonic wavelength in the
random medium where c is the average speed of sound and f is the ûequency; V. =
( 4 d ) / 3 is the effedive ce11 volume such that r is the effective radius; K- p. and ic, p, are
the compressibiIity and density of the scatterer and of the surrounding medium,
respectively; ne is the volume of a voxel, which is a fixed elemental blood volume; and
var(n) is the variance in the number of scatterers in Re obtained by averaging over space
and time.
The average wavelength is used because the speed of sound depends on the
medium it propagates in. Thus when the ultrasonic wave propagates through a random
medium, the wavelength changes accordingly. This explaim why A,,, depends on the
hematocrit level in this model.
2.133 The Hybrid Appmach
Mo and Cobbold 1391 proposed this approach which combines the stmngths of the
two approaches above. In this model, the RBCs are no longer treated as individual
scatterers as in the particle approach, rather the sample volume is divided into elemeotal
acoustic volumes d l e d wxels, such that each voxel contains a variable number of RBCs
treated as a single scattering unit. Each voxel mwes with a single velocity. The BSC for
this approach is obtained by determining the contribution Grom a single scattering unit, and
theo summing the cont~butions nom ali the voxels. The influence of the mean number of
scatterers per voxel as well as its variation in tirne is included as part of the ~onsiderations~
Mo et al. propased the BSC for this approach to be:
where ab is the backscattering cross-section as defined in Eq. 2.1 for 0 = 180°, var@) is
the variance in the number of scatterers per voxel, and 8, is the volume of a voxel. In this
model, a voxel is defined in terms of a &action of the wavelength A.
2.2
the
The Modeüng of Ultrasound Backscattering by RBCs with
System-Bmed Approach (adapted fiom [35,36])
This approach to the modeling of ultrasound backscattering was selected for the
present study mainly because it provides the afcess to the specification of the shape, size,
orientation and number of scatterers, as well as the possibüity of insoniQing a sample
tissue image at any desired angle. Such features, which may not be readily available with
the models described above, make this approach very suitable for the objectives of the
present study. Other strengths and limitations of this approach are presented in Section
4.2. The following is a brie€ description of the formulation and the theory behiad the
system-based approach, which uses the principle of superposition applicable to linear
systems.
Assume that an ultrasound system is used to transmit a pulse of acoustic wave into
a fluid medium containhg only a small particle suspended in it. Then the backscattered
acoustic wave is chacaderized by the differences in the compressibility and density
between the particle and its sumunding fluid (Section 2.1.1), and the ultrasound system is
characterized by the radio frrquency (RF) signal at the output of the system, which is
essentially the point spread funcîion (PSF) of the system. If one is able to model the
acoustic irnpedance of a particle (which depends on the differences in the compressibility
and density between the particle and its sumiunding Buid), and the PSF of the ultrasound
system, then a simulated version of the backscattered RF signals can be obtained: When
there are more than one particle suspended in the fluid (under certain assumptions to be
given later), the backscattered R F signal at the output of the ultrasound system is
essentially the weighted PSFs that are summed over the space where the particles are
Iocated. Each weight essentially represents the acoustic irnpedance of each individual
scatterer at its respective location.
Since each scatterer is wnsidered in this model, it can be perceived as a particle
approach. However, the modeling is totally different €rom that presented in Eq. 2.2. In the
present study, if the region interrogated is sufficiently small, the PSF is essentially space-
invariant, and thus the summation over d l scatterers can be replaced by a convolution
operation. Therefore, the backscattered RF signal from a volume of interest (VOI),
W3D(&y,z), c m be expressed as:
where H&(%yYz) is the 3-D PSF of the ultrasound system, O denotes the convolution
operation, and Zm(%y,z) is the acoustic impedance hinction. The second order derivative
with respect to y refers to the direction of propagation of the acoustic waves. Note that
RFzo(~y,z) is a "volume" of RF signais- For the ease of amputation, W'(x;y,z) can be
well reprcsented by a 2-diwmional version of the RF image, RFm(5y). This is shown to
be possible in [35,36] if H3D(qy,z) is separable, that is, Hui(qy,z)=H(%y)H(t)- H&y,z)
wiii be defined later and wiii be shown to be separable. Thus the RF image cm be
described as:
Since the region of interest is composed of a coiledion of scatterers, the t
impedance, Z&(&yy~) can be beumed to be an ensemble of small scatterer
inhomogeneities:
where n indicates the location of the scatterer in the VOI. Ideaily, if the scatterers al1 have
the same size and shape, then only a universal scatterer prototype is required to generate
Z(i,y) [35]. In that case, a scatterer prototype function can be defined and repeated at
each scatterer location, i.e.:
where a. represents the echogenicity of the scatterer at position n and S3&) is the
scatterer prototype huiction. Substituthg (2.9) into (2.8) and then into (2.7) gives:
For the case where the scatterer size is small compared to the thickness of the beam in the
z direction, then the value of H(-z) will remain constant within each scatterer volume.
With respect to each scatterer, this implies that H(-z) varies only from scatterer to
scatterer and not in the z direction &&in any particular scatterer location. Therefore H(-z)
c m be taken out of the integral, which results in:
n
w here
This means that the 2-dimemional tissue impedance function Z(x,y) c m be generated by
projecting the scatterer prototype fundion dong the >axis and then weighting the result
with H(-zJ in each scatterer location. Z&y) c m also be expressed as the Z-dirnensional
convolution between S(x;y) and a position matrix P(k,y):
where
n
P(i,y) is essentially a matrix with randomly positioned Dirac delta fiindions weighted by
H(-zJ and a.. Eq. 2.13 meam that one wiii get an "ideal" tissue matrix, where ail
scatterers simulated have the same s h , shape and orientation. Having obtained Z(l;y), the
2-dimensional RF image M'&%y) can be computed according to Eq. 2.6.
In the event that a more realistic tissue image is required, i.e. a unique shape, size
and orientation for each scatterer, then each of the scatterers has to be integrated
independently and summed together to create the tissue matrix Z(l;y). A 2-D tissue
irnpedanœ matrix for each scatterer would then be obtained by:
and
where S.(gy,z) detemines the shape, size and orientation of the n * scatterer. This
essentially considers the entire V01 to contain only that particular scatterer, and have it
weighted with a, and H(-zJ, and have the result integrated along the z-axis. That wiil
generate a 2-0 tissue impedance matrix for that specific scatterer (Eq. 2.15). This
procedure is then repeated for each of the scatterers, and al1 the 2-D impedance matrices
are summed to produce the tissue impedance matrix, Z(qy) (Eq. 2.16). Note that the
creation of Zky) involves a position matrix as well. The 2dimeosional RF image
RFu>(Sy) can then be obtained h m Eq. 2.6.
The following are the main assumptions used in the system-based approach in the
modeling of ultrasound b a c b t t e r i n g by RBCs.
1. The medium is weMy scattering, thus multiple scattering is negligible.
There is no aîtenuation by the medium.
The noise level is zero.
The scatterers are in the fhr field of the transducer.
The PSF of the ultrasound system can be cepresented by a cosine modulated by a 3-
dimensional Gaussian envelope.
The ultrasound beam is sufficiently large such that the scatterers are small wmpared to
the beam thickness, i.e. H(-z) in Eq. 2.10 cemains constant within each scatteter
volume.
AH scatterers have equal echogenicity a,.
The region of interest (ROI) is sufficiently small such that the PSF is space invariant.
The assumption of weak scatterers essentiall y means that the mismatches in deaîity
and compressibility between an RBC and the plasma is Edidy smdl. As recognized in the
literature 1491 and shown in Table 4.1, this condition is satisfied for blood:
Medium
RBCS
Plasma
0.9 % saline
Distilled water
Table 2.2 Acoustic properties of blood constituents at 20 O C (adapted from [49]).
Mass density, p (g/cm3)
1.092
1.021
1.005
0.998
A weak scattering medium also implies that multiple scattering does aot have a major
contribution on the backscattered power. This is because the scattering from a particle is
so weak that when the acoustic wave bounces on neighboring particles, the rescattered
waves are negligible. The attenuation of ultrasound in blood and tissues is attributed to
scattering and absorption. The contribution to attenuation due to scattering is srnail
whereas absorption in blood is a p p d m a t e l y linearly proportional to the hemoglobin
Adiabatic compressibility, K (IO-= cm/dyne)
34.1
40.9
44.3
46.1
concentration b e l w 15 % [22]. In the cumnt study, the depth position of the ROI was
not considered thus the attenuation within the blood sample could not be wmpensated.
The assumption of zero noise level is just for the sake of simplicity, it can be easily
incorporated into the model by introducing some noise into the RF image. Under normal
clinical conditions, ultrasound measunments are perfomied in the far field region, thus
assumption 4 is consistent 4 t h what is done in ceality. Since the PSF can be seen as the
echo obtained when the scatterer is a small point object, a m i n e function modulated by a
Gaussian envelope can be a good appmrrimaîion. Note that a Gaussian fwction is a h
separable (see Eqs. 2.17 and 2.18 on page 30), which was another motivation for choosing
this model.
For the simulations of RBC rouleaux, the change in the scatterer volume was
refleded in the length of the rouleau while its diameter remained constant. For the RBC
clump mimic, the diameter of the spherical clump increased with the scatterer volume. In
the present study, as described later, the largest clump mimic used in the simulations had a
diameter of 120 p, which was still considerably smaller than the beam thicimess of a
typicat ultrasound transducer which is on the order of a millimeter. Thus, the assumption
that the barn thiclmess was large enough such that If(-z) remained constant within each
scatterer was valid for our study. As long as this assumption holds, H(-z) can be omitted
fiorn the mode1 because it has no effect on the variation of the backscattered power as a
fundion of the scatterer size and structure. Assumption 7 is reasooable because in
practice, although the echogenicity related to the density and compressibility of RBCs may
change, this has never been demonstrated.
In practice, the point spread function (PSF) of an ultrasound system in the far field
camot be considered space invariant and it is hiown to decrease in intensity with depth
[Il]. Moreover, a point scatterer at different depths wiU appear to have different sizes on
a B-mode image because of the divergence of the beam. However, if the interrogated area
is sufficiently srnall (1.28 mm by 1.28 mm in this study), the hypothesis of the space-
invariance of the PSF may be valid Using this hypothesis allows the possibility of
producing the RF image by convoluting the same PSF with a tissue image (Eq. 2.6), for
each position x. and y. of the scatterers within the ROI.
To implement the system-based mode1 of Eq. 2.6 on computer, the task can be
summarized into the following steps:
I. Create the PSF matrix % ~ ( x , y) (Eq. 2.6), and rotate the PSF about the center ay
of the matrix to simulate different angles of insonification. In the present study,
angles varying between O and 90" were tested.
II. Create the tissue matrix Z(qy) (Eq. 2.13).
A. Create a scatterer prototype S(i,y) (Eq. 2.12).
B. Create a position matrix P(i,y) (Eq. 2.14).
C. Compute the discrete convolution of S(5y) and P k y ) in the fkquency
domain.
III. Create the 2-dimemional RF image RFm(&y) by computing the discrete
a 2 convolution between - H(x, y) and Z(x,y) in the fiequency domain (Eq. 2.6).
ay
IV. Compute the average power of M'&,y) as a measure of the backscattering
coefficient.
The functions used to generate the matrices H(gy), Z(qy), S(i,y) and P(k,y) will be
defined below. AU convolutions were perfomied in the fkequency domain to enhance the
computational speed. AU matrices had the same size which reflects the size of the sample
volume.
For the present study, al1 scatterers in each tissue image were assumed to be
onented in the same direction and they dl had the same shape and size. Such an ideai
condition was used in order to reduce tissue image generation time, as convoluting each of
the scatterers one at a time, as described in Eq. 2.16, is computationally very intensive.
The justification of using such ideal conditions is presented in Section 4.1.15.
Step 1:
a 2 The focus of this step is to create the PSF fiinction -H(x,y) in Eq. 2.6. In this
?Y2
study, an approximation of the k-field PSF was used. In Eq. 2.5, the 3-dùnensional PSF,
H'.&,y,z), could be represented by a cosine function modulated by a 3-D Gaussian
envelope fwction, as justified before:
Expressing H3&,y,z) in tenns of H(x,y)H(z), we obtain:
and H(z) - e
where ly, , I#,, , and 2LZ were the standard deviations of the 3-dimeasional separable
Gaussian function which were the parameters controlling the beamwidth, the bandwidth
(length of the PSF) and the ultrasound beam thichess, respective1 y.
From Eq. 2.17, the first order denvative of H(i,y) is:
where E(x, y) = e
and the second order derivative of N(;re;y) is:
From the equation above,
Having obtained the second order derivative of H(x,y), its rotation about the origin
could be accomplished by first rotating the a i s by a desired angle, and then implemeating
the function on the rotated axes, as illustrated in Fig. 2.1.
Fig. 2.1 The rotation of the X-Y plane about the z-axis to simulate different insonification
angles. Y represents the direction of propagation of the ultrasonic waves and X is
the axis corresponding to the beamwidth.
Any arbitrary point on the X-Y plane was mapped onto the P-Q plane through the
celationship shown in Fig. 2.1. Since the 2-dimeasional Gaussian hindion was applied
onto the rotated axes, no inteplation was perfomed on the PSF matrix and therefore,
there were no interpolation errors introduced. Fig. 2.2 shows the angle of iosonification
with respect to the vesse1 fiom a physical perspective, where 8 is the angle of
insoni fi cation.
Tube wall
Flow
direction
1 Tube wall
Direction of / Y'' ul trasound
d propagation
Fig. 2.2 An illustration of the angle of iasonification h m a physical perspective.
Step 2:
aL Refemng to Eq. 2.6, having obtained ZH(x9y), the focus of this step is to
dy
create the tissue impedance function Z(qy), which is the comrolution between a scatterer
prototype function S(x,y) and a position matrix P&y) (Eq. 2.13).
Step 2a:
The focus of this section is on the creation of the scatterer prototype function
S(4y). A 3-dimensional separable Gaussian furiction was used to mode1 S 4 5 y , r ) (to be
justified in Section 4.1.1.4):
Frorn Eq. 2.12,
which is a 2-dimemional Gaussian function weighted by the thickness of the scatterer in
the z-direction. Note that a, O, and a= are the standard deviations of the 3dimensional
Gaussian fundion representing the width, length and depth of the scatterer prototype. The
2-dimeasional scatterer prototype matrix S(+) can be used to npresent eit her individual
RBCs, RBC rouleaux, or clumps of RBCs in pathological cases. This result in a tissue
matrix composed of scatterers al1 identical to one another in temis of size, shape and
orientation.
Step 26:
From Eq. 2.13, the task now is to obtain the scatterer position matrk P(5y).
Refemng to Eq. 2.14, P(qy) is a 2dimensional matrix with randomly positioned Dirac
delta fundions scaled by a,, and H(-r$. If the scatterers have equal echogenicity a,, P(5y)
can be modeled as a Poisson distributed 2dimensional matrix modeling the random
number of scatterers per pixel in the sample volume. The distribution used is dependent on
the number of scatterers per pixel. For a k e d hematocrit kvel of 10 96, the scatterer
number per pixel is generaily l e s than 12 (for scatterers greater than approximately 4 Pm
in length o r diameter), which can be adequately modeled by the Poisson distribution. The
number of scatterers per pixel is determined by the hematocrit level simulated. This matrix
implies that two or more scatterers may overlap on top of one another when the 3-
dimensional matrix is projected dong the z axis and is collapsed into a 2-dimensional
version in Eq. 2.6 above. Note that even though the scatterers may overlap, the
contribution of each of them is considered in the model. Moreover, note that H(-z,,) is a
funaion representing a transducer parameter that is considered constant within the
thickness of the beam. Thus this parameter is ignored in the computation of the signal
power since it is the relative change in signal power due to the variations in the scatterer
size and structure that is of interest and not the absolute power.
Step 2c:
Having created the scatterer position matrix P(x;y), the tissue impedanc~ lunaion
Z(i,y) in Eq. 2.13 can be obtained by computing the product of the Fast Fourier
Transfomi (FFI') of both S(*;y) and P(l;y) matrices, and then computing the inverse FET
(IFFI") of the product. In order to minimize computation, the IFFT was not computed,
and the tissue impedance, Z(X,n was ieft in the frequency domain.
Step 3:
a Having created - f f (x ,y) in step 1 aird Z(X,,Y) in step 2, the 2dimensioaal RF
av
image RF'(i,y) in Eg. 2.6 can be obtained by taking the FFI' of %H(x, y) , by au
multiplying the result with Z(X,,E3. and by computing the WFi' of the produa. The B-
mode image is esentially the envelope of the RFtD(i,y) image, which can be derived by
taking the absolute value of the Hilben transfonn of RFm(*;y). In order to reduce
computatioa, the fast IFFI' was dso not computed when evduating the b a c h t t e c e d
power, as specified below.
Step 4:
The BSC defined in Eq. 2.4 is essentially the backscattered power, which can be
wmputed in the Etequency domain from the RF image obtained in step 3. The power
spectrum of the RF image, POW(X,v, is computed by:
where X, Y are the 2-dimensional ftequency variables, and N is the number of sample dong
each side of the image. The average power of the RF image in dB was computed as
follows:
where k is the frequency sample.
Aithough the scatterers are aiways moving in vivo, such motion was not directly
sirnulated as the motion of the Eluid was reflected in the shape, size and orientation of the
scatterers. Thus simulating a snap shot of a "fiozen image" of the scatterers is consistent
with reality . Simulating several snaps ho& for averaging purposes may reflect position
changes of the scatterers as a fuaction of time, althwgh no specific time-variation pattern
was sirnulated in the present study. Table 2 2 lis& the mode1 specifications used for the
simulations:
1 Transducer frtquency (9: 1 O M H z 1 Ultrax>und velocity (c): 1570 m/s I
Table 2.3 The parameters used in the simulation for the present study.
Transducer beam width (21pJ: 0.5 mm
Length of the PSF (29,): 652 pm
Transducer beam thickness (21pJ: 0.5 mm
2.4 Methodology
Size of the ROI: 1.28 mm x 1.28 mm
Number of pixels in the ROI: 512 x 512
image resolution: 2 5 p m
This section describes the steps taken to accomplish the objectives stated in
Chapter 1. Since RBC aggregates usually exist in small rouleaux a s in normal human
blood, or in the form of rouleau networks as in hyper-aggcegating blood samples, it would
be adequate to simulate the former in the shape of a cylinder, and the latter in the form of
a spherical clump. This approach is especially valid s i n e the spatial resolution of a 10
MHz system is not high enough to differentiate one rouleau from another in a clump of
aggregate. An approximate diameter of 7 Pm and a thickness of 2 pm was used to
represent the physical dimension of an RBC in the simulation. The pixel size into which
the scatterers were positioned according to the Poisson distribution ninction had a volume
of 3125 (25 pm x 2 5 p m x 0.5 mm).
For al1 simulations, the center Eequency of the PSF remained fixed at 10 Mm and
the hematocrit level stayed constant at 10 % so that the influence of the packing factor in
Eq. 2.2 could be negleded. This is mainly because scatterers with di&rent shape and size
have different packing factors, which may not necessatily be known. The use of a low
hematocrit such as 10 % in al1 simulations is to minimize the effects of the paclring factor
on the backscattered power. Although the shape and size of scatterers simulated in this
study do not reflea the actual physiological condition at 10 % hematocrit, the present
study represents a first step in approximating the bacb t te red power from scatterers of
different shape and size, and also in examining the potential causes of the "black holen
phenornenon. Note that a constant hematocrit also implies a variable number of scatterers
per pixel as the scatterer volumes changed. Amrding to Eq. 2.14, it is the position of the
scatterers not their entire volume that is considered in each pixel. By using a Poisson
distribution (step 2b), the variance in the number of scatterers per pixel is equal to its
mean number, which reflects the low hematocrit level simulated. Fig. 2.3 shows a drawing
of an RBC ctump and rouleaux:
Fig. 2 3 A drawing of a small RBC clump and a few rouleaux.
2.4.1 Verifidon of the Vaiidity of the Mode1
Before any theoretid model can be used for meaninml investigations, its vdidity
must first be veri fied. The following two subsections present the methods to examine two
different aspects of the system-based model. The results will be shown in Chapter 3.
2.4.1.1 Isotropy / hksotropy Due to the Structure of the Scatterers
Lntrasonic baclcrcattered power fkom scatterers with an asymmetnc structure has
been shown in the past to be angular dependent. where the angle is between the ultrasonic
beam and the longitudinal axis of the scatterers. Some examples of such observation can
be found in experiments conducted with myocardial tissue [31,41], bovine liver [q, human
Achilles tendon [25] and rend parenchyma [26]. AU of the results showed a maximum
ultrasonic bacbatter when the direction of the ultrasonic beam was perpendicular to the
longitudinal axis of the scatterers, and a minimum when the beam direction was parallel to
the axis of the scatterers.
Such anisotropic behavior was also observed in flow experiments conducted with
porcine whole blood, as well as with carbon fibers suspended in a saline-glycerol solution
[2]. Each carbon fiber was approximately 7 pm in diameter and 250 Pm in length,
mimicking a long rouleau of RBCs. However, aaisotropy was not observed for porcine
RBCs suspended in a saline solution 121, where the scatterers existed in the form of
individual RBCs not forming aggregates. Therefore, a proper theoretical model should
exhibit the following characteristics:
1. The power should be angular dependent when the structure of the scatterers is
anisotropic and large ewugh in cornparison with the wavelength. Such an anisotropy
may exist only if the longitudiaai axis of each scatterer is aligned approximately parallel
to one another.
2. The power should be angular independent when the structure of the scatterers is
isotropic, independent of their size.
In order to verify these characteristics, the signai power fkom two different sets of
tissue matrices repiesenting two different types of scatterer structures was compared: one
was a group of isotropic scatterers; the other was anisotropic. Thirty tissue matrices for
each set were generated for the purpose of establishing the error margin. AU of them had
the same scatterer size and hematodt Ievel; the only difference king the location of the
scatterers. The power variation was also examined at diffèrent insonification angles
ranging fiom 0" to 90" at 10' increments. This was accomplished by convoluting each of
the scatterer matrices with the PSF rotated across these angles. Note that at O", the
direction of propagation is parailel to the long axis of the rouleaux.
The scatterers in each set of the tissue matrices were generated with two-
dimensional Gaussian functions (Eq. 2.21), one with identical standard deviations to mimic
the shape of a spherical clump, and the other with the standard deviation in one direction
greater than the other to mimic a rouleau, More specifically, the width and depth of each
rouleau which is determined by a, and uz in Eq. 2.21 were set to 7 Pm (i.e. a, and O= were
both set to 3.5 pm), while the length of each rouleau, which is determined by a, was set to
60 pm (Le. a- was set to 30 pm). This was to simulate a group of long rouleaux each
consisting of 30 RBCs stacked together in the form of a rod.
For the case of RBC clump mimic, since a sphere is used to mimic such type of
aggregates, a, a, and 4 were al1 set to 30 pm, thus rnirnicking a group of RBC clumps
each with a 60 pm diameter. Both scatterer types were simulatecl at such dimension so
that anisotropy can be better observed at 10 MHz. Figs. 2.4 to 2.7 show the PSF, one of
the 30 gray scale images for each type of scatterer, as well as the correspondhg RF and
B-mode images. The results of the comparison between these two types of scatterers are
presented in Section 3.1.1.
Trarieducer PSF
Fig. 2.4 The PSF of the tmnsducer used in the pnsent study (Eq. 2.19).
Fig. 2.5 Samples of tissue images at an angle of zero degree (Eqs. 2.13 and 2.21).
RF iman. of dmp rninûc RF imga of mc rodeau mimic
Fig. 2.6 RF images of RBC clump and rouleau mimics at an angle of zero degree (Eq.
2.6).
Fig. 2.7 B-mode images of RBC clump and rouleau mimics (Hilbert transformation of Eq.
2.6) at an angle of zero degree.
2.4.1.2 Power W. Scattlrer Volume Ueïatlonship
It was mentioned earlier that for a fixed low hematocrit level at 10 %, the influence
of the packing factor Win Eq. 2.2 could be neglected. Substituting Eq. 2.1 into Eq. 2.2,
one gets BSC to be l i d y proportional to the scatterer volume V,. Thus a linear
relationship is expected between the signal power and V, for Rayleigh scatterers if the
system-based model is valid.
In order to verify this characteristic, the signal power fmm 10 difirent sets of
tissue matrices was computed with each set representing a different volume. Each set
consisted of 30 tissue matrices with identical scatterer volume but different spatial
distribution so that an error margin could be established. As in Section 2.4.1.1, this was
done for both scatteirer types, Le. each of the rouleau and clump mimic had those 10 sets
of volumes. Both scatterer types were generated with 24mensional Gaussian fundions as
in Eq. 2.21.
With the carrier frequency at 10 MHz and a sound velocity in blood at 1570 m/s,
the corresponding wavelength is 157 Pm. Since Rayleigh scattering occurs at sizes
approximately l e s than one tenth of the wavelength [49], the dimension of the scatterers
should be less than 15.7 pm- For the rouleau mimic, both a, and a, in Eq. 2.21 remained
at 3.5 Pm, and the change in volume was reflected in the length of the rouleau, determined
by a,. The length of the rouleau mimic was preset to range £kom 4 pm to 13 Pm, at 1 (rm
increment. This corresponded to O- varying fkom 2 Pm to 6 5 Fm, and an approximate
range of scatterer volume Ekom 154 to 500 respectively. The diameter of the
clump mimic was also preset to the same range with the corresponding scatterer volume
ranging from 34 to 1150 approximately. The increase in scanerer volume for
the clump mimic was reflected in the increase of am a, and o, which were identical to one
another at al1 times. The results of the power-volume relationship is presented in Section
3.1.2.
2.4.2 The Effects d the Scatterer Structure and Size on the Signal Power
In this section, the method is described to examine the influence of the scatterer
structure and size on the signal power. Several factors affect the backscattered power,
among them are the scatterer size, number, shape and their orientation. Physiologically, ail
these parameters cannot be easily controlled either in the human circulation or in an
experimental setting. With the present model, the influence of these parameters could be
determined.
The backscattered power for 30 different scatterer volumes was wmpared for both
RBC rouleau and clump mimics. Referring to Eq 2.21, both a, and a, remained at 3.5 Pm
for the rouleau mimic, and the change in volume was reflected in a- which raaged nom 2
um to 60 Pm, mrresponding to 2 RBCs t o 60 RBCs stacked together in the shape of a
rod. For the RBC clump mimic, a, a, and a= ail had identical values, and they ranged
from 2 Pm to 60 Pm as well. This corresponded approximately to a range between l e s
than 1 to 11,755 RBCs in a clump, since the number of RBCs was approximated by
dividing the volume of the clump by that of a human RBC which is around 77 pn3.
The simulated range was sdely for the purpose of examining the power at that
size. The range does not imply their physical existence physiologically. Such a range was
also selected to observe the power in the non-Rayleigh scattering situation. The denvation
of this model made an implicit use of the Born approximation [35], which is valid when
the medium is weakly scattering and the scatterer size is rnuch smaller than the
wavelength. However, agreements have been found to exist between experiments and
theory using the Born approximation even if the scatterers were of the size on the order of
the wavelength 1271. Thus the model should still be valid up to the dimension mentioned
above, which is close to the size of the wavelength but no longer in the realm of Rayleigh
scattering.
As before, 30 tissue matrices were generated for each volume of each scattenr
type in order to establish the error margin. Al1 tissue matrices were iasoni6ed h m O" to
90" at 22.5" increment so that the effect of the scatterer structure on the signal power
could be observecl. The results are presemted in Section 3.2.
2.4.3 The Potential Causes O€ the "Black Ede" Phenornenon
information on the influence of the scatterer size and structure on the
corresponding backscattered power enables us to iwestigate the possible composition of
scatterer structures and s i s that produce the "black hole" phenomeoon. Based on the
results in Section 3.2, five separate possible cases are pmposed. The scatterers are either
made up of strictly RBC rouleau mimic or clump mimic, or a combination between the
two. These p r o p a l s and their graphical representatioos are presented in Section 3.3.
Several models have been proposed in the past to predict the ultrasound
backscattering €rom blood, but most of them dedt with the Rayleigh scattering condition,
and did not prwide the freedom to specify the shape, the size, the orientation a s well as
the number of scatterers in the insonified region. A system-based model, which is fkee
fiom the abwe mentioned limitatioas was presented in this chapter. The implementation of
this model on computer as well as the essential parametes used in the simulation were
also described. The model was used to examine the behavior of the backscattered power
as the scatterers Vary in their size and structure, and aiso to iwestigate the potential
causes of the '%lack hole" phenornenon observed on B-mode images.
Chapter 3
Results
In this chapter, the results on the verification of the validity of the model is
presented fist. This is followed by the resule on the effects of the scattenr shape and size
toward the backscattered power. Proposais on possible scatterer structure and size that
may cause the Wack hole" phenornenon are presented in the third section.
3.1 Verifkation of the VaISdity of the System-Based Mode1
The following sections present simulation results for the venfication of the model,
as demibed in Section 2.4.1. Two aspects of the model were examined. Aithough there
may be other aspects of the mode1 that can be tested, these verificatioos are adequate for
the objectives of the present study. Further justification of the model can be found in [58],
where the model was validated agaimt experimental results and results based on the hybnd
t heoretical approach for bot h Rayleigh as well as non-Ray lei& scatterers.
3.1.1 Isotropy 1 Anisotropy Due to the Structure O€ the Scatterers
Fig. 3.1 compares the mean power computed from 30 tissue matrices mimicking
RBC clumps and rouleaux.
RBC clump mimic A
RBC rouleau mimic
O 20 40 60 80 1 O0
Angle (degree)
Fig. 3.1 Cornparison of the mean backscattered power between simulations of RBC
clumps (60 Fm in diameter) and rouleaux (60 p in length). The error bars for each
angle represent one standard deviation obtained from 30 tissue matrices.
From this graph, it is apparent that the difference in the scatterer shape is refleded in the
power with this model. A change of approximately 25 dB in the mean power for RBC
rouleau rnimic was observed as a hinction of the angle. Virtually no change in the mean
power was found for the RBC clumps. This coofimis the consistency of the model with
experimentai obsewations reported in the literature as describcd in Section 2.4.1.1.
3.1.2 Power vs. Scatterer Vdume Relatioaship
Under the conditions described in Section 2.4.1.2, a linear relationship is expected
between the signal power and the volume of the scatterers, V, if the system-based mode1
is valid. The figure below shows the power versus volume nlatiomhip for the RBC clump
rnimic. As shown, a satisfactory linear relationship (8 = 0.99) is obtained for the scatterer
volumes considered.
1 Computed power l# - Fitted line
O 2000 4000 6000 8000 10000
Scatterer volume
Fig. 3.2 Linear power-volume relationship for isotropie spherical scatterers. The error bars
for each scatterer size represent one standard deviation obtained from 30 tissue
matrices. The comlation coefficient (2) was fitted onto the mean baclrscattered
power values.
Fig. 3.3 shows the power-volume relatiooship for anisotropic cylindrical scatterers:
100 150 200 250 300 350 400 450 500 550
Scatterer volume @m3) Fig. 3.3 Linear power-volume relationship for anisotropic cylindrical scatterers. The error
bars for each scatterer size represent one standard deviation obtained from 30 tissue
matrices. The currelation coefficient (8) was fitted onto the mean backscattered
power values.
The linear relatiomhip (8 = 0.97) between the power and the scatterer volume is obvious,
thus showing the wnsistency of the model.
3.1 J Summary
Two aspects of the mode1 have been examined. The first is the change in signal
power when the target scatterers have different structures, and the second is the
relationship between the signal power and the scatterer volume. This section has
demonstrated t hat the mode1 is consistent with experimental observations repoaed in the
literature and the results expected fiom the Rayleigh scattering theory.
3.2 The Effects of the Scatterer Size and Structure on the
Power
Figs. 3.4a and 3.4b compare the power obtained for both RBC clump and rouleau
mimics for a range of sizes.
240
O 20 40 60 80 100 120 140
Diameter of RBC clumps (pm) Fig. 3.4a Power vs. the diameter of RBC clumps. The data are for 5 different
insonification angles (O0, 22.S0, 4S0, 67.5" and 903. Since no angular dependence is
observed for these angles, al1 data were pooled together and plotted as one curve.
The error bars for each scatterer size represent one standard deviation obtaiued h m
30 tissue matrices.
O 20 40 60 80 100 120 140
Length of RBC rouleaux (pm) Fig. 3.4b Power vs. RBC rouleau Iength at 5 different iosonification angles. The enor bars
for each scatterer size represent one standard deviation obtained from 30 tissue
matrices.
From Figs. 3.4a and 3.4b. one c m see that the signal power increases up to a peak
as the scatterer volume increases, but any funher increase in the diameter or length of the
scatterers results in a decrease in the signal power, except when the scatterer structure is
in the fom of a rouleau at the iosonification angle of 90". In such a case, the s igd power
continues to iacrease with the scatterer kngth. The position of the peak is also observed
to change depending on the iasoaification angle, and the differenas across the
insonification angles appear to be enhanœd for longer rouleaux. Note that the number of
RBCs in a clump is greater than that in a rouleau. Considering the small standard
deviations obtained in Ag. 3.4% it can be concluded that sphenfal clumps produa
isotropic scattering as expected.
Fig. 3.5 compares the power €rom both structures in tenns of the number of RBCs
per aggregate:
240
- RBC clump mimic 22.5'
- RBC rouleau mirnic o0
O 10 20 30 40 50 60 70 80
Number of RBCs per aggregate Fig. 3.5 Power vs. the number of RBCs pex aggregate across 5 different iosonification
angles. For the RBC clump mimic, data were pooled for angles of O", 22.9, 4S0,
675" and 90". The enor bars represent one standard deviation obtained nom 30
tissue matrices.
From Fig. 3.5, it is apparent that when the number of RBCs per aggregate is the
same, the structure of the aggregate has a significant impact on the bacbt te red power.
For every number of RBCs, spherical ciumps produce larger bacbt tered power than
rouleaux. As the number of RBCs increasa, the signai power may change depending on
the structure of the aggregates formed and the insonification angle.
3 3 The Wack Hole" Phenornenon: Potential Causes
Based on the results obtained in Section 3.2, this section postulates the possible
composition of scatterer structures and sizes that may produce the Wack hole"
phenornenon. More specifically, the scatterer structure and size that lead to a Wack hole"
were investigated, based on the results obtained in Figs. 3.4a and 3.4b. Five different cases
are presented. Case la is based on the proposal by Mo et a l and Shehada et al. [40,48],
the experimental observatiois by Qin et aL 143,441 are also incorporated in this case. Case
3a is based solely on the proposal by Mo et aL and Shehada et al. 140,481, while the
remaining cases are new proposais. Note that for ail figures in this section, R and -R
represent the vessel walls, and the "0" on the abscissa is the center of the vessel. The error
bars at each radial position represent one standard deviation obtained fiom 30 tissue
matrices.
Case la:
The "black hole" is the result of the presence of small aggregates of RBCs at the
center of the vessel, as proposed by Mo et aL and Shehada et al. [4û,48]. The condition
proposed is that the shear rate at the center of the vessel is unfavorable for large aggregate
formation. As the shear rate increases radially towards the vesse1 wdl, a region between
the tube center and the wall exists such that the shear rate promotes the aggregation of
RBCs. This results in a hyperechogenic ring, making the region at the tube center a "black
hole" in contrast. Closer to the wall, it was suggested that the higher shear rates produced
disaggregation of RBQ. The size distribution of rouleaux shown in Fig. 3.6 was used in
the simulation study to represent the above proposed condition. Fig. 3.7 shows the
correspondhg simulated 'Wack hole" region for different insonification angles:
M e Caltu of ïùbe d the tube d
Fig. 3.6 An illustration of the size distribution of RBC rouleaux across the tube.
Insonificaüon angle: 00 Insonificrition angle: 22.50
-R O R -R O R
Spatial location across the tube Spatial location across the tube
Fig. 3.7 Simulated "black hole" for Case la.
InsonHfcadlon angle: & hroniiication angle: 6ï.s"
Spatial location across the tube Spatial location across the tube
Spatial location across the tube
Fig. 3.7 (cont.)
Each rouleau in Fig. 3.6 represents the mean rodeau size in a regwn within the tube. As
shown in Fig. 3.7, a Wack hole" is observed at the center of the vesse1 for every angle
tested. The magnitude of the "black hole" is affected by the iiwnification angle and
increases as the ultnisound beam is oriented more perpendicuiar with the axis of the
vessel.
Case lb:
It is interesting to show that if the region promoting the aggregation of RBCs
results in longer rouleaux (e-g. greater than 10 RBCs per aggregate), a slightly dithxent
backscattered power distribution across the tube may be obtained. It is shown below that a
hypo-echogenic ring sandwiched between two hyper-echogenic rings around the "black
hole" cm occur at low insoaification angies. Figs. 3.8 and 39 show the rouleau
arrangement dong with the backscattered power distribution that corresponds to this
simulation:
Tube wall
C e of the tube
Tube Wall
Fig. 3.8 An illustration of the size distribution of RBC rouleaux across the tube for Case
lb.
Insoniffcation angle: O0
-R O R Spatial location across the tube
Insonification angle: 450
-R O R Spatial location across the tube
InsonMcation angle: 22.50
Spatial location across the tube
-R O R Spatial location across the tube
Insonification angle: 90°
-R O R Spatial location across the tube
Fig. 3.9 Simulated "black hole" with the hypo-echogenic ring, appearing at low angles (OoY
22Soy 45") for Case lb.
Refemng to Fig. 3.9, the magnitude of the %la& hole" was obsenred to increase with the
iasonification angle, whereas the magnitude of the hypwchogenic ring decreased as the
angle increased. The two peaks amund the 'Wack hole" were greater in power than that of
Case 1% reûecting the incceased rouleau sizes in that region.
Case 2:
The '%la& hole" may a h be due to the presence of a group of nlatively long
RBC rouleaux at the center of the vessel. Refemng to Fig. 3.4b, the power is observed to
increase up to a peak and then decreases as the length of RBC rouleaux continues to
increase for angles below 90". Thus the "black hole" phenomewn could be the result of
having a group of relatîvely long rouleaux at the tube center and smaller rouleaux towards
the wall. The figure below illustrates the aggregate structure and size arrangement that
corresponâs to t his case:
Tube waii
Ce* of tbe tube
Tube wan
#
Fig. 3.10 An illustration of the size distribution of RBC rouleaux across the tube for Case
2.
Fig. 3.11 shows the simulated "black hole" region with such a rouleau arrangement:
Insonification angle: 0'
-R O R Spatial location across the tube
Insonificiition angk: 4 5 O
InsonHication angk: 22.50
-R O R Spatial location across the tube
Insonifïcation angk: 67.s0
-R O R Spatial location across the tube
-R O R Spatial location across the tube
Insonification angle: 90'
-R O R Spatial location across the tube
Fig. 3.11 Simulated "black hole" for Case 2.
The Wack hole7' does appear but seems to be more prominent at lower angles (iess than
454. Its magnitude is observed to de- with the iasonification angle and disappears
completely at higher iasonification angles, which is watrary to what was observed h
Cases la and lb.
Case 3a:
Another passible situation muld aise following the same logic as in Case 1, but
having rouleau networks as aggregates rather than just plain rouleaux. Such a condition
could be possible especially for the case of hyper-aggregating RBCs at extremely low flow
rates. The figure below illustrates the aggregate stnidure and size arrangement used in the
simulation:
Ce* of the tube
I
Fig. 3.12 An illustration of the size distribution of RBC rouleaux across the tube for Case
3a.
Due to the fact that the scatterers are isotropic, the simulaîed 'Wack hole" is angular
independent. Fig. 3.13 shows the simulateci "Mack hole" region foc such isotropic
spherical scatterers. The simulated "black hole" has a magnitude of approximately 3 dB
for d l insonification angles.
-R O R
Spatial location across the tube
Fig. 3.13 Simulated Wack hole" for Case 3a.
Case 36:
As in Case 1, such scatterer arrangement may also lead to the presence of a hypo-
echogenic ring around the '"black hole" region. However, the likelihood of such
occurrence is unlaiown as the scatterers in the region where RBC aggregation exists must
be relatively huge (e.g. an RBC clump with a 40 Pm diameter) in order for such
phenornenon to appear. The figure below shows the aggregate structure and size
arrangement that was used to generate the "black hole" and the hypo-echogenic ring
illustrated in Fig. 3.15. For dl insonification angles, the magnitude of the "black hole" as
well as the hypo-echogenic ring were approximately 2 dB. Due to the size of the RBC
clumps, only those scatterers in the lefi half of the tube are shown.
Cmtm of me the tube W1P
Fig. 3.14 An illustration of the size distribution of RBC rouleaux across the tube for Case
3b.
-R O R Spatial location across the tube
Fig. 3.15 Simulated "black hole" with the hypoechogenic ring for Case 3b.
Case 4:
In Fig. 3.44 we see that when large clumps decrease in volume, the backscattered
power reaches a peak and then declines. Thus moving ftom the center of the tube, such
break-up would result in an increase in signal power uatil a peak is reached, and as the
shear rate continues to increlse at positions closer to the wall, the clurnps get broken into
smaller pieces and the signal power eventually declines. At a position close to the wall, the
aggregates could either exist in the f o m of small clumps or small rouleaux. This situation
results in a rapid àecline in signal power at mgions closer to the vesse1 wall. Fig. 3.16
shows the simulated scatterer arrangement for Case 4:
i Nmbu of6LBCs (rppra-1
r L+=&=-of~wfPfi-(Crm)
Fig. 3.16 An illustration of the size distribution of RBC aggregates across the tube for
Case 4.
Due to the €ad that RBC aggregates in the form of clumps are isotropic, the magnitude of
the "black hole" in Fig. 3.17 is iodependent of the insonification angle. The simulated
Wack hole" has a magnitude of appmximately 4 dB.
Case 5:
Spatial location across the tube
Fin. 3.17 Simulated Wack hole" for Case 4.
This case was derived strictly bom the mode[, the chances of its existence under
physiological conditions are unhiown. This final alternative has large clumps of RBC
aggregates at the center of the tube, and as the shear rate increases radially towards the
wall, the clumps get broken into small RBC rouleaux forming the hyperechogenic ring. As
the shear rate continues to increase towards the wall, the rouleaux get disrupted by the
high shear rate and the signal power eventually declines. Fig. 3.18 is an illustration of the
aggregate structure and size arrangement that produces the "black hoIe7' in Fig. 3.19:
Tube tbe tube
Fig. 3.18 A graphical illustration of the scatterer structure and mean size arrangement for
Case 5.
As shown in Fig. 3.19, the "black hole" is observed at the center of the vesse1 for
every angle tested. The magnitude of the "black hole" is also observed to be influenced by
the insonifkation angle. It increases in magnitude fkom approximately 2 dB to over 5 dB
as the angle approaches 90". Note that such observation is very similar to that of Case la,
except that the scatterer arrangement is totally different.
Relative power (dB)
Relative power (dB)
2 2 2 2 I U P Q ) O P O "D-
Relative power (dB)
Relative power (dB)
3 :: a
! O a PD 3 m 0 m 0 .
- - - - I
.I
C
Relative power (dB)
4 o
L I 1 1 1
In this chapter, the system-bascd mode1 was first verified, and then used to
simulate scatterers with predetemined volumes and structures. The signal power
computed fkom these scatterer-mimic was then used to estimate the passible orgaaization
of scatterers that wwld result in the "black hole" phenomenon observed experimentally by
other researchers. Five different cases that could pdentially cause such a phenomenon
were proped. These results shall be discussed in the following chapter.
Chapter 4
Discussion
This chapter consists of two main sections, the analysis of the resula in Chapter 3
will first be presented while the strengths as well as the limitations of the mode1 will be
discussed in the second section.
4.1 Analysis of the Results
In the following, the results obtained in Section 3.2 will be analyzed, and this will
be followed by a discussion on the proposed causes of the Wack hole" phenomenon.
4.1.1 Power vs. Volume Relatioaship €rom a System-Based Perspdve
Figs. 3.4a and 3.4b show that the backscattered power from a 6xed hematocrit
tissue sample increases up to a peak and then decreases as the scatterer volume continues
to increase. From a system-based perspective, such behavior can be better understood in
the frequency domain.
4.1.1.1 Anisotropic cylindrical scatterers
The convolution operation described in Eq. 2.6 is the multiplication in the
frequency domain of the Fourier transform of the PSF and the Fourier transfom of the
tissue image. Figs. 4.1 and 4.2 show these images with their comsponding magnitude
spedrum for RBC rouleau mimic at an arbitrary irisonification angle of O":
Fig. 4.1 The traosducer PSF (2nd order derivative of w)) in the spatial domain with the
corresponding magnitude spedrum (zoomed version).
Fig. 4.2 Tissue matrix mimicking 20 pm RBC rouleaux at 10 % hematocrit with the
coriesponding magnitude spectmm. Note that the DC component has been removed
nom the spectnim for better visualization. The RBC rouleaux are aligned paralle1 to
the y-axis.
Ail figures are s h o w 4 t h the origin at the center of the image. In Fig. 4.1, the
displacement of the two cinilar spots in the spedrum nom the origin reflects the carrier
frequency of the PSF. The spectrum in Fig. 4 2 is esccntiaily a matrix with 2-dimensional
delta functions weighted by the scattenr prototype fundion (Eq. 2.12).
The product of the two spectra in Fi*. 4.1 and 4.2 can be better perceiveci by
supenmposing the two spectra on top of one another. The overlapping regions between
these two spectra wiU produce a non-zero produd; anywhere else in the spectrum would
yield a nul1 product. Thus, the baclscattered power is diredy affecteci by the overlappiag
area between the PSF spectnim and the tissue image spectrum.
From the Fourier transfocm relationship between space and hequency, it is knawn
that scaling in the spatial domain leads to an inverse relationship in the kquency domain:
where f(m) is the fiinction in the spatial domain, a is the scaling factor, x is the spatial
domain variable, FT is the Fourier transform, F is the Cunction in the fiequency domain,
and o is the Erequency domain variable. Eq. 4.1 shows t hat as the rouleau size is increased
along the y-direction in the spatial domain, the tissue image spectrum will shcink along the
same direction in the fiequency domain and its amplitude wiU increase. If the rouleau size
is reduced, the ellipsoid in the spectrum of the tissue image wiU tend to be more circular,
thus increasing the overlap with the two spots in the spectrum of the PSF. As the length of
the scatterers is increased, the inccease in the amplitude of the spectrum offsets such
scaling effect in the fkequency domain and as a nsult, the backscattered power is increased
as obseived for the case of Rayleigh scattering (Fis. 3.3 and 3.4b for rouleau lengths l e s
than 13 pm).
In Fig. 3.4b, as the scatterer size continues to increase, such behavior of the
magnitude spectrum wiii result in a decrease in the overlapping region between the PSF
spectrum and the ellipsoidal tissue image spectrurn, leading to the decrease in
backscattered power. In general, one can wnclude that, as the length of rouleaux is
increaseà, the baclrscattered power depends on the i m a s e in amplitude of the tissue
spedrum, the decrease in the ovedapping region between both spectra, and the aumber of
scatterers in the region of interest. At the limit, when the scatterers are relatively large
(e.g. 120 pm) and the angle is close to zero degree, the overlapped area is so smdl that
the product between the two spectra is a d y nulk, leading to a reduction of the
bacbcattered powrer by several decibels. The number of scatterers in the region of interest
for large scatterem is also not as numems as when the scatterers are smaller, since the
hematocrit is kept constant. As a consequence, fewer scatterers wiil produce less
backscattered signai, thus reducing the b a c b t t e r e d power.
4.1.1.2 Eff'ects of the insonifkation angle for anisotropic scatterets
In the system-based model, the change in the insoaification angle was represented
by the rotation of the PSF about its origin, which is the center of the image. According to
Fig. 3.4b, increasing the angle leads to an increase in the b a c b t t e r e d power. This can
be explained by the fact that the two spots on the PSF spectmm wiU increasingly overlap
with the region oaxipied by the ellipsoidal tissue image spectrum as the angle is changed
fkom O" to 90". Note that at 90°, the power drop is not observed because the volume
change of RBC rouleaux is refleded in the length, which is dong the y-sis . In the
bequency domain, the change in the scatterer volume does not affect the overlapping
region between both spectra because the two spots of the PSF are aligned dong the x-axis
and superimposed over the tissue spedrum. Thus, the iacrease in the amplitude of the
tissue spectnim as the rouleau length is increased raises the backscattered power.
Fig. 4.3 shaws the tissue image and the corresponding magnitude spedrum for
spherical scatterers mimicking RBC clumps:
Fig. 4.3 Tissue matrix mimifang RBC clump of 20 Pm diameter and 10 % hematocrit
with the comsponding magnitude spectmm. Note that the DC component has been
removed h m the spectnim for better visualization.
Using a 3-dimemional Gaussian hindion with identical standard deviations in the
simulation of RBC clumps results in a spectmm that is circular in shape. As the clump in
the spatial domain increascs in size, the circular spectmm of the tissue image will decrease
in its diameter, following the same relatiomhip as in Eq. 4.1. The same explmation for the
behavior of the backscattered power used for the rouleau mimic applies to the clump
mimic, except that the latter is angular independent and appears to have a stronger
backscattered power compared to the former when the number of RBCs per aggregate is
the same (Fig. 3.9.
From the spectrum in Fig. 4.3, it is not difficult to see that the backscattered power
fkom such a tissue mimic is angular independent, since the isotropie stmcture of the
scatterers is refleded in the frequency domain. As the PSF is rotated across the different
iosonification angles, the symmetry of the tissue image spectrum causes the ovedapped
region to remain unchangea leading to the aagular independent backscattered power. For
a given number of RBCs per aggregate, the stronger backxatteced power for clumps c m
be explainecl by the increase in the scatterer size in dl three dimeasions rather than just
dong the y-direction as in the case of the RBC rouleau mimic.
4.1.1.4 Otber considerations
It is important to note that the position of the peaks observed in Figs. 3.4a and
3.4b is dependent on the carrier frequency of the PSF. In this study, a PSF carrier
frequency of 10 MHz was used for al1 simulations. However, the peak would be expeded
to occur at a larger scatterer size if the carrier frequency had k e n reduced and viœ versa.
Refemng to the magnitude spectrum of the PSF in Fig. 4.1, the carrier frequeacy is
reflected in the position of the two "spots" with respect to the origin. Reducing the carrier
fiequency would mean that it would take a larger scatterer size to decrease the
overlapping region mentioned above, thus resulting in the shift of the peak position. From
a physical point of view, lower carrier frequency implies greater wavelength, which means
that larger scatterers can be considered as Rayleigh scatteres, where the power is related
to the volume at a fixed hematocrit.
With regards to the effects of the system parameters toward the bacb t t e r ed
power, the bandwidth and beamwidth of the ultrasound system, which corresponds
respectively to the standard deviations and I#, in Eq. 2.17, are also of importance
because both variables affect the spectrum of the PSF. However, any power variations due
to these system parameters wiil not affect the trend of the bacbcattered power due to the
change in the scatterer parameters as observed in Figs. 3.4 and 3.5.
As shown in Eq. 2.20, 3J-dimeosional Gaussian shaped scatterers were used in this
study. This assumption was used mainly for the sake of simplicity, s i n e the Fourier
transfonn of a Gaussian fuaction is still a Gaussian function, and it can be truncated with
negligible error at the third o r fourth standard deviations. In addition, the transducer may
not be able to r e d v e the edges of the scatterers in practice, thus, a hinction that decays
smoothly should be sufficient as an approximation. One could have used other
mathematid functions such as the ~c tangular window to represent the scatterers. But the
edges of such a fundion would introduœ side lobes in the fkquency domain. This is fine if
the scatterers are large, since the PSF would overlap only with smail side lobes toward the
"tail" of the spedrum. However, if the scatterers are small, the side lobes may complicate
the intecpretation of the tesults since the PSF would overlap with large si& Lobes close to
the baseband of the spectrum. Thus, the use of mathematical fundons that may introduce
side lobes in the hequency domain needs to be considered very carefully, especially when
the scatterers are very small. The use of the Gaussian fundion can be coasidered
appropriate sinœ the simulated results agree well w ith experimental observations reported
in the literature (described in Section 24.1.1) and the theoretical prediction as presented in
Section 3.1.
4.1.1.5 Simulation of a more redistic tksue image
Eqs. 2.15 and 2.16 can be used to simulate a more realistic tissue image. For
illustration, such a sample tissue image can be created by introducing a random component
in the length of each scatterer and in its angle with respect to the tube ais (O0 represents
the long axis of the tube). Having the scatterers aligned at an average angle other than O"
will result in a change in the length and angle with respect to the tube axis when projecting
the scatterer onto the 2-dimemional plane. The simulated scatterers had a diameter of 7
pm and a length of 250 Pm. AU of them were aligned at an average angle of 4S0, with a
random angular component of r 5". A random component cm be incorporated in the
length of the scatterers a s well, but this was not done for the puipose of the present
illustration. Fig. 4.4 shows the sample tissue image with the componding magnitude
spectmm:
Fig. 4.4 A sample tissue mat& mimicting RBC rouleaux of 250 Pm long with the
corresponding magnitude spectmm. Note that the DC component has been removed
fiom the spectnim for better visualization. The RBC rouleaux are aligned at an
average angle of 4S0, with a random compownt of * 5".
The system-based mode1 is capable of producing realistic tissue matrices, but at the
expense of long coxnputational time because each scatterer must be added iodiMdually
onto the tissue matrix with its individual projected scatterer length, angle of alignment,
and position. From the magnitude spedrum in Fig. 4.4, as long as the variance in the
angular alignment of each scatterer with respect to the tube axis is not too large, the
magnitude spectnim will not be very different fkom an ideal situation (where al1 scatterers
are assumed to have the same length and orientation), in which case the spectrum would
appear as an ellipsoid (e.g. Fig. 4.2) aligned at 45". Thus the trend of the backscattered
power due to the changes in scatterer parameters for both ideal and non-ided cases should
be fairly similar, i.e. we do not expect any drastic changes in the trend of the results
presented in Figs. 3.4 and 3.5 when more realistic tissue matrices are used in the
simulation. Thus the use of ideal tissue matrices should be sufficient for the objectives of
the present study.
Fiuctuations in the angular alignment of the scatterers may occur when the Bow is
turbulent, or when their concentration is low, as observed by Goldsmith et oL 119-211. It
was observed that a low concentration of elongated particles had the tendency to rotate
and deform under shear flow, with the angular velocity k i n g dependent o n the particle
length The histogram of the number of elongated particles as a function of their
corresponding angles showed that most of them were aiigned with the BOW at a given
time. As mentioncd in Section 2.3, d l scatterers were alignai with the flow at O" without
any random angular fluctuatioas for the simulations of the present study. However, the
random scatterer le* component is not expeded to wntribute signincantly because it is
averaged out when many (e.g. 30) tissue images are used to compute the average power.
4.1.2 Most Probable Causes of the "Black Hole" Phenornenon
In order to examine the validity of each of the cases presented in Section 3.3, one
way is to compare the results to experimental observations. Using porcine whole bled,
Mo et al. [40] and Shehada et aL [48] pmposed that there should exist a location
somewhere between the tube axis and the waif that promotes RBC aggregation. As a
result, a hyperechogenic ring is formed around the tube am's, leading to the "black hole"
phenornenon. The range of shear rates used in their studies was €tom 0.001 to 5.3 s".
From the results obtained by Chien 1101 on the aggregation of normal human RBCs at 45
% hematocrit, it was observed that a range of shear rates promoting the aggregation of
RBCs does exist. Fig. 4.5 shows a plot of the aggregation index computed by Chien as a
function of the shear rate:
AGGREGAnOhJ OF NORMAL RBC (n - 46 %) IN DE%TFUN W (4g 1100 ml)
œ
-1.
Q2-
SHEAR RATE (sr&)
Fig. 4.5 Aggregation index of normal human RBCs at 45 % hematocrit as a function of
the shear rate (adapted fiom [IO]). RBCs were separated nom the plasma and were
suspended in a dextran saline solution.
One can see from Fig. 4 5 that zero shear rate does not necessarily produce the largest
aggregates. The same celationship as in Fig. 4.5 was found for porcine whole blood 1481.
Thus, as the shear rate increases from the tube a i s to the wall, the "black hole" region as
well as the hyperechogenic ring around the tube axis would correspond to the region to
the lef3 of the peak in Fig. 4.5. Closer to the wall, the shear rate is increased beyond the
peak of Fig. 4 5 and the echogenicity would be reduced, thus emphasizing the
hyperechogenic ring. In the studies by Mo et aL and Shehada et aL, spherical aggregates
were considered. The explanation given above shows that their hypothesis is lihly to be
valid. This hypothesis was simulateù in Case 3a (Fig. 3.12).
Perfonning expnments in a smaller diameter tube under sirnilar flow conditions,
Cloutier et aL [12,13] observed that the "black hole" phenomenon was not always present
with porcine whole blood. This awld be exp la id by the fact that the minimum shear rate
value present was approximatefy 2 i1 at the center of the vessel. Depending on the
variance of the shear rate at the center of the vessel, this region could be favorable or
unfavorable for the aggregation of porcine RBG. In the oivorable case, t his would lead to
the situation where the region arwnd the tube axis had the largest aggregates. As the
shear rate was i n c d toward the wall, the aggregates got broken apart and the
backscatteced power declined. Note that the size of the aggregates at the center is not
expeded to be as large as in Cases 4 or 5 (Fip. 3.16 and 3-18), as that would lead to the
presence of a %la& hole". If the shear rate was unfitvotable for the occurrence of the
largest aggregates, then the situation would be as described by Mo et aL [40] and Shehada
et al. [48].
Usiog horse blood, Qin et aL [43,44] suggested that the orientation of large,
organized rouleau structure contributes to the formation of the '8lack hole" (Section
1.3.1). The magnitude of the "black hole" was observed to inccease with the insonification
angle. With that, it was proposed that the RBC aggregates would appear as shown in Case
l a (Fig. 3.6). Fig. 4.6 shows the plot of the magnitude of the %Iack hole" as a huiction of
the insonification angle for Case la:
lnsonification angle (degree) Fig. 4.6 "Black hole" magnitude vs. insoaification angle computed eorn the simulation of
Case la. The results were expressed in terms of mean t one standard deviation and
were averaged over 30 tissue matrices.
The simulated results in Fig. 4.6 follav a similar trend to those obtained by Qin et al.
[43,44]. ln that study, the magniîude of the Wack hole" varied nom 1 to 2 dB for angles
varying fiom 4û0 to 70°. The angle was defineà between the ultrasound beam and the tube
axis. Because the RBC rouleaux were probably not aligned completely with the flow in the
experiments conducted by Qin et al , the quantitative cornparison with Fig. 4.6 is diffiailt
since the angle with respect to the long axis of rouleaux was unknown in t hat study. In the
present study, the magnitude of the "black hole", M, was computed as follows:
where Pr is the power at the pak representing the hyperechogenic ring and Pz is the
power in the valley representing the "black hole". Note that Qin et aL took the average
power of the hvo peaks in the computation of the magnitude as the values w e n d i e n n t
experimentall y because of the ultrasound attenuation by blood.
Since the magnitude of the "bIack hole" for Case 2 decreases with the angle of
insonificatioa, this case seems unlikely to be vdid because the angular dependence is
contrary to the results obtained experimenially with home blood 143,441. Due to the fact
that no experimental data is avdable with regards to the angular dependence of the "black
hole" for porcine whole blood, the scatterer structure and organization could appear as in
Case la (Fig. 3.6) or Case 3a (Fig. 3.12) or a combination of both, but the conclusion is
yet to be determiued. As mentioned eariier, the largest aggregates are expected to be at
the center of the tube in cases where the Wack hole" is not observed for porcine whole
blood. However, if hyper-aggregating RBCs is present at the center of the tube as in Cases
2,4 and 5 (Fie. 3.10,3.15 and 3.18), a "black hole" may also occur. Only Cases 1 , 2 and
5 present angular dependence in the mapitude of the "black hole", with the magnitude
change of Case 2 in the opposite direction of Cases 1 and 5 as the insonificatioa angle
i ncreases.
4.2 The Strengths and Limitations of the Model
The system-based approach is a good model in the sense that it not only provides
the fieedom to specib the shape, structure, orientation and number of scatterers three
dimensionally, but it also allows one to visualize the actual tissue to be imaged. The model
provides the freedom of insoniQing a tissue image at any angles which is not always
possible in an actual experimental or clinical setting. Although b l d motion was not
simulated in the present study, this is not beyond the capabilities of the model. Mwing
scatterers can be simulateû by creating successive "fkozen images" by shifting and
interpolating the position maaix. A motion pictuce can be generated to create a realistic
and intuitive "mwie" of the scatterers in motion. Other advantages of the model include
the ability to specify the system parametes such as the bandwidth and beamwidth, as weli
as the ability to retcieve either the RF or B-mode images for other processictg purposes.
One limitation is the assumption of wealt scatterers, making the modeling of other
particles such as ultrasonic wntrast agents rather diffiatlt. The packing factor is aaother
parameter that ne& to be incorporated into the model if the modeling of higher levels of
hematocrit is of interest. niis may be very chdlenging, especially if RBC aggregates are
considered. The modeling of a focused tramducer which operates at the near field region
may be beyond the reach of this approach as the scatterers are assumed to be in the far
field of the transducer,
Another weabiess of the model may be its enormous appetite for computational
power. However, the model can be implemented one-dimensionally with good aariracy.
That is, imtead of having a PSF and a tissue matrix, one has an impulse response and a
tissue vector on which the 3D scatterer prototype is projected. But one will have to forfeit
the freedom ofchanging the iosonification angle. Thus, unless the iosonification angle is of
no interest for al1 simulation studies, the mode1 has to be implemented in 2D. Any attempt
to include in the model scatterers as small as an RBC requires a very high spatial
resolution. Since the physical size of the sample volume camot be infïnitely small, that
translates to the use of images large enough to accommodate the sample volume size, with
sufficient number of pixels to avoid spatial aliasing of the RBCs. The computing of the
two-dimensional convolution of images with large number of pixels is the major
determinant of the computation time. This disachantage however, is system-dependent,
and will fade away when more powerful processors are available in the future.
In this chapter, the results shown in Section 3.2 were analyzed in the ûequency
domain to explain the behavior of the backscattered power with respect to the change in
scatterer size and structure. The effacts of the system parameters on the b a c b t t e r e d
power were also disaissed. An example of a more realistic tissue image was provideâ
together with the justification of using a more idealized tissue image in al1 simulation.
Conccming the "black hole" phenomeaon, Cases l a and 3a were deemed to be more Iikely
valid. Both Cases l a and 5 agreed with the angular dependence of the "black hole"
presented by Qin et oL [43,44], with Cases 4 and 5 king new proposais solely based on
the system-based model presented in this study. Chse 2 was wncluded to be uniikely to
occur based on the experimental results by Qin et aL. The strengths and limitations of the
model were discussed in the final section.
Chapter 5
Conclusion
Motivated by the potential clinical use of the ultrasound bacb t t e r ed power in
the assessment of patients' RBC aggregation level, and the dismvery of the 'Wack hole"
phenomenon by other researchers, the objectives of the present study have been the
assessment of the influence of the scatterer characteristics on the backscattered power, as
well as the potentiai causes of the Wack hole" phenomenon.
It can be concluded that, besicles the previously known dependencies of the
ultrasound backscattered power on the scatterer volume, orientation, and shape, the
measured power could be a deceptive parameter especially when the scatterers are
relatively large and are no longer Rayleigh scatteres. Using a system-based model, the
backscattered power was observed to drop beyond certain scatterer size at a carrier
frequency of 10 MHz. The power difference at different insonification angles was also
observed to increase when the scatter size is larger. Further research is needed before
ultrasound backscattered power can becorne a usehl clinical tool in the assessment of
RBC aggregation level in patients.
The proposai of Mo et al. [40] and Shehada et al. [48] as described in Case l a and
3a was tested with the system-based model, and the results of simulation were compared
to the experimeotal results obtained by Qin et al. [43,44]. It can be mncluded that the
hypotheses proposed by Qin et al. in these studies are very likely valid, as the simulation
and experimental results matched very well. The mode1 was also used to create several
other scenarios t hat could potentially lead to a "black hole". The conditions leading to the
possible occurrence of a hypo-echogenic ring around the 'Wack hole" were also
presented. Based on the results of this thesis, it can be concluded that the Wack hole"
phenomenon could have more than one potential cause. Further work will have to be
conducted both experimentdly and with the system-bd mode1 to ver@ the causes of
this phenornenon.
References
[l] R. J. Albright and J. H. Harris, "Diagrmis of urethral Bow parametes by ultnsonic
backatter," IEEE Tram Bwmed Eng, vol. BME-22, pp. 1-11, 1975.
[2] L. Allard, G. Cloutier. and L G. Durand, ''EEcd of the iasonification angle on the
Doppler backscattend power under red blood ceIl aggregation conditions9" IEEE Tram
U h o n Ferroelec Freq Cmt, vol. 43, pp. 211-219,1996.
[3] H. 1. Bicher, BIood c d aggregarrion Ui thronrbotic processes. 1st ed. Illinois: Chades
C. Thomas, 1972.
[4] M. Boynard and J. C. Lelievre, "Size determination of red blood cell aggregates
induœd by dextran using ultrasound bacbttering phenornenon," Bwrheology, vol. 27,
pp. 39046,1990.
[SI M. Boynard, J. C. Lelime, and R. Guillet, "Aggregation of red blood cells shidied by
ultrasound backscattering," Bwrhedogy, vol. 24, pp. 45 1-461, 1987.
[6] W. R. Brody and J. D. Meindl, Theoretical analysis of the CW Doppler ultrasonic
flowmeter," IEEE Tram Biomed Eng, vol. BME-21, pp. 183492,1974.
[7] T. M. Burke, E. L. Madsen, and J. A. Zagzebski, "A prelirninary study on the angular
distri but ion of scattered ultrasound fkom bovine liver and m yocardium," Ulîrasonic
Imaging, vol. 9, pp. 132-145, 1987.
[8] S. Chen, "halysis of organization dynamics of red blood cells under flow, in health
and disease, by computerized imaging", Ph.D. thesis, Hebrew University, Jerusalem, pp.
1-82, 1994.
[9] S. Chien. "Biophysical behavior of red cells in suspensions," in me red blood ceil.
D.M. Surgenor Ed. New York, San Francisco, London: Academic Press, 1975, vol. II, pp.
1032-1133.
[ I O ] S. Chien, ''Electrochemical interactions between erythrocyte surfaces," Thrombosis
Research, vol. 8, pp. 189-202, 1976.
[Il] D. A. Christensen, Ultrasonic BwUistrumentotwn. 1st ed. New York, Chichester,
Brisbane, Toronto, Siagapore: John Wiley & Sons, 1988.
[12] G. Cloutier and Z. Qin. "Shear rate dependence of nonnal, hypo-, and hyper-
aggregating erythrocytes studied with power Doppler ultrasound," in Acousticul Imuging.
S. Lees and L. A. Fenari Ed. New York: Plenum Publishing Corporation, 1997, vol. 23,
pp. 291-296.
[13] G. Cloutier, 2. Qin, L. G. Durand, and B. G. Teh, "Power Doppler ultrasound
evaluation of the shear rate and shear stress dependences of red blood cell aggregation,"
IEEE Trans Bwmed Eng, vol. 43, pp. 441450,1996.
1141 G. Cloutier, X. Weng, G. 0. Roederer, L. Allard, F. Tardif, and R. Beaulieu,
''Differences in the erythrocyte aggregation level between veins and arteries of
normolipidemic and hyperlipidemic individuals," Ultrasound Med Bwf, vol. 23, pp. 1383-
1393,1997.
[IS] A. L. Copley, R. G. King, and C. R. Huang. '?3ythmcyte sedimentation of human
blood at varying shear rates," in Microcircularion. J. Grayson and W. Zingg Eds. New
York: Plenum Press, 1976, vol. 1, pp- 133-134.
[16] H. Demiroglu, I. Barista, and S. Diindar, "Erythrocyte aggregability in patients with
coronary heart disease," Cfinicul Hemorheology, vol. 16, pp. 313-317, 1996.
[17] M. Domer, M. Siadat, and J. F. Stoltz, ''Erythrocyte aggregation: Approach by light
scattering determination," Bwrheology, vol. 25, pp. 367-375, 1988.
[18] D. H. Evans, W. N. McDicken, R. Skidmore, and J. P. Woodcock, Doppkr
Ultrasowtd. Physics, Instrumentation, and Cfinicol Applicatwns. 1 st ed. Chic hester,
New York, Brisbane, Toronto, Singapore: John Wiley & Sons, 1989.
[19] H. L Goldsmith, "The microcirculatory society Eugene M. Landis award lecture. The
microrheology of human blood," Microvascuhr Res, vol. 31, pp. 121-142, 1986.
[20] H. L. Goldsmith and J. Marlow, "Flow behavior of erythrocytes. I. Rotation and
deformation in dilute suspensions," Proc Royal Soc (London), vol. 182, pp. 351-384,
1972.
[21] H. L. Goldsmith and S. G. Mason. "Some model experiments in hemodynamia. 111,"
in Hemorheology. Proceedings of the first hintnational conference. A.L. Copley Ed.
New York: Pergamon Press, 1966, pp. 237-254.
[22] J. F. Greenleaf, Tbsue characteriurtion wdh ultrasound. 1st ed. Boca Raton,
Fiorida: CRC Press, 1986.
1231 R. Hahn, P. M. Müller-Seydlitz, K. H. Jtkkel, H. Hubert, and P. Heimburg,
"Viscoelasticity and n d blood ce11 aggregation in patients with foronary heart disease,"
Angiofogy, vol. 40, pp. 914-920,1989.
[24] H. Hammi, P. Penotin, R. Guillet, and M. Boynard, ''Determination of red blood a U
aggregation in young and eldedy subjects evaiuated by ultrasound: Influence of
dih ydroergocryptine mesylate," C h i c a l ffemorheology, vol. 14, pp. 1 17.126, 1994.
[25] B. K Hofnneister, A. K. Wwg, E. D. Verdonk, S. k Wickiine, and J. G. Miller,
"Anisotropy of ultrasonic backscatter from human tendon compared to that fiom normal
human myocardium," Ultrasmicr Symposimz, vol- 2, pp. 1127-1 131, 199 1.
[26] M. F. Insana, T. J. Hall, and J. L Fishback, 'Identifying acoustic scattering sources in
normal noal parenchyma €tom the anisotmpy in acoustic properties," Ultrasuund Med
Bwl, vol. 17, pp. 613-626, 1991.
[27l M. F. Insana, R. F. Wagner, D. O. Brown, and T. J. Hall, 'Pesnibing small-scale
structure in random media using pulse-echo ultrasound," J Acowt Soc Am, vol. 87, pp.
179-192,1990.
[28] M. M. Khan, R. R. Puniyani, N. G. Huilgol, M. A. Hussain, and G. G. Ranade,
''Hemorheo1ogical profila in cancer patients," Clhical Hemorheology, vol. 15, pp. 37-44,
1995.
[29] S. Y. Kim, 1. F. Miller, B. Sigel, P. M. Coosigny, and J. Justin, "Ultrasonic evaluation
of eryt hrocyte aggregation d ynamics," Biorheology , vol. 26, pp. 723-736, 1989.
[30] C. Le Devehat, M. Vimeux, G. Bondoux, and T. Khodabandehlou, "Red blood cell
aggregation in diabetes mellitus," Int Angiol, vol. 9, pp. 11-15,1990.
[31] E. 1. Madaras, J. Perez, B. E. Sobel, J. G. Mottley, and J. G. Miller, "Anisotropy of
the ultrasonic backatter of myocardial tissue: II. Measurement in vivo." J Acoust Soc
Am, vol. 83, pp. 7620769,1988.
[32] N. Maeda, K. Imaizumi, M. Sekiya, and T. Shiga, "Rhedogical characteristics of
desialylated erythrocytes in relation to fibrinogen-induced aggregation," Biochimica &
Bwphysica Acta, vol. 776, pp. 151-158, 1984.
[33] W. N. McDicken, Diagnostic uftrasonLcs: Prihciples and use of Urstruments. 3rd ed.
Edinburgh, Lnndoa, Melbourne & New York: Churchill Livingstone, 1991.
(343 G. Mchedlishvili, R. Shalrarishvili, M. Aioeva, and N. Momtselidze, "Elaborated - "Georgian index" of erythrocyte aggregability characterizing the microrhedogical
disorders associated wit h brain infimt,'' C h i a L Hemrheology, vol. 15, pp. 783-793,
1995.
[35] J. Meunier and M. Bertrand, "Echographic image mean gray level changes with tissue
dynamia: A system-based mode1 study," IEEE Tram Biomed Eng, vol. 42, pp. 403-410,
1995.
[36] J. Meunier and M. Bertrand, "Ultrasonic texture motion analysis: Theory and
simulation," i E E Trans Med ImagUrg, vol. 14, pp. 293-300, 1995.
[37J B. Miller and L. Heilmam, "Hemorheological parameters in patients with
gynecologic malignancies," Gynecologic O n c o l . , vol. 33, pp. 177-181, 1989.
1381 L. Y. L. Mo and R. S. C. Cobbold, "A stochastic mode1 of the backscattered Doppler
ultrasound from blood," IEEE Tram Bwmed Eng, vol. BME-33, pp. 20-27, 1986.
[39] L. Y. L. Mo and R. S. C. Cobbold, "A unified approach to modeling the
backscattered Doppler ultrasound îkom blood," IEEE Trans Biomed Eng, vol. 39, pp.
450461,1992.
[40] L. Y. L. Mo, G. Yip, R S. C. Cobbold, C. Gutt, M. Joy, G. Santyr, and K. K. Shung,
cWon-newtooian behavior of whole blood in a large diameter tube." Bwrheology, vol. 28,
pp. 421-427,1991.
1411 J. G. Mottley and J. G. Miller, "Anisotropy of the ultrasonic backatter of
m yocardial tissue: 1. Theory and measurements in vitro." J Acoccst Soc Am, vol. 83, pp.
755-761,1988-
[42] F. J. Neumann, H. A. Katus, E. Hoberg, P. Roebmck, M. Braun, H. M. Haupt, H.
Tillmanns, and W. Kübler, "Increased plasma viscosity and erythrocyte aggregation:
lndicators of an unfavourable clinid outcome in patients with unstable angina pectons,"
Br Heurt J, vol. 66, pp. 425430,199 1.
[43] Z. Qin, "Characterization of red blood ce11 aggregation dynamics with Doppler
ultrasound", M.Sc. thesis, Université de Montréal, Institut de recherches cliniques de
Montréal, Montreal, pp. 1-130, 1997.
[44] 2. Qin, L. G. Durand, aad G. Cloutier, "Kinetics of the "black hole" phenornenon in
ultrasound backscattenng measuremenis with red blood ce11 aggregation," Ultrasuund
Med Biol, vol. 24, pp. 245-256, 1998.
[45] J. W. S. Rayleigh 'Vibrations of solid bodies," in 'Ineory of s d . New York:
Dover Publications, 1945, vol. 2, pp. 415431.
[46] S. M. Razavian, M. Del Pino, A. Simon, and J. Levcrison, Yncrease in erythrocyte
disaggregation shear stress in hypertension," Hypertension, vol. 20, pp. 247-252, 1992.
[47l G. Ruhemtroth-Bauer, O. MBssmer, J. 0111, S. Koenig-Erich, and 0. Heinemann,
"Highly significant negative correlations between erythrocyte aggregation value and senim
concentration of high density lipopmteia cholesteml in a sample fiom a nomal population
and in patients with coronary heart disease," Eur J Ciin Invest, vol. 17, pp. 275-279,
1987.
[48] R. E. N. Shehaàa, R. S. C. Cobbolà, and L. Y. L. Mo, 'Aggregation efftxts in whole
blood: Influence of time and shear rate measured using ultrasound," Biorieology, vol. 31,
pp. 115-135,1994.
[49] K. K. Shung and G. A. Thieme, Uftrasonic scattering i~ biologicul trssues. 1st ed.
Boca Raton, A m Arbor, London, Tokyo: CRC Press, 1993.
[SOI B. Sigel, J. Machi, J. C. Beitler, J. R. Justin, and J. C. U. Coelho, 'Variable
ultrasound echogenicity in flowing bfood," Science, vol. 218, pp. 1321-1323, 1982.
[SI] J. F. Stoltz and M. Donner, Wemorheology: Importance of erythrocyte
aggregation," C h i c a l Hemorheology, vol. 7, pp. 15-23, 1987.
[52] A. J. Vander, J. H. Sherman, and D. S. Luciano, Humun physwfogy. The
mechaniFms of body funcrion. 6th ed. New York, St. h i s , San Francisco, London.
Toronto, Tokyo: McGraw-Hill, Inc. 1997.
[53] X. Weng, G. Cloutier, P. Pibarot, and L. G. Durand, "Cornparison and simulation of
different levels of erythcocyte aggregation with pig, home, sheep, cal€, and nonnal human
blood," Biorheobgy, vol. 33, pp. 365-377, 1996.
[54] Y. W. Yuan and K K. Shung, Wltrasonic backscatter nom Bowing whole b l d : II.
Dependence on kequency and fibriaogen macentration." J Acourt Soc Am, vol. 84, pp.
1195-1200,1988.
[SS] Y. W. Yuan and K. K. Shung, 'Uîrasonic backscatter nom flowing whole blood: 1.
Dependence on shear rate and hematocrit," JAwwt SocAm, vol. 84, pp. 52-58,1988.
[56] Y. W. Yuan and K. K S hung, "Echoici ty of whole blood," J Ultrasound Med, vol. 8,
pp. 425-434,1989.
[57] F. Zamad, P. Voisin, F. Brunette, J. F. Bnintz, J. F. Stoltz, and J. M. Gilgenkrantz,
c~aemorheoiogical aboomialities in artend hypertension and their relation to d i a c
hypertrophy," J H y e w n vol. 6, pp. 2939297,1988.
1581 1. Fontaine, M. Bertrand, G. Cloutier, 'Validation of a system-based approach to
mode1 the ultrasound signal b a c b t t e r e d by red blood cells," to be submitted to IEEE
T m Biomed Eng.