close

Вход

Забыли?

вход по аккаунту

?

Breast screening with custom-shaped pulsed microwaves

код для вставкиСкачать
Breast screening with custom-shaped
pulsed microwaves
Adam Santorelli
Department of Electrical and Computer Engineering
McGill University, Montréal, Québec, Canada
April, 2012
A thesis submitted to McGill University in partial fulfillment of the requirements
of the degree of Master's in Electrical Engineering.
© Adam Santorelli 2012
Library and Archives
Canada
Bibliothèque et
Archives Canada
Published Heritage
Branch
Direction du
Patrimoine de l'édition
395 Wellington Street
Ottawa ON K1A 0N4
Canada
395, rue Wellington
Ottawa ON K1A 0N4
Canada
Your file Votre référence
ISBN:
978-0-494-84806-7
Our file Notre référence
ISBN:
NOTICE:
978-0-494-84806-7
AVIS:
The author has granted a nonexclusive license allowing Library and
Archives Canada to reproduce,
publish, archive, preserve, conserve,
communicate to the public by
telecommunication or on the Internet,
loan, distrbute and sell theses
worldwide, for commercial or noncommercial purposes, in microform,
paper, electronic and/or any other
formats.
L'auteur a accordé une licence non exclusive
permettant à la Bibliothèque et Archives
Canada de reproduire, publier, archiver,
sauvegarder, conserver, transmettre au public
par télécommunication ou par l'Internet, prêter,
distribuer et vendre des thèses partout dans le
monde, à des fins commerciales ou autres, sur
support microforme, papier, électronique et/ou
autres formats.
The author retains copyright
ownership and moral rights in this
thesis. Neither the thesis nor
substantial extracts from it may be
printed or otherwise reproduced
without the author's permission.
L'auteur conserve la propriété du droit d'auteur
et des droits moraux qui protege cette thèse. Ni
la thèse ni des extraits substantiels de celle-ci
ne doivent être imprimés ou autrement
reproduits sans son autorisation.
In compliance with the Canadian
Privacy Act some supporting forms
may have been removed from this
thesis.
Conformément à la loi canadienne sur la
protection de la vie privée, quelques
formulaires secondaires ont été enlevés de
cette thèse.
While these forms may be included
in the document page count, their
removal does not represent any loss
of content from the thesis.
Bien que ces formulaires aient inclus dans
la pagination, il n'y aura aucun contenu
manquant.
Abstract
Microwave imaging has been proposed as a possible complimentary imaging
technique to X-ray mammography for early-stage breast cancer detection and
screening. Based on the intrinsic differences in the various tissues of the breast
at microwave frequencies, microwave imaging offers an imaging modality that is
safe, pain-free, and without limitations on the frequency of the exam.
An initial time domain microwave imaging system has been previously
developed within our group. Accurate numerical models, for finite-difference
time-domain simulations, matched to the experimental system have been
developed.
Numerical simulations are used to assess the safety of the microwave imaging
technique. We calculate the maximum energy absorbed by the breast when
exposed to incident microwaves and ensure that these values fall within the
established thresholds.
In this thesis we will test our hypothesis that an augmented microwave
imaging system can improve tumour detection by making use of custom-made
pulses with critical frequency content. We integrate passive microwave circuitry
with a previously designed experimental microwave imaging system in order to
create a new system that transmits an optimized pulse. We contrast
measurement results of this newly developed system with those of the
previously developed experimental system when imaging various tissue
phantoms.
ii | P a g e
Sommaire
L'imagerie micro-ondes a été proposée comme une nouvelle technique pour la
détection du cancer du sein qui est complémentaire au mammographie. Cette
technique est basée sur les differences intrinsèques des tissus mammaires
différents à des frequences micro-ondes. L'imagerie mirco-onde est une
technique qui est sûr, sans douleur, et sans limitations sur la fréquence de
l'examen.
Un système préliminaire pour l'imagerie micro-ondes dans le domaine
temporel a déjà été mis au point. Des modèles numériques précis , pour les
simulations avec la "finite-difference time-domain" technique, qui sont en
accorde avec la système expérimental sont construit.
Des simulations sont utilisé pour évaluer la sécurité de la technique d'imagerie
micro-ondes. Nous calculons le maximum d'énergie absorbée par le sein quand il
est exposé aux micro-ondes incidentes et nous s'assure que les résultats sont en
accord avec les normes établis.
Dans cette thèse, nous allons tester notre hypothèse que un système
d'imagerie micro-ondes augmentée peut améliorer la détection des cancers en
utilisant des impulsions fait sur-mesure. Nous utilisons des circuits micro-ondes
passives, avec le système déjà développé, pour créer un nouveau système qui
transmet un impulsion optimisé. On compare les résultats de nos mesures avec
les deux systèmes quand on utilise des fantômes de tissus divers.
iii | P a g e
Acknowledgements
First and foremost, I would like to thank my 'mentor', Dr. Guangran Kevin Zhu.
I met Dr. Zhu on my first day working on the project; he led by example and
showed me that motivation and hard work pays off.
I am pleased to recognize and acknowledge all the hard work and help
provided by Ms. Emily Porter (PhD candidate). Working together on this project
over the past two years has been a productive experience (though often-times
quite the head scratching experience). I would like to thank her for making that
experience an enjoyable one; working in the lab has been a pleasure.
I would like to thank Professor Mark Coates and Professor Joshua Schwartz,
both of whom have been vital in my progression as a research student
throughout my Master's degree. Their critical and thought provoking comments
really helped me to better understand what it means to be a good researcher.
And, last but certainly nowhere close to being least, I would like to deeply
thank my supervisor, Professor Milica Popovid, without whom I would have
never had the opportunity to work on this project. Her counsel, advice, and
support were paramount in the completion of my Master's degree. Our weekly
meetings were not only beneficial in helping to answer the many issue which
arose but to serve as a reminder that a balance in all aspects of life is key for
success.
iv | P a g e
Table of Contents
CHAPTER 1: INTRODUCTION ...................................................................................................... 1
CHAPTER 2: BACKGROUND........................................................................................................ 7
2.1.1 Anatomy of the Breast ........................................................................................... 7
2.1.2 Pathology of Breast Cancer .................................................................................... 8
2.2 Microwave Imaging for Breast Cancer Detection ................................................... 11
2.3 Existing Microwave Imaging Systems ..................................................................... 16
2.4 Phantom Tissue Construction ................................................................................. 20
2.5 Microwave Safety Concerns.................................................................................... 23
2.6 Planar Microstrip Technology for Pulse Shaping .................................................... 25
CHAPTER 3: INITIAL EXPERIMENTAL SYSTEM .............................................................................. 27
3.2 Improvement to the Initial Design .......................................................................... 33
CHAPTER 4: DEVELOPMENT OF EXPERIMENTAL-MATCHING NUMERICAL MODELS............................ 34
CHAPTER 5: SAFETY ASSESSMENT WITH 3-D REALISTIC BREAST MODELS ........................................ 40
5.1 Numerical Model .................................................................................................... 41
5.2 Plane-Wave Excitation ............................................................................................ 43
5.3 Antenna Excitation .................................................................................................. 51
5.4 Conclusions ............................................................................................................. 63
CHAPTER 6: INTEGRATION OF PULSE SHAPING TECHNOLOGY ........................................................ 64
6.1 Pulse-Shaping Circuit............................................................................................... 64
6.2 The SBR Structure ................................................................................................... 65
6.3 Signal Routing.......................................................................................................... 68
6.4 Pulse Analysis .......................................................................................................... 69
6.5 Amplification ........................................................................................................... 72
CHAPTER 7: INTEGRATED EXPERIMENTAL SYSTEM DESIGN............................................................ 75
CHAPTER 8: EXPERIMENTAL RESULTS WITH AUGMENTED SYSTEM................................................. 80
8.1 Homogeneous Adipose Breast Phantom ................................................................ 82
8.2 Fat and Skin Breast Phantom .................................................................................. 89
8.3 50% Gland Breast Phantom .................................................................................... 95
8.4 70% Gland Breast Phantom .................................................................................. 101
v|Page
8.5 System Performance Assessment and Comparison ............................................. 107
Chapter 9: Summary and Conclusion.............................................................................. 111
REFERENCES ....................................................................................................................... 113
List of Figures
Figure 2.1.1: Anatomy of the female breast [11]................................................................ 7
Figure 2.1.2: A depiction of the lymphatic system in the breast region of a female [11]. . 8
Figure 2.2.1: Plot of the median relative permittivity for three separate adipose content
groups, reproduced from [13]. Group 1, 0 – 30 % adipose tissue, represented by the dotdash line, Group 2, 31 – 84 % adipose tissue, represented by the dashed line, Group 3,
85 – 100 % adipose tissue, represented by the solid line. ................................................ 13
Figure 2.2.2: Plot of the median conductivity for three separate adipose content groups,
reproduced from [13]. Group 1, 0 – 30 % adipose tissue, represented by the dot-dash
line, Group 2, 31 – 84 % adipose tissue, represented by the dashed line, Group 3, 85 –
100 % adipose tissue, represented by the solid line......................................................... 13
Figure 2.2.3: Plot of the median relative permittivity for three separate malignant tissue
groups, reproduced from [14]. The solid line, dash-dot line, and dash line represent
malignant tissue content of higher than 70 %, at least 50 %, and at least 30 %
respectively. ...................................................................................................................... 15
Figure 2.2.4: Plot of the median electrical conductivity for three separate malignant
tissue groups, reproduced from [14]. The solid line, dash-dot line, and dash line
represent malignant tissue content of higher than 70 %, at least 50 %, and at least 30 %
respectively. ...................................................................................................................... 16
Figure 2.4.1: An image of the new breast phantom. (Left) The conical glands make up
50% of the breast phantom by volume. (Right) In this image the breast phantom is
composed of 80% glandular tissue by volume. Reproduced from [25]............................ 22
Figure 2.4.2: Realistic breast phantom of irregular shape. (Left) A side view of the breast
phantom, note the asymmetry of the model. (Right) Aerial view of the breast phantom.
Reproduced from [26]....................................................................................................... 23
Figure 3.1: A high level depiction of the initial experimental system. Note the antennas
are placed within the radome slots. ................................................................................. 27
Figure 3.2: Time-domain plot of the impulse train used in [9]. ........................................ 28
Figure 3.3: Spectrum of the input pulse used in [9]. ........................................................ 29
Figure 3.4: Design schematic for the radome. Reproduced from [9]. .............................. 30
vi | P a g e
Figure 3.5: Photograph of the fabricated radome. Reproduced from [9]. ....................... 30
Figure 3.6: A photograph of the fabricated antenna. ....................................................... 31
Figure 3.7: Antenna performance plot. ............................................................................ 32
Figure 4.1: Image of the radome (purple), breast phantom (yellow), skin-mimicking
tissue (orange), matching medium (green), and antennas (red) modeled in SEMCAD-X.
This image provides a sliced view through the midpoint of the radome from four
different perspectives. ...................................................................................................... 35
Figure 4.2: Comparison of the best-fit Debye model and the measurement data [24] for
the fat and gland tissues. .................................................................................................. 38
Figure 4.3: Comparison of the best-fit Debye model and the measurement data [24] for
the skin and tumour tissues. ............................................................................................. 38
Figure 4.4: Input pulse recorded directly from impulse generator. ................................. 39
Figure 5.1.1: An image of the 3-D numerical breast model. Two view of the breast model
are provided, with the exterior region made transparent allowing to view the numerous
‘cells’ which make up the model. ..................................................................................... 42
Figure 5.2.1: Image of the simulation scenario, with each component labelled, with the
plane wave incident from the top. Reproduced from [31]. .............................................. 44
Figure 5.2.2: Perspective view of the simulation scenario with the plane wave incident
from the side. .................................................................................................................... 45
Figure 5.2.3: Peak local unaveraged SAR values. Simulation results (denoted by ‘o’ and
‘x’ markers) and corresponding linear fit for both z-directed and y-directed plane waves.
Reproduced from [31]....................................................................................................... 48
Figure 5.2.4: Cross-sectional profile of the unaveraged SAR values for a plane wave
propagating along the y-axis at 8 GHz. ............................................................................. 49
Figure 5.2.5: Cross-sectional profile of the unaveraged SAR values for a plane wave
propagating along the z-axis at 8 GHz. ............................................................................. 49
Figure 5.2.6: Peak 10-g averaged SAR values. Simulation results (denoted by ‘o’ and ‘x’
markers) and corresponding linear fit for both z-directed and y-directed plane waves.
Reproduced from [31]....................................................................................................... 51
Figure 5.3.1: A plot of the S11 of the antenna embedded in three separate matching
mediums. Reproduced from [47]. ..................................................................................... 52
Figure 5.3.2: Time domain representation of each of the three pulse shapes. The pulses
have been normalized to their respective maximum and time-shifted so they can be
plotted on the same graph. .............................................................................................. 54
Figure 5.3.3: Spectral content of the three pulses analyzed. Reproduced from [47]. ..... 54
Figure 5.3.4: Side view of the simulation environment showing the four possible antenna
locations. Antennas in position 2 and 4 are cross-polarized in comparison to antennas
placed in position 1 and 3. Reproduced from [47]. .......................................................... 55
Figure 5.3.5: Front view of the simulation environment illustrating the various tissues of
the breast model. Note that the breast is in front of the antenna from this viewing angle;
the antenna is at no point in contact with the skin. Reproduced from [47]. ................... 56
vii | P a g e
Figure 5.3.6: (A)-(D) Show the distribution of unaveraged SAR values for illumination of
the breast at 8.9 GHz with the antenna located in positions 1 through 4 respectively.
Each cross-section contains the peak unaveraged SAR value for each respective scenario.
The scale is noted in dB relative to this local peak SAR value. Note how the antenna
position changes the location of the local peak as well as the distribution of the
absorbed energy within the breast. A sample image of the antenna is shown to indicate
its approximate location. .................................................................................................. 61
Figure 6.1.1: An illustration of the pulse shaping circuit and the subsequent components
used to achieve successful integration into the experimental system. ............................ 65
Figure 6.2.1: Time domain plot of the target pulse shape. .............................................. 66
Figure 6.2.2: Spectral content of the target pulse. The majority of the frequency content
is in the 2 - 4 GHz range, with suppression of the outside frequencies. .......................... 66
Figure 6.2.3: S11 of fabricated SBR. This plot shows the frequency content which is
reflected by the device. .................................................................................................... 67
Figure 6.2.4: Photograph of the fabricated SBR. The dimensions of the device are 2.9cm
x 14.3 cm. .......................................................................................................................... 68
Figure 6.3.1: Comparison of the frequency content of the signal reflected by the SBR
after it has passed through the signal routing components (circulator and coupler). A
plot of the S11 for the fabricated SBR is also included to provide a comparison. ............. 69
Figure 6.4.1: Time domain comparison of the initial pulse (blue) and the newly reshaped
pulse (orange). .................................................................................................................. 71
Figure 6.4.2: Spectral profile of the generic impulse and the reshaped pulse. The target
spectral shape is included to show the successful implementation of the pulse shaping
technology. ....................................................................................................................... 72
Figure 6.5.1: Frequency domain comparison of the pulse prior to and post amplification.
.......................................................................................................................................... 73
Figure 6.5.2: Amplification of the reshaped pulse from a 1.2 V peak-to-peak signal to a
16 V peak-to-peak signal. .................................................................................................. 74
Figure 7.1: A high-level depiction of the complete experimental setup. The setup
includes a clock to drive the pulse generator, a directional coupler to route the signal
from the SBR structure towards the antenna, a radome to house the antennas and the
breast phantom, and an oscilloscope to record the time-domain data. Note that the two
antennas are placed within the slots of the radome. In this case the antennas are
arranged such that the transmitted signal will be recorded. To record the reflected signal
the antennas are placed on the same side of the radome. .............................................. 76
Figure 8.1.1: Time-domain measurement of the received signal for the baseline
recording of Case 1. The plot compares the recorded signal when the breast is
illuminated by both the generic impulse and the reshaped pulse when using the SBR. . 82
Figure 8.1.2: Frequency content of the recorded baseline signals for Case 1 using both
the generic impulse and newly formed pulse to excite the antenna. .............................. 84
Figure 8.1.3: A plot comparing the tumour response signal, in the time domain, for Case
1, with both experimental systems................................................................................... 84
viii | P a g e
Figure 8.1.4: A comparison of the spectral content of the tumour response signal, for
Case 1, with the SBR-enabled system and the original system. ....................................... 86
Figure 8.2.1: A comparison of the recorded baseline signal for both experimental
systems when the antennas are arranged as per Case 1.................................................. 89
Figure 8.2.2: A comparison of the frequency content of the baseline signal for Case 1
when the breast phantom is illuminated by the generic impulse and the reshaped pulse.
.......................................................................................................................................... 90
Figure 8.2.3: A comparison of the tumour response signal, for Case 1, for both
experimental systems. ...................................................................................................... 91
Figure 8.2.4: Frequency response of the tumour response signal for both experimental
system when the antennas are arranged in Case 1. ......................................................... 92
Figure 8.3.1: Time-domain measurement of the received signal for the baseline
recording of Case 3. We compare the recorded signal for both the experimental system
outlined in our previous work and the newly designed system. Reproduced from [48]. 95
Figure 8.3.2: Frequency content of the recorded baseline signal for Case 3 using both the
generic impulse and the newly formed pulse to excite the antenna. Reproduced from
[48]. ................................................................................................................................... 96
Figure 8.3.3: Time-domain measurement of the tumour response for Case 3. We
compare the tumour response signal for both experimental prototypes. Reproduced
from [48]. .......................................................................................................................... 97
Figure 8.3.4: A comparison of the frequency content of the tumour response signal for
Case 3, for both experimental systems. Reproduced fro [48]. ......................................... 98
Figure 8.4.1: A comparison of the time domain recordings of the baseline signal for Case
3 with both experimental systems.................................................................................. 101
Figure 8.4.2: A comparison of the spectral content of the recorded baseline signal for
both input signals. The antennas are arranged as per Case 3. ....................................... 102
Figure 8.4.3: Tumour response signal comparison for both experimental systems. The
antennas are arranged as per Case 3.............................................................................. 103
Figure 8.4.4: A comparison of the frequency content of the tumour response signal
between the two experimental systems. The antennas are arranged as per Case 3. The
power from the signal computed using the original system is buried within the noise of
the system. ...................................................................................................................... 104
ix | P a g e
List of Tables
Table I: Difference Between Malignant and Benign Neoplasm .......................................... 9
Table II: SAR Exposure Limits ............................................................................................ 23
Table III: Best-Fit Debye Model Parameters for Various Breast Tissues ........................... 37
Table IV: Mean-Squared Error for Best-Fit Model for Various Breast Tissues.................. 37
Table V: Debye Parameters Describing Dispersive Model (Reproduced from [31])......... 44
Table VI: Incident Power Selection Per Frequency ........................................................... 46
Table VII: Peak Unaveraged SAR in mW/KG, Corresponding Excitation Frequency in
Parenthesis........................................................................................................................ 57
Table VIII: Peak 10-G Averaged SAR in mW/KG, Corresponding Excitation Frequency in
Parenthesis........................................................................................................................ 62
Table IX: Illustration and Description of Antenna Arrangements ..................................... 78
Table X: Maximum Received Signal and Peak Tumour Response Signal Comparison ...... 87
Table XI: Comparison of the T Parameter ......................................................................... 87
Table XII: Comparison of the T/B Parameter for Both Input Signals ................................ 88
Table XIII: Maximum Received Signal and Peak Tumour Response Signal Comparison ... 93
Table XIV: Comparison of the T Parameter....................................................................... 93
Table XV: Comparison of the T/B Parameter for Both Input Signals ................................ 94
TABLE XVI: Maximum Received Signal and Peak Tumour Response Comparison ............ 99
TABLE XVII: Comparison of T Parameter........................................................................... 99
TABLE XVIII: Comparison of the T/B Parameter for Both Input Signals .......................... 100
TABLE XIX: Maximum Received Signal and Peak Tumour Response Comparison .......... 105
TABLE XX: Comparison of T Parameter ........................................................................... 106
TABLE XXI: Comparison of the T/B Parameter for Both Input Signals ............................ 106
TABLE XXII: Maximum Tumour Response Signal Comparison (in mV) Across Trials for
Each Phantom Type and experimental System, [48] ...................................................... 108
TABLE XXIII: Comparison of T Parameter (dB) across trials for each phantom type and
experimental System, [48] .............................................................................................. 109
x|Page
CHAPTER 1: INTRODUCTION
Breast cancer is the most commonly diagnosed cancer for Canadian women. In
2011 alone, over 23 000 Canadian women were diagnosed with breast cancer,
increasing the probability to over 11% for a Canadian woman to develop breast
cancer in her lifetime [1]. Early detection of the disease is imperative to ensure
proper treatment and to improve survival rates. Currently, the 5-year survival
rate is 100% for invasive tumours classified as Stage I or lower (Stage 0).
However, for tumours classified as Stage II and Stage III the 5-year survival rate
drops to 86% and 57% respectively [2]. Clearly, vigilant screening is paramount to
ensuring that the disease is detected in its early stages to prevent the spread of
the cancerous tissues and to limit the subsequent debilitating effects of the
disease [2]. This has led to a search for a screening technique that is readily
available to women of all ages, allows for a regular screening protocol with
limited intervals between tests, and does not pose any health concerns.
A thorough and well-established screening procedure allows for early
detection of the disease in patients who present without any symptoms, or high
risk factors, of breast cancer. Unfortunately, in Canada, the current screening
procedure is heavily limited. It is suggested that women begin discussing their
risk of developing breast cancer with their doctor's at age 40. Screening for
breast cancer is suggested for women 40-69 years of age. For younger, and
older, women there is no specific screening procedure in place [3].
The suggested screening procedure is reliant on using X-ray mammography to
detect the presence of cancerous tissue, with the exam being administered once
every two years for women over the age of 40. X-ray mammography is the most
commonly applied imaging technique for both screening and diagnostic
purposes. This imaging modality suffers from several drawbacks; it exposes
patients to ionizing radiation, the exam requires the uncomfortable compression
1|Page
of the breast, it has difficulty imaging the breasts of young women (due to
density of the tissues), and, in the case an abnormality is observed, further
testing (biopsy or ultrasound) is required. In fact, according to the Breast
Imaging-Reporting and Data System (BI-RADS) categorization of mammograms, a
biopsy is required for diagnosis even if there is only a 23% chance that the
abnormality is cancerous [3]. Furthermore, according to the results in [4], using
X-ray mammography missed 67% of the cancerous lesions in women at high
familial risk of developing breast cancer. Hence, the desire to use X-ray
mammography as an early-stage screening technique with minimal time intervals
between exams for women of all ages does not seem prudent. Additionally,
women more likely to develop breast cancer at an early age usually fall into the
high-risk category, and, as per [4], mammography is not efficient at detecting the
presence of cancer for these women.
Currently, magnetic resonance imaging (MRI) and ultrasound (US) imaging are
considered as the primary supplemental imaging techniques in breast cancer
screening and diagnosis [3], [4]. Ultrasound is performed to assess abnormalities
found within the breast from mammography; it is also used for US-guided
biopsy, providing a detailed report on the location of the abnormality to be
sampled, when the US image is inconclusive. Ultrasound imaging provides an
inexpensive, portable, and non-invasive alternative to mammography.
Unfortunately, as reported in [4], it has a false-negative rate of just over 60% and
a false-positive rate of 10% in women at high-risk of developing breast cancer.
MRI has been shown to be much more effective than both mammography and
US for breast cancer screening [4], [5], with a false negative rate under 10% in
women with a high predisposition of developing breast cancer. Furthermore, of
the 43 cancers identified in the 529 women, 19 of these cancers were detectable
with MRI only (mammography and US presented false-negative). Additionally,
these previously undetected tumours had a mean size of 9mm, and were all
classified as stage I [4].
2|Page
This lacklustre performance of X-ray mammography and US imaging
techniques suggests that alternative imaging techniques should be used in the
development of new, more robust, screening procedures. While MRI has been
proven to be efficient at detecting tumours from the onset of their formation
even in the most difficult of cases, limitations of this technology (size, cost, ease
of operation) prevent it from being a viable candidate as an imaging technology
for an early-stage breast cancer screening protocol. Hence, recent research is
focused on new imaging modalities which address the following issues: cost,
ease of operation and availability, safety of the technique, and the ability to
detect the formation of cancerous tissues in the early stage of their
development.
Microwave imaging is one of the recently proposed imaging techniques to
serve as a complementary, and potentially supplementary, screening procedure
to the currently existing techniques. Since microwave imaging is reliant on the
illumination of the breast with microwaves, there are no concerns in regards to
the exposure of ionizing radiation. Additionally, the required equipment for a
well-designed microwave system is relatively inexpensive and discrete in size.
The exam itself can be a pain-free, non-invasive procedure. Hence, microwave
imaging is an attractive novel imaging technique for an early-stage breast cancer
screening protocol which could include more frequent screenings and for which
participation in the screening program can begin at a relatively young age.
The premise of the microwave imaging techniques hinges on the detection of
small tumours by exposing the breast tissue to incident microwave energy and
recovering information based on the amount of energy which is absorbed,
reflected, or scattered. This recovered signal can be analysed, and a profile of the
breast, and subsequent images, can be created to pinpoint the location of the
cancerous tissues. The malign tissue in the breast can be detected with
microwave illumination, due to the intrinsic electrical differences between the
3|Page
healthy and tumourous tissue at these frequencies [6]. This difference in the
dielectric properties of varying tissues gives rise to the reflection, transmission,
and absorption of the incident microwaves. Furthermore, since the majority of
breast tissue is adipose tissue, the signal propagating through the heterogeneous
breast structure undergoes minimal attenuation, improving the quality of the
recovered signals [6].
There are three primary microwave imaging techniques for breast cancer
detection: passive, hybrid, and active methods. In all three methods the
illumination of the breast is similar; incident microwave energy illuminates the
breast. The differences between the methods lies in the information recorded.
Passive methods are focused on measuring changes in temperature within the
breast. A thermal map can be created and suspicious areas, or hot spots, are
investigated. Hybrid methods combine microwave imaging with ultrasound
techniques. These methods are commonly referred to as microwave-acoustic
imaging techniques. An ultrasonic transducer is used to measure pressure waves,
generated by the expansion of tissues due to the absorption of energy, near the
surface of the breast. The most popular of the methods has been the active
technique. Active antenna-elements surround the breast tissue and are used to
transmit and receive microwave signals. The presence of abnormalities within
the breast will cause changes, both location and magnitude, in the detected
signal at the receiving antennas with respect to the baseline healthy case [7].
The major safety concern for microwave imaging techniques is the prospect of
tissue heating due to the absorption of the incident microwaves. Since the
various tissues of the breast have different conductivities, this absorption will
not be uniform across tissues. The most commonly employed metric to assess
the amount of energy deposited in human tissues is the specific absorption rate
(SAR). Health Canada and IEEE have well established maximum SAR values for
specific situations. It is imperative that the microwave imaging techniques satisfy
4|Page
these criteria to be considered safe and guarantee there is no potential damage,
due to overheating, to the illuminated breast tissue [8].
This project is focused on developing and testing an active microwave timedomain system for breast cancer screening. We wish to develop an experimental
system which is discrete in size, inexpensive to fabricate, and easy to operate. An
ultrawideband pulse is transmitted from one antenna while additional antennas
around the breast are used to collect the transmitted and reflected signals. In
this thesis, we analyse the power deposition within the breast tissues and assess
whether the absorbed energy complies with existing safety standards, we focus
on developing and utilizing passive pulse-shaping circuitry to reshape the input
signal to our time-domain measurement system in order to improve tumour
detection.
This thesis is divided into several sections to provide the reader with an
understanding of the motivation for this research, previous developments in the
microwave imaging field, and recent work done to improve our initial
experimental setup. Chapter 2 presents background information on the subject.
The reader will be acclimated with an in-depth examination of the development
of breast cancer and the breast anatomy. Chapter 2 also provides a detailed
explanation on why microwave imaging for breast cancer is a suitable imaging
modality, as well as introducing some theory behind key topics that are required
for the development of the experimental system (e.g. breast phantom
construction, safety concerns, pulse shaping theory). Chapter 3 presents the
initial experimental system developed within our research group, first described
in [9], and lists potential recommendations for improvements to this system.
In Chapter 4 we present the development of numerical models to closely
mimic this experimental setup. The results of an analysis of the energy levels
which are deposited into the human breast tissue from exposure to microwaves
emitted from an antenna placed in the near field are presented in Chapter 5.
5|Page
Specifically, we are interested in analysing the amount of energy absorbed by the
breast when illuminated by the antenna from our experimental setup at specific
locations of interest, which correspond to the antenna locations from this setup,
at several frequencies in the microwave region. This data is used to provide a
safety assessment of the developed experimental system, and provide insight on
the potential maximum input power to the antenna to meet the safety criterion
of both IEEE and Health Canada. The analysis is based on conducting finitedifference time-domain (FDTD) simulations with a three-dimensional (3-D)
anatomically realistic breast model placed in the near-field of the antennas.
The focus of this thesis is the development of an augmented experimental
system to improve tumour detection capabilities; specifically, we use a
synthesised planar transmission line to shape a specific waveform and create a
new input pulse. This microstrip transmission line is used to reshape the generic
impulse of the system presented in [9]. The benefits of this technique are twofold: i) the newly reshaped pulse has a specific wave form that improves signal
transmission through the breast structure by focusing signal power into a chosen
frequency band, and ii) this pulse shaping tool provides an inexpensive, easily
integrated solution for arbitrary wave form generation, in contrast to expensive
and bulky GHz-band waveform generation instruments. The integration of this
technology and the subsequently developed system are presented in Chapters 6
and 7. An in-depth time-domain analysis of the ability of this new system to
detect tumours is carried out and presented in Chapter 8. These tumours are
placed in various life-like tissue phantoms, composed of fat, skin, and glandular
tissues. A comparison of the tumour detection capabilities of both experimental
systems, that in [9] and the system presented here within, in these various
breast phantoms allows for an assessment of the improvements made from the
implementation of the pulse-shaping technique.
6|Page
CHAPTER 2: BACKGROUND
This chapter is focused on providing the reader with background information
on both the theoretical aspects and previous experimental work which have
allowed for the development of microwave imaging systems for breast cancer
detection.
2.1.1 Anatomy of the Breast
The breast is primarily composed of adipose tissue which surrounds a network
of glandular tissue. This glandular tissue is made up of lobules and connective
ducts. The lobules, composed of hollow grape-like clusters which store and
create breast milk, are arranged into groups called lobes. These milk-producing
glands are connected to the nipple via the connective ducts. There are
approximately 15 to 25 of such lobes within the breast. Breast tissue is not
limited to the breast itself; it extends up to the collarbone and stretches from the
armpit and attaches and to the breastbone and the centre of the chest. The
adipose tissue fills the space between the skin and these glandular tissues
forming a protective layer for these important glands [1], [10].
Figure 2.1.1: Anatomy of the female breast [11].
7|Page
Additionally, embedded within the breast tissue is a series of lymphatic vessels
used to drain the breast tissue of lymph (a fluid which contains a wide array of
white blood cells). These lymphatic vessels are connected to a network of lymph
nodes, which are then connected to the rest of the lymphatic system of the
human body. This system serves to transport, and drain, the lymph fluid
throughout the body through a series of intricate connections between lymph
nodes. It is possible for some of the cancerous cells to become trapped within
the lymph nodes, and due to the lymphatic system, this can cause breast cancer
to spread to other regions of the human body [1], [10].
Figure 2.1.2: A depiction of the lymphatic system in the breast region of a female [11].
2.1.2 Pathology of Breast Cancer
This section provides additional information about specifics about breast
cancer development. We address some common misnomers about
nomenclature and terminology, provide a more detailed analysis about the
development of breast cancer from a pathological view point, and a summary
about the staging protocol for the development of breast cancers.
While the word tumour is often used to describe cancerous tissue, the more
exact definition refers to any form of swelling from any process. This includes
8|Page
hypertrophy, calluses, and inflammation, amongst others. A more precise term
would be neoplasm. Neoplasm literally translates to new growth or formation. It
is commonly used to refer to any abnormal mass of tissue cells which seems to
grow extremely quickly and without cessation. Furthermore, this growth is
uncoordinated with surrounding nearby tissues and can continue even in the
absence of the stimulus which provoked the formation. Most neoplasms begin
from a single cell in which the DNA has been altered from its parent cell [12]. For
the remainder of this thesis we have adopted the more common nomenclature
of referring to a tumour as a growth of cancerous tissue (malignant neoplasm).
A neoplasm can be benign or malignant in nature. A malignant neoplasm is
classified as cancer [12]. The most specific, practical, and most frequently used
technique to classify a neoplasm is based on the histology of the cells (hence the
requirement for biopsy for breast cancer diagnosis). Malignant neoplasms differ
from benign neoplasms in that they are fast growing and aggressive in nature;
specifically, they have the ability to invade and destroy surrounding tissue and
have the ability to develop a secondary point of tissue growth (metastasis) at a
distance from the primary tumour. This potential for metastasis is extremely
dangerous as it is the cause for cancers spreading throughout the human body. A
summary of some of the differences between benign and malignant neoplasms
based on behaviour and appearance is outlined below in Table I [12].
TABLE I: DIFFERENCE BETWEEN MALIGNANT AND BENIGN NEOPLASM
Benign
Behaviour


No metastatic potential
Slow growth
Appearance

Encapsulated, spherical,
well defined region
Nucleus of regular shape
and size

Malign



Metastatic potential
Rapid growth
Can infiltrate surrounding
tissue


Irregular shape
Enlarged, bizarre, and
variable shape and size of
nucleus
9|Page
Breast cancer most commonly develops in the upper outer quadrant of the
breast, near the armpit. These cancers are classified as in-situ (in place)
carcinomas and invasive carcinomas. Nearly all of these cancers start in the
glandular tissues of the breast, in either the ducts or the glands [3]. The in-situ
carcinomas include ductal carcinoma in situ (DCIS), lobular carcinoma in situ
(LCIS), and mammary Paget disease. The term in situ refers to the fact that the
tumour has not invaded any surrounding tissue and remains well confined inside
the original site of tissue growth [3], [12]. DCIS is the most commonly diagnosed
breast cancer, in which the cancerous tissues have begun to form on the inside
lining of the breast duct [3]. Interestingly enough, by definition, DCIS and LCIS
are not actual carcinomas since the tumour remains well encapsulated and has
not infiltrated any of the surrounding tissues. There is some evidence that these
cells have the potential to form invasive carcinomas, hence, they are commonly
referred to as malignant tumours and are removed if diagnosed [3].
Invasive carcinomas, primarily invasive ductal carcinoma (IDC) and invasive
lobular carcinoma (ILC), are advanced forms of DCIS and LCIS, in which the
cancerous growth has infiltrated the surrounding tissues and is no longer limited
by the duct or the lobular wall.
An important aspect of prognosis and treatment of cancer is the assessment of
the spread of the cancerous tissues. Staging is used to evaluate the extent of
which the tumour has spread. For breast cancer a TNM system of staging is used.
This system is based on the size and spread of the primary tumour (T), whether
the cancerous cells have spread to any lymph nodes (N), and whether any distant
metastases (M) have formed [3], [12]. Tumours can be classified from Stage 0
through Stage IV, where Stages II and II have additional sub-stages. As
mentioned in Chapter 1, for breast cancer at Stage II or higher the survival rate
begins to decrease dramatically. Thus, for the design of an experimental earlystage screening device we need to ensure that the system can detect tumours
10 | P a g e
that would be classified as Stage I or lower. The maximum tumour size for breast
cancer categorized as Stage I is 2cm in diameter.
2.2 Microwave Imaging for Breast Cancer Detection
The premise for using microwave imaging for breast cancer detection is based
on inherent differences in both electrical conductivity and permittivity among
the breast tissues at microwave frequencies. These differences in the dielectric
properties can be directly tied to specific tissue types. The contrast between
malignant and healthy breast tissues will create various interfaces within the
heterogeneous breast structure causing non-uniform scattering and absorption
of the incident microwave energy. The greater the contrast in the dielectric
properties of malignant and healthy tissues, the more readily discernible the
tumour is believed to be. The relatively low-loss nature of the breast, due
primarily to the large percentage of adipose tissue within the breast, enables the
recovery of these scattered and reflected signals. From these recordings an
image of the breast can be constructed and used to pinpoint the location of any
abnormalities.
Unfortunately, various tissues within the breast complicate the imaging
problem by decreasing the contrast between the adipose and tumourous tissues.
A group of papers published by Lazebnik et al., [13] and [14], are the current
benchmark in characterizing the electrical properties in the various tissues within
the breast.
In [13], healthy breast tissue was excised from breast reduction surgeries.
These removed tissues were then used to characterize the behaviour of healthy
breast tissue over the 0.5 – 20 GHz range. The importance of this study is twofold. First, the sample size was the largest to date, spanning over 350 samples
taken from over 90 patients. Secondly, a histological assessment was used to
categorize the percentage of fatty tissue in the sample under investigation. Using
11 | P a g e
this process the samples were categorized in three specific groups: Group 1
contained samples with 0 – 30 %, Group 2 with 31 – 84 %, and Group 3 with 85 –
100 % adipose tissue, respectively. Groups 1 and 2, with low percentage of
adipose tissue, correspond closely to the glandular and fibroconnective tissues
found within the breast [13].
Figures 2.2.1 and 2.2.2 are a reproduction of the data presented within [13].
Specifically, Fig. 2.2.1 plots the relative permittivity for the three groups
mentioned above fit to a Cole-Cole dispersive model (providing a relationship
between relative permittivity and frequency). Clearly, a correlation between the
percentage of adipose tissue and the relative permittivity exists; there is an
inverse relationship between the adipose tissue content and the relative
permittivity. Group 3, with the highest percentage of adipose tissue, has, by far,
the lowest permittivity, whereas Group 1 has the highest permittivity. Similarly,
Fig. 2.2.2 plots the electrical conductivity. A similar relationship between the
fatty tissue content and the electrical conductivity can be observed.
Additionally, in each of the plots error bars, representing the 25 th and 75th
percentile of each group, are included [13]. Error bars indicate the variations in
the electrical properties of tissue sample pertaining to the same group. From Figs
2.2.1 and 2.2.2, there appears to be an increased variation in the lower adipose
content groups (Group 1 and 2). In fact, there is an overlap between the two
groups in both plots (permittivity and conductivity).
12 | P a g e
Figure 2.2.1: Plot of the median relative permittivity for three separate adipose content groups,
reproduced from [13]. Group 1, 0 – 30 % adipose tissue, represented by the dot-dash line, Group 2, 31 –
84 % adipose tissue, represented by the dashed line, Group 3, 85 – 100 % adipose tissue, represented by
the solid line.
Figure 2.2.2: Plot of the median conductivity for three separate adipose content groups, reproduced from
[13]. Group 1, 0 – 30 % adipose tissue, represented by the dot-dash line, Group 2, 31 – 84 % adipose
tissue, represented by the dashed line, Group 3, 85 – 100 % adipose tissue, represented by the solid line.
13 | P a g e
In [14], tissue samples were taken from cancer surgeries (including biopsy and
mastectomy). A total of 155 samples from 119 patients were analysed. Of these
samples, the tissues were first broken into three groups, healthy, malign, and
benign tissues.
The healthy tissues correspond to the excised tissue which presented with no
lesions or tumourous development. These samples were categorized by
percentage of adipose, glandular, and fibroconnective tissues. These samples
were analysed following the same procedure of [13]. There was no significant
difference between the dispersive Cole-Cole models for normal tissue from
breast reduction and cancer surgeries [14]. This repeated result helped served as
validation that the procedure first implemented in [13] could be reproduced
without introducing error.
The main focus of [14] was to categorize the dielectric properties of the
cancerous tissues. The malignant tissue samples were organized into three
groups based on the percentage of cancerous tissue within each sample
(assessed by a pathologist based on the histological exam). Figures 2.2.3 and
2.2.4 reproduce the plots from [14] detailing the relationship between frequency
and the corresponding permittivity and conductivity for each of the three groups
respectively. As in Fig. 2.2.2 the error bars correspond to the 25th and 75th
percentile recordings. The variation between data for the malignant tissue was
not statistically significant [14]. In fact, from the two figures below we observe
that for both the relative permittivity and electrical conductivity the differences
between the median of each of the three groups is minimal. Furthermore, the
separate recordings are much more consistent and less varied when compared
to the data from healthy tissue samples.
These two studies have played an integral part in understanding how different
tissues behave from an electrical standpoint, categorizing the permittivity and
conductivity, and their changes with frequency, of the adipose, glandular, and
cancerous tissues which can be found within the breast. This categorization has
14 | P a g e
helped to create more accurate models for use in commercial FDTD simulators
by assigning specific parameters for the Debye dispersive model of the various
tissues [15]. From [13], and [14] it was shown that the variability of these
dielectric parameter due to patient age, sample temperature, and from patient
to patient was shown to be statistically irrelevant. These studies have also
shown that the imaging problem is not as easy as initially believed. While the
contrast between adipose and malignant tissue can be as high as 1:10 [14], this
contrast is reduced to a factor of about 1:1.1 between glandular/fibroconnective
and malignant tissue.
Figure 2.2.3: Plot of the median relative permittivity for three separate malignant tissue groups,
reproduced from [14]. The solid line, dash-dot line, and dash line represent malignant tissue content of
higher than 70 %, at least 50 %, and at least 30 % respectively.
15 | P a g e
Figure 2.2.4: Plot of the median electrical conductivity for three separate malignant tissue groups,
reproduced from [14]. The solid line, dash-dot line, and dash line represent malignant tissue content of
higher than 70 %, at least 50 %, and at least 30 % respectively.
2.3 Existing Microwave Imaging Systems
In this section we explore some of the existing microwave imaging systems
which have been recently developed. In particular, we focus on the work of two
groups, Craddock et al. at the University of Bristol, and Meaney et al. at the
University of Dartmouth, who have successfully developed a microwave imaging
system and were able to implement a clinical prototype.
Craddock and colleagues have presented a range of papers, [16]-[20], which
detail the development of a microwave radar imaging system for breast cancer
detection. The system is designed to use stacked patch antennas that are cavity
backed. These antennas have been designed for optimal performance in the
ultrawideband (UWB) range; particularly, they are designed such that the S11
parameter is below -3 dB from 4 - 9 GHz (indicating that the majority of the
signal content is transmitted in this frequency range). The design incorporates 16
antennas, distributed in a 4 x 4 grid, which are arranged in a hemispherical
16 | P a g e
pattern. In the most primitive designs, [16], the antenna arrangement was
asymmetric; each row of antennas was offset due to physical limitations (space
for connectors and cabling). However, they have been recently able to reduce
the size of the antennas [17] to allow for a symmetrically aligned grid of
antennas (4 rows and 4 columns) [18]. Measurements are recorded using a
vector network analyser (VNA) that performs a frequency sweep from 4 - 10
GHz. The S21 parameter (forward transmission) is recorded between a specific
pair of antennas. A set of electromechanical switches are used to cycle through
all possible configurations of antenna pairs, in each measurement there is one
transmitting and one receiving antenna, such that 120 signals can be recorded.
Initial experiments, [18], [19], focused on analysis using breast phantoms.
These phantoms consist of a skin-layer, 2 mm thick, which surrounds a
hemisphere, with a radius of 58 mm, that is made up of entirely fat-mimicking
tissue. To reduce air gaps between the antenna array and the breast phantom a
liquid matching medium, in this case this matching-medium is the fat-mimicking
tissue, is poured into the antenna array. A spherical tumour phantom, of various
size (ex. 4 mm, 6mm, 10mm), is embedded within the fat phantom [18], [19].
Each scan of the breast requires the recording of all 120 signals, rotating the
antenna array with respect to the breast, and then collecting all 120 signals
again. Each scan takes upwards of 3 minutes. This rotation procedure is used to
obtain additional signals that are used to separate the tumour response from
other background signals (such as coupling between antennas and the skin
reflection). Reconstruction algorithms can be used to process these recorded
signal and re-create a 3-D image of the breast phantom. The system is capable of
detecting the various sized tumours [18], [19].
The biggest limitation of these studies is that the breast phantoms used in the
experimental setup do not provide a realistic estimation of the breast. The
contrast in permittivity between the healthy tissue and the tumourous tissue is
over-estimated, the phantoms are completely homogeneous, and are of ideal
17 | P a g e
shape. Furthermore, these breast phantoms are composed entirely of adipose
tissue, where as the human breast is made up of numerous glands. These
glandular structures serve to further complicate the imaging process since they
are quite similar, electrically, to the tumourous tissue.
Recently, the group at Bristol have implemented some minor changes to their
experimental system which has allowed them to successfully complete a small
clinical trial [20]. For the clinical trial, the system now consists of an antenna
array that is made up of 31 antennas arranged in a hemispherical pattern with
radius of 67 mm. While the group was able to successfully produce images of the
breast, the clinical trial allowed them to assess some of the difficulties that arise
in a clinical setting which do not exist in an ideal experimental setup. Of note
were the effects from patient movement and breathing which can have a
detrimental effect on the quality of the image. An additional problem was the
irregular, and varied, shape of the human breast, which led to differences in the
fit between the antenna array and the breast [20].
At the University of Dartmouth, Meaney and colleagues have also developed
an experimental system that has been successfully implemented and used in a
clinical setting. An initial report on the setup was presented in [21]; a microwave
imaging system is used to create 2-D images of the breast for chemotherapy
monitoring. In this system, 16 monopole antennas form a hemispherical array, of
75 mm radius, around the breast. As opposed to the system developed by
Craddock et al., the system in [21] is designed such that, for each measurement,
one antenna will be transmitting while all the others (15) are receiving. A
switching matrix is used to cycle the transmitting antenna through all 16
possibilities. Additionally, measurements are taken at discrete frequencies
between 300 and 900 MHz in 100 MHz increments. A total scan involves
mechanically moving the antenna array through 7 different planes (at different
heights in regards to the breast thickness), a process which takes roughly 15
minutes .This system was used as a clinical prototype, and tested on five
18 | P a g e
volunteers. The images obtained report a spatial resolution of approximately 1
cm; additionally, they report that abnormalities as small as 4 mm can be
detected [21].
In [22], Meaney et al. once again perform a clinical trial using the experimental
system from [21]. In this experimental paradigm the patients have previously
undergone MRI. This allows for a basis of comparisons between the MRI images
and the images generated by the 2-D microwave imaging system. Improvements
to the experimental system from [21] allow the 2-D images to be constructed in
less than 3 minutes; furthermore, Meaney et al. switched from a matching
medium that was a saline solution [21] to one that is made up of a mixture of
glycerine and water [22]. Unfortunately, in this paper, only a qualitative
assessment of the performance is given. There is no quantitative assessment of
the ability of the microwave imaging system to detect the location and size of
tumours in comparison to MRI results.
These two systems provide the motivation behind the design of our initial
experimental system [9]. However, where these systems are focused on
frequency-domain measurements our design implements a purely time-domain
system; improving the dynamic range of the system and decreasing the cost of
fabrication. Furthermore, we are interested in the development of more
complex and realistic breast phantoms to allow for more rigorous testing of the
experimental system. This thesis focuses on augmenting the initial experimental
system described in [9] by making use of passive microwave devices (planar
microstrip lines) in order to reshape a generic impulse into a specifically tailored
pulse shape which optimizes antenna performance and improves signal
transmission within the breast. We assess the improvements in the tumour
detection abilities of this system, when compared to [9], for various breast
phantoms of increasing realism.
19 | P a g e
2.4 Phantom Tissue Construction
In order to assess the viability and performance of an experimental
system breast phantoms, which closely mimic real breast tissues, can be
fabricated. These breast phantoms allow for testing in a modular fashion with
control of the unknown variables. Specifically, we can choose to create breast
phantoms of specific shape with known quantities of adipose and glandular
tissue, and we can specify tumour shape, size, and location. These various breast
phantoms allow us to determine whether the presence of the tumour can be
detected in specific situations. This experimental protocol allows us to assess the
aspects of the imaging system which can be improved. Most importantly, this
work is easily repeatable and avoids any of the ethical or legal challenges of
human or animal testing. For microwave imaging systems it is of the upmost
importance that the breast phantoms tissue have dielectric properties,
permittivity and conductivity, which closely match the recorded data for real
tissue.
In [23] a procedure has been developed to create breast tissue phantoms from
an oil-gelatin mixture. By varying the percentages of oil in the mixture it is
possible to create tissue phantoms with a wide array of variability in dielectric
properties. The permittivity and conductivity of these created phantoms are
measured from 500 MHz - 20 GHz. From these recordings they were able to
determine recipes to create specific tissue phantoms. A higher oil concentration
more closely resembled adipose tissue, whereas decreasing the percentage of
oil, and subsequently increasing the concentration of water, leads to the creation
of phantoms with a higher permittivity and conductivity, i.e. closer to glandular
and malignant tissues.
Recently, Porter et al. have focused on creating improved and complex breast
phantoms. In [24], a detailed procedure to create four specific breast tissue
phantoms, namely fat, skin, gland, and tumour, is presented. These various
tissues can all be created from the same, easily attainable and safe to use,
20 | P a g e
chemicals. The procedure to integrate these chemicals is identical to the
methodology presented in [23]. The four tissue types were created to ensure a
good match with the data from [13] and [14]. To confirm this result the
permittivity and conductivity of the four tissue types is recorded from 50 MHz - 6
GHz using a dielectric probe and a vector network analyser. For each recording at
least ten samples are taken from different locations on the phantom and then
averaged [24]. Furthermore, this paper presented a methodology to create
hemi-spherical breast phantoms consisting of a 2-mm exterior skin layer with the
interior filled with adipose tissue.
The next step was to create a breast phantom with increased realism in
regards to the geometry of the interior of the breast structure. Conical glandular
segments were created in to closely mimic the anatomy of the human breast;
glandular tissue can be thought of as a group of cones which start at the nipple
and flow outward [25]. By creating these glandular conical segments we can
choose a discrete number of cones to be included in the breast phantom. We
have determined each gland structure to have a volume of 28mL, and, coupled
with knowledge of the total volume of the breast phantom, we can directly
control the percentage of glandular tissue in the breast phantom. We fill the
remainder of the breast phantom with the fat-mimicking tissue. Figure 2.4.1
provides an image of two separate breast phantoms made up of 50 % and 80 %
glandular content by volume.
21 | P a g e
Figure 2.4.1: An image of the new breast phantom. (Left) The conical glands make up 50% of the breast
phantom by volume. (Right) In this image the breast phantom is composed of 80% glandular tissue by
volume. Reproduced from [25].
As of now, the created breast phantoms have all been hemi-spherical in shape
with a skin-layer of equal thickness. In [26], Porter et al. have developed realistic
breast phantoms of irregular shape and varying skin thickness. The purpose of
developing these new breast phantoms is to investigate what role the breast
shape, and subsequent varying skin thickness, will have on the imaging scenario.
Furthermore, it would be unrealistic to assume that the human breast is
perfectly hemi-spherical; these irregularly shaped breast phantoms correspond
to a more accurate representation of the geometry of a real breast. Two breast
shaped moulds were used to create these new breast phantoms. A countermould procedure is used to create a skin-layer with average thickness of 2 mm.
The skin layer is then placed back into the breast mould and filled with the
desired amount of adipose and glandular tissue [26]. Figure 2.4.2 provides an
image, side and aerial view, of the irregularly shaped breast phantom.
22 | P a g e
Figure 2.4.2: Realistic breast phantom of irregular shape. (Left) A side view of the breast phantom, note
the asymmetry of the model. (Right) Aerial view of the breast phantom. Reproduced from [26].
2.5 Microwave Safety Concerns
When exposed to electromagnetic (EM) waves human tissue is susceptible to
damage from two primary causes; nerve stimulation from interfering electrical
fields and tissue damage due to overheating. At frequencies above 100 kHz the
effects of nerve stimulation can be ignored, thus for microwave imaging systems
the primary concern is the amount of energy deposited into the tissue and the
subsequent heating [8], [27]. The most recognized and commonly employed
metric to assess the safety of an RF device is the specific absorption rate (SAR),
which provides information about the amount of energy absorbed by human
tissues when exposed to incident radiation over a set period of time. Table II
provides data about the maximum permissible SAR levels for various specific
situations [8], [27], where WB refers to whole-body averaging, and 1G and 10G
refer to 1-gram and 10-gram averaging, respectively.
TABLE II: SAR EXPOSURE LIMITS
Controlled
Environment [W/kg]
Uncontrolled
Environment [W/kg]
SARWB
SAR1G
0.4
8
0.08
1.6
SAR10G
20
4
23 | P a g e
These limits were established based on the computed maximum whole-body
averaged SAR value of 4 W/kg as the threshold for harmful thermal effects. For
controlled environments (workers, well isolated environment) and uncontrolled
environments (general public) a reduction factor of 10 and 50, respectively, has
been introduced [27], [28]. For localized exposure it is possible that, while the
whole-body average remains below these thresholds, local SAR values can
become very large. This led to the introduction of the 1-gram and 10-gram
averaged SAR values [28].
Microwave imaging devices aim to create an easily accessible technology for
breast cancer screening. It is possible, though highly unlikely, that under this
criterion that the general public may be exposed to the incident microwave
energy. Hence, it is necessary to ensure that the designed microwave imaging
device adhere to the safety limits associated with the uncontrolled environment.
Furthermore, these systems limit the exposure to the breast region. We are
therefore concerned with peak local SAR values averaged over small quantities
of mass. Specifically, we choose to follow the 4 W/kg SAR limit for averaging over
10 grams of tissue.
Previously, numerical work done by Zastrow et al. has concluded that the
application of microwave imaging modalities conform to the standard SAR limits
outlined above, and that, for a specific maximum input power, microwave
imaging is a safe technology [29], [30]. Additionally, we have presented
numerical simulation results in [31] confirming the safety of the microwave
imaging technique. These studies were based on either plane-wave illumination
([29], [31]) or used a breast model [32], developed before the extensive research
done by Lazebnik et al. in [14], that is based on outdated tissue properties ( [29],
[30]).
In this thesis we explore the use of a more complete breast model, developed
in [33]. This breast model makes use of the recently reported tissue parameters
from Lazebnik et al. [14]. This heterogeneous, 3-D anatomically realistic breast
24 | P a g e
model is exposed to radiation when placed within the near-field of a specifically
designed antenna. The antenna, which is the same as the one used in our
experimental research, is then swept through four locations. The analysis of the
data generated from these simulations will help to conclude whether the
radiation pattern of the antenna, the location of the antenna relative to the
breast, or the frequency of the incident microwave energy, changes the location
and magnitude of peak local SAR values.
2.6 Planar Microstrip Technology for Pulse Shaping
Recent literature [34] – [36] has demonstrated that is it possible to design and
fabricate a passive microwave device that can be used to reshape a generic
impulse and create a new desired waveform. In [34] a synthesis method is
developed which provides an exact analytical solution for converting a target
frequency response into a physical profile for a microwave transmission line.
Thus, for any passive, stable and causal frequency response desired we can
perturb the physical shape of a transmission line, in accordance to results of the
synthesis method from [34], and obtain a passive microwave device with the
expected response. In [36], this technology was successfully implemented by
reshaping the impulse from a generic impulse generator. Two different UWB
pulses, which adhere to the FCC mask for UWB signals, were synthesized using
two custom-designed microstrip lines. Each of these microstrip lines had a
unique and specific conductor-strip width profile along their length as dictated
by the results of the synthesis method of [34].
Ideally, we would like to transmit a chosen target waveform that improves
signal transmission through the breast and optimizes antenna performance. In
our experimental system we have a priori knowledge of both the available
(generic) impulse signal and the frequency response of the antenna. Hence, we
25 | P a g e
can calculate the required frequency response of the microstrip line to be
integrated within the system to create, in accord with the antenna response, a
pulse which matches the desired waveform we wish to transmit into the breast.
Following the procedures which have been outlined in [34] – [36] it is possible to
determine the desired frequency response, and the subsequent physical profile
of the conductor-strip, of the microstrip line to be fabricated.
These techniques can be beneficial for a microwave imaging system where the
detected signals can be quite low, due to input power limitations from the safety
standards that must be adhered to. By pre-shaping the transmit pulse’s
frequency content we can focus the signal power into a specific spectrum to
optimize antenna performance and signal transmission through the breast
structure. This process can improve post-processing of the recorded signals by
not only increasing the strength of the recorded signals, but by focusing the
majority of the signal content into a pre-established band of interest with a
specific pulse shape.
The purpose of implementing this arbitrary wave form generation using a
hardware approach is primarily because integrating a software based approach
to synthesize an arbitrary pulse in the GHz spectrum is comparatively more
costly. Fabricating a microstrip line, which is simply a metal trace on a dielectric
substrate, can be easily integrated with our experimental system and the costs
associated with the fabrication process are minimal when compared to any
commercial arbitrary waveform generator in the GHz frequency range.
Additionally, one can envision future microwave imaging systems which are
conformal in nature. In this scenario, the system is designed to adhere to the
contour of the breast. The potential exists to have this pulse shaping circuitry on
board the device by implementing the microstrip as a contour-mapped structure.
26 | P a g e
CHAPTER 3: INITIAL EXPERIMENTAL SYSTEM
In this section, an overview of the components and the design choices made in
the development of the initial time-domain microwave imaging system proposed
by Porter et al. in [9] is provided. Primary motivation for this system came from
previous work done by Craddock et al. [16]-[20], Meaney et al. [21]-[22], and
Fear et al. [37], amongst others, who have created frequency based systems by
using multiple antennas to surround the breast and record transmitted and
reflected signals with the use of a VNA. The development of this imaging system
was based on two important factors: i) the use of a purely time-domain system
to enhance the dynamic range of the system and ii) to reduce the footprint, both
economical and physical. A high-level depiction of the system presented in [9] is
provided in Fig. 3.1.
Figure 3.1: A high level depiction of the initial experimental system. Note the antennas are placed within
the radome slots.
27 | P a g e
The major components of the experimental system are the clock, impulse
generator, radome, antennas, and the oscilloscope. The clock (Tektronix
gigaBERT 1400 generator [38]) is operated at 250 MHz and is used to both drive
the impulse generator and to trigger the oscilloscope. Hence, the clock is used to
set the repetition rate of the input pulse.
The impulse generator (Picosecond Pulse Labs Impulse Generator Model 3600
[39]) creates a 70 ps full-width half maximum pulse with -7.5 V amplitude. This
pulse is a DC-centric unipolar signal. Figures 3.2 and 3.3 provide a time-domain
and frequency domain representation of the generated impulse. In Fig. 3.2 the
amplitude of the impulse has been diminished from -7.5 V to approximately -6.3
V. This attenuation is due to the loss in the cable connecting the impulse
generator and the oscilloscope.
Figure 3.2: Time-domain plot of the impulse train used in [9].
28 | P a g e
Figure 3.3: Spectrum of the input pulse used in [9].
The radome is a hemi-spherical structure, made from alumina (Si2O3),
consisting of 16 antenna slots. The radome has been fabricated by the company
Friatec [40]. The purpose of the radome is two-fold: i) to improve antenna
performance, and, ii) to provide a rigid housing structure for both the antennas
and the breast phantom under investigation, thus reducing the chance for
erroneous measurement due to human error (i.e. it ensures that the antennas
are always in the same position relative to each other and the breast phantom).
Alumina,  = 9.6,  = 0, was chosen as the fabrication material since the
permittivity of the material was close to that of the antenna substrate [9].
The design for the radome is shown in Fig. 3.4. The 16 slots of the radome are
organized into equally spaced quadrants (separated by 90°), each containing four
slots. The radome can be thought to be a 4 × 4 grid. Each of the four slots in a
quadrant are equally separated by 18° on a 8.5 cm radius. Each slot is 3 mm
thick, 18 mm wide, and 13 mm deep. In two of the quadrants, as can be
29 | P a g e
observed in Fig. 3.4, all four slots are oriented parallel to both each other and the
open face of the radome. However, in the two other quadrants each alternating
slot is oriented perpendicular to the previous slot. We refer to the slots that are
parallel to the open face of the radome as the co-polarized slots. The slots
oriented perpendicular to the co-polarized slots are referred to as being crosspolarized. In total, there are 12 co-polarized and 4 cross-polarized antenna slots.
This radome design allows for an investigation of the effect of antenna
polarization and tumour detection. The antennas placed in the perpendicular
slots will be cross-polarized compared to the antennas placed in the parallel
slots. A photograph of the fabricated radome is shown in Fig. 3.5.
Figure 3.4: Design schematic for the radome. Reproduced from [9].
Figure 3.5: Photograph of the fabricated radome. Reproduced from [9].
30 | P a g e
The antenna used in this experimental system is a Travelling Wave Tapered
and Loaded Transmission Line Antenna (TWTLTLA) [41], specifically designed for
microwave imaging of the breast. It is a compact, 0.635 x 12 x 18 mm3, wideband
antenna which has been tailored for optimum performance when placed in a
medium with dielectric properties similar to that of the adipose breast tissue (i.e.
 ≈ 9). A suitcase,  = 10,  ≈ 0, is used to hold the antenna in place and
minimize any air gaps between the surface of the radome and the antenna. The
suitcase is fabricated from a dielectric that is similar to both the radome and the
antenna substrate to improve signal transmission by limiting the mismatch
between the components. A photograph of the antenna is shown in Fig. 3.6. The
antenna is placed next to a Canadian penny for a sense of scale.
Figure 3.6: A photograph of the fabricated antenna.
S11 measurements for the antenna embedded in the radome were recorded
and are presented in Fig. 3.7. This plot represents the amount of power which is
being reflected, at a given frequency, by the antenna. The antenna performance
has been optimized for frequencies greater than 3 GHz. From approximately 3 12 GHz, the S11 of the antenna is below -10 dB, signifying that more than 90% of
the incident energy is transmitted. At frequencies below 2 GHz, antenna
31 | P a g e
performance is compromised, S11 > -5 dB, the antenna is reflecting the majority
of the power instead.
Figure 3.7: Antenna performance plot.
The oscilloscope (Pico Technology, [42], PC Oscilloscope 9201) samples data at
80 GS/sec and digitally records the time-domain signal. This process allows for
easy storage and post-processing of data. The noise floor of the oscilloscope is 2
mV.
The system is operated by sending and collecting the signals at the antenna
locations. The clock drives the impulse generator, which creates an input impulse
train. This signal is then sent to the antenna placed within the appropriate
radome slot. The signal is differentiated and then transmitted by the antenna.
This signal will proceed to scatter, reflect, and be transmitted across several
different interfaces, primarily it will hit, in sequential order, the
antenna/radome, radome/breast phantom, breast phantom/radome, and
32 | P a g e
radome/antenna interfaces respectively. This signal is then recovered by the
receiving antenna and is then fed to the oscilloscope to be recorded. By changing
the tissue properties and the geometry of the breast phantom, as well as
changing the tumour size and location, the received signal will also change
accordingly. Hence, an image of the breast can be produced to pinpoint the
location of a potential tumour.
3.2 Improvement to the Initial Design
The first improvement to the initial experimental system was in reference to
the pulse repetition rate. From the results of [9], operating the clock at 250 MHz
causes an overlap in the received signal for each subsequent pulse. Hence, for
each pulse, the result was distorted by the transmitted signal from the previous
pulse. Changes to the pulse repetition rate was made in [43], reducing the pulse
repetition rate to 25 MHz. From the results presented in [43], reducing the pulse
repetition rate to 25 MHz enabled the received signal from each pulse to be
distinguished.
The focus of this thesis is to improve this experimental system by changing the
input pulse signal. From Figures. 3.3 and 3.7 it is clear that the current input
signal, the impulse created by the impulse generator, is not well matched with
antenna performance. The majority of the input signal is not being transmitted
into the breast, but is reflected by the antenna instead.
This thesis is focused on using a passive microwave device, which can be easily
integrated into the current experimental system design, to reshape this generic
impulse signal into a new signal with the majority of its power focused in a
frequency range to optimize antenna performance. By matching this newly
formed pulse to the antenna performance we are able to increase the signal
which will be transmitted into the breast. This thesis will investigate whether this
new pulse, by virtue of increasing the signal sent into the breast, can improve the
tumour detection capabilities of the experimental system.
33 | P a g e
CHAPTER 4: DEVELOPMENT OF EXPERIMENTAL-MATCHING NUMERICAL
MODELS
Simulations provide an effective means to easily change numerous parameters
that would otherwise require a much greater effort to reproduce in experimental
work. Thus, simulations are used to perform a primary analysis on the effect of
varying certain parameters in the experimental setup, allowing us to obtain some
idea how this variation will affect the performance of the microwave imaging
system. While the results of these simulations can provide useful insight on the
data we expect to record from experimental work, they should not be
considered to be of greater importance than the experimental work itself. The
simulation environment corresponds to an ideal virtual world. In reality, nothing
is perfect; there is variation due to manufacturing processes, reflections can
occur at various interfaces, and outside sources of interference exist.
SEMCAD-X is a commercially available finite-difference time-domain (FDTD)
solver software that is particularly useful for designing, modelling, and solving
3-D EM wave problems [44]. This software is used to build a numerical model to
replicate the experimental system described in [9] (radome, breast phantom
tissues, antenna, and suitcase).
The antennas and the radome had been previously designed using SEMCAD-X,
thus the two models were easily imported into the same simulation
environment. The suitcases are fabricated as rectangular prisms to fit perfectly
into the radome slots with  = 10 and  = 0. The suitcases are hollowed out so
that the antennas would fit perfectly within them; there are no air gaps between
the interior hollow structure of the suitcase and the antenna surface. The
interior of the radome is filled with the hemi-spherical breast phantom. This
breast phantom is made to replicate the breast phantom created in [43]. This
breast phantom consists of a matching-medium layer, a skin layer, and a
homogeneous, hemi-spherical adipose tissue phantom. The adipose tissue is a
34 | P a g e
6.25 cm radius hemi-sphere that is covered by a 2.5 mm thick skin layer.
Between the skin layer and the interior of the radome wall is a 5 mm thick layer
of matching-medium, used to improve signal transmission by eliminating the air
gaps between the radome wall and the breast phantom. A spherical tumour can
be easily added to the simulation scenario by creating a sphere in the
appropriate location within the fatty tissue and simply 'over-writing' this area
with the tumour. Figure 4.1 provides a depiction of these structures modelled
within SEMCAD-X from four different perspectives. The images in the picture
correspond to a cross-sectional view that cuts through the middle of the radome.
With the inclusion of the various breast phantom tissues, along with the radome,
antenna, and the suitcases, the geometry of the simulation scenario matches the
experimental system.
Figure 4.1: Image of the radome (purple), breast phantom (yellow), skin-mimicking tissue (orange),
matching medium (green), and antennas (red) modeled in SEMCAD-X. This image provides a sliced view
through the midpoint of the radome from four different perspectives.
35 | P a g e
The next challenge is to accurately model the dispersive media which make up
the breast phantom within SEMCAD-X so that these various tissues have
dielectric properties which match well with measurement data. The relative
permittivity of the various breast phantom tissues (fat, skin, gland, and tumour)
have all been measured from 500 MHz – 6 GHz [24]. In SEMCAD-X, dispersive
media can follow a 1st order Debye model given by (1). The goal is to determine
the value of the various parameters in the Debye model so that the
corresponding numerical model behaves in a similar manner to the experimental
data.

 −

∞
  = 0 ∞ +    + 1+
;
(1)
0
where ɛ(ω) is the frequency-dependent complex permittivity, 0 is the
permittivity of free space,  is the static dielectric constant, ∞ is the dielectric
constant at infinite frequencies,  the static conductivity,  the angular
frequency and  is a relaxation constant.
An iterative process is used to shift through a wide range of possible values for
the Debye parameters ( , ∞ , ), while using the data from [15] as a starting
point. The parameters which provided the best-fit were chosen based on
minimizing the mean-squared error (MSE), given by (2), between the measured
data and the model.
 =
1


=1( 
−   )2 ;
(2)
where  is the number the number of samples,   is the measured
permittivity, and   is the estimated permittivity, based on the Debye model, at
the specific frequency corresponding to sample . This process was carried out
for each of the four tissues types (fat, skin, gland, and tumour). Table III compiles
the resulting data from this analysis; listing the values to be used in the Debye
36 | P a g e
model, for each of the four tissue types, in order to obtain the best-fit. Table IV
lists the MSE for each of the best-fit models for the four tissue types. The best-fit
procedure results in a maximum mean squared error in relative permittivity
across trials of 1.12 (gland case).
TABLE III: BEST-FIT DEBYE MODEL PARAMETERS FOR VARIOUS BREAST TISSUES
Fat
Skin
Gland
Tumour

12.0
42.0
53.5
61.0
∞
9.0
27.0
40.0
39.0
 [/]
0.2
0.83
0.68
0.79
 []
60
43
40
40
TABLE IV: MEAN-SQUARED ERROR FOR BEST-FIT MODEL FOR VARIOUS BREAST TISSUES
MSE
Fat
Skin
Gland
Tumour
0.11
0.32
1.12
0.58
Figures 4.2 and 4.3 show the resulting best-fit plot, for  , from 0 – 6 GHz, in
comparison with the recorded measurement data.
37 | P a g e
Figure 4.2: Comparison of the best-fit Debye model and the measurement data [24] for the fat and gland
tissues.
Figure 4.3: Comparison of the best-fit Debye model and the measurement data [24] for the skin and
tumour tissues.
38 | P a g e
The final step taken to match the numerical system to the experiment is to
model the impulse. The impulse is recorded directly from the impulse generator.
The measured impulse is plotted in Figure 4.4. This data is imported into
SEMCAD-X and is used directly as the input source signal to the transmitting
antenna. Since the impulse is measured at the oscilloscope after travelling
through one coaxial cable, the pulse fed to the antenna in the simulations
matches exactly that of the pulse at the transmitter in the experiment (i.e., cable
loss is already taken into consideration).
Figure 4.4: Input pulse recorded directly from impulse generator.
39 | P a g e
CHAPTER 5: SAFETY ASSESSMENT WITH 3-D REALISTIC BREAST MODELS
In order to adhere to the existing safety protocol for microwave imaging
technologies it is imperative to monitor the amount of energy absorbed by the
breast. This limitation on the maximum amount of absorbed energy, as denoted
by the SAR value of the specific tissue, in turn limits the maximum amount of
power which can be delivered into the breast. These limitations allow us to
decide the amount of power to be transmitted by the input pulse to the
microwave system. Thus, we can decide to amplify the input signal, in order to
improve detection, such that we are transmitting close the allowed maximum
power.
This section presents results from FDTD simulations carried out in SEMCAD-X.
First, a synopsis about the development of an anatomically accurate,
heterogeneous, 3-D numerical breast model is given. This breast model is then
used in the subsequent safety assessment. There are two specific assessments
carried out: i) the breast is illuminated by a plane wave and ii) the breast is
illuminated by the TWTLTLA antenna from [41] at the antenna locations from [9].
By choosing to use the corresponding antenna and antenna locations from the
experimental system of [9] the results can be used as a guideline to set the
power limits for the experimental system. The analysis in this section is based on
determining the maximum SAR values in the breast tissue. Specifically, we
compare the spatial distribution of peak unaveraged local SAR values, and the
peak 10g-averaged SAR values, in the breast tissue at specific frequencies.
40 | P a g e
5.1 Numerical Model
A numerical 3-D breast model was developed in [33] to closely mimic the
geometry and heterogeneity of real breast tissue. The development of this
model has been specifically tailored to facilitate efficient use in commercial
software such as SEMCAD-X. In this section, a high-level outline of the procedure
which was followed in the development of this breast model is presented.
Additionally, specifics about the implementation of this model within the
SEMCAD-X framework (where all simulations take place) are provided.
The shape and interior geometry of the breast model has been derived from
MRI images of the breast, for patients lying in the prone position. From the MRI
scan a set of 2-D sliced images of the thoracic region are obtained. In order to
reconstruct a 3-D image from this data it is first necessary to segment the breast
region from the background in each of the slices available. This segmentation is a
manual process. The first step is to locate and identify the breast region in the
image. Once the breast has been localized it is necessary to segment the
skin/background interface and the skin/tissue regions. This process will result in
a boundary for the skin region and the tissue region for one specific MRI slice.
Repeating this segmentation procedure for each slice will create a stack of 2-D
images that can be reconstructed to form a 3-D image, hence the exterior of the
breast, the skin region and boundary of the tissue region, has been created and
the model can be imported into SEMCAD-X [33].
Once the interior of the breast region has been segmented from the MRI slices
it can be used to differentiate the various breast tissues based on the pixel
intensity at a given location. By defining a specific range of pixel intensities for
each tissue type, it is possible to identify the type of tissue at a specific pixel, and
then map the appropriate electrical permittivity and conductivity, based on [14].
This results in a 3-D profile of the dielectric parameters that realistically
represent the complex heterogeneous nature of the human breast. As presented
41 | P a g e
in [33], assigning each of these pixels its own unique parameters would result in
a huge number of discrete entities in the model, over 200 000 cells. This results
in a huge number of interfaces within the breast region causing the FDTD solver
to become highly inefficient. Therefore, highly similar regions were grouped
together to decrease the number discrete ‘cells’ which made up the breast
region. In [33], a regression tree analysis is performed to determine the trade-off
between decreasing the number of cells making up the breast tissue region and
the increasing error. Ultimately, it was decided to use a breast model which is
now composed of approximately 500 cells.
An image of the 3-D heterogeneous and anatomically realistic breast model,
modelled within SEMCAD-X, is shown below in Figure 5.1.1. The transparent
exterior regions allows for a view of the numerous cells which make up the
breast region. A perspective and side view of the breast model are presented.
Figure 5.1.1: An image of the 3-D numerical breast model. Two views of the breast model are provided,
with the exterior region made transparent allowing to view the numerous ‘cells’ which make up the
model.
42 | P a g e
The completion of the breast model involves attaching the skin layer over the
breast tissue and attaching the chest wall to the breast region. The skin layer is a
1.5-mm thick layer surrounding the breast, whereas the chest wall consists of the
1.5-mm skin layer, a 3-mm thick fat layer, and a 5-mm muscle layer.
5.2 Plane-Wave Excitation
In [31], we have presented the results of a preliminary safety assessment using
the numerical breast model developed in [33]; this section serves to both
summarize the methodology and results previously reported, as well as providing
additional clarification and information for certain sections. In [31] we chose to
expose the breast to a plane wave at a distinct number of frequencies in the
0.4 – 9 GHz range. We calculate both the local unaveraged SAR values and the
10-g average SAR values at each voxel location within the simulation
environment. The frequencies of choice were determined to adhere to
frequency bands which have been reserved for medical imaging [45].
The simulation environment is completed by placing the breast model,
complete with the skin layer and chest wall, in a matching medium which will
minimize reflections at the skin interface. In this scenario, the matching medium
used is canola oil, a low-loss medium similar to adipose tissue [46]. A cluster of
tumour cells 4-mm in diameter has also been incorporated into the model. The
conductivity of the tumour cells is higher than the surrounding adipose tissue,
thus, the inclusion of these tumour cells allows for the investigation of a
potential ‘hot spot’ occurring in tumour region. The tissues are modelled in
SEMCAD-X as dispersive materials which follow a single-pole Debye model (as
per (1) in Chapter 4). Table V lists the Debye parameters for the various
dispersive media in the simulation scenario.
43 | P a g e
TABLE V: DEBYE PARAMETERS DESCRIBING DISPERSIVE MODEL (REPRODUCED FROM [31])
Canola Oil
Skin
Fat
Muscle
Tumour

2.51
39.8
4.71
54.9
55.1
∞
2.28
15.9
3.12
21.7
6.75
 [/]
0.008
0.831
0.050
0.886
0.790
 []
27.8
13.0
13.0
27.8
10.5
In [31], the plane wave is incident onto the breast model from two distinct
angles, along the z- and along the y-direction. In the first case, the plane wave is
incident laterally, to the breast side, and propagating in the z-direction. In the
second case the plane wave is propagating in the direction opposite to the y-axis,
towards the chest wall, and is referred to be ‘from the top’. Figures 5.2.1 and
Figures 5.2.2 show an image of the simulation scenario for the plane wave
propagating in the y-direction and the z-direction respectively.
Figure 5.2.1: Image of the simulation scenario, with each component labelled, with the plane wave
incident from the top. Reproduced from [31].
44 | P a g e
Figure 5.2.2: Perspective view of the simulation scenario with the plane wave incident from the side.
Table VI, lists the input power to the plane wave at each of the ten selected
frequencies. The input power for the plane wave has been selected to adhere to
the maximum permissible incident energy as described in [8], akin to the
decision made in [28].
45 | P a g e
TABLE VI: INCIDENT POWER SELECTION PER FREQUENCY
2
f [GHz]
PINC [W/m ]
0.434
2.893
0.915
6.100
1.5
10.00
2
10.00
2.45
10.00
3
10.00
4
10.00
5
10.00
5.8
10.00
7
10.00
8
10.00
9
10.00
Results
The local unaveraged SAR values are representative of a worst-case scenario;
the result does not take into account the dispersion of heat amongst the nearby
tissues. It is also impossible to obtain this information in an experimental setting
and it is for this reason that safety standards which set the maximum permissible
SAR values do not include data for the peak local unaveraged SAR parameter (as
noted per Table II). However, the local unaveraged SAR values are of interest as
they provide information of potential ‘hot spots’, regions within the breast tissue
which absorb more energy than surrounding areas, data which would be
46 | P a g e
otherwise lost due to the smearing effect of averaging over a predetermined
amount of tissue.
The peak local unaveraged SAR values, the maximum local unaveraged SAR
value for a given simulation, are computed at each of the select frequencies of
interest and for both the y-directed and z-directed plane waves. A first order
least-squares fit is used to represent the data in order to determine whether
there is a correlation between increasing frequency and the peak unaveraged
local SAR values. Figure 5.2.3 plots the resulting data for both plane waves. The
simulation results are denoted by the discrete data points in the plot, with the
linear best-fit denoted by the solid line. The maximum peak local SAR values are
24W/kg and 17W/kg for the laterally-incident and top-incident plane waves,
respectively, with both occurring when the breast is illuminated with the plane
wave with highest frequency. The first order fit representing the data has r2values of 0.9013 and 0.7225 for the z-directed and y-directed plane wave
respectively; thus, we can conclude that the relationship between the frequency
of the incident plane wave and the amount of energy deposited in the breast can
be well explained by a linear model. Furthermore, there is a correlation between
increasing the frequency of the incident plane wave and the peak local SAR
value.
47 | P a g e
Figure 5.2.3: Peak local unaveraged SAR values. Simulation results (denoted by ‘o’ and ‘x’ markers) and
corresponding linear fit for both z-directed and y-directed plane waves. Reproduced from [31].
A cross-sectional profile of the local unaveraged SAR values within the
simulation environment, for both the y-directed and the z-directed plane wave,
is shown below in Figures 5.2.4 and 5.2.5 respectively. These images are
extracted from the 3-D simulation environment to contain the location of the
peak local unaveraged SAR value at a specific frequency. From these crosssectional profiles we can observe the location of the peak unaveraged SAR value,
as well as other regions which absorb the majority of the incident EM wave. For
the y-directed plane wave the peak SAR values occur in the skin area near the
nipple as well as within the muscle layer of the chest wall. When the plane wave
is propagating along the z-axis the peak SAR values occur at the skin-matching
medium interface, as well as near the skin-chest wall interface.
48 | P a g e
Figure 5.2.4: Cross-sectional profile of the unaveraged SAR values for a plane wave propagating along the
y-axis at 8 GHz.
Figure 5.2.5: Cross-sectional profile of the unaveraged SAR values for a plane wave propagating along the
z-axis at 8 GHz.
49 | P a g e
In order to determine whether the plane wave illumination of the breast
adheres to the safety limits established in Table II a peak 10-g averaged SAR
value is computed at each frequency for both plane waves. The results of these
simulations are presented in Figure 5.2.6. In Figure 5.2.6 the simulation data is
denoted by markers at the corresponding frequency of illumination. Once again,
a first order least-squares model is used to fit the data. In this case, a linear fit
does not provide a good representation of the data, hence, we do not attempt to
infer any conclusion about the relationship about the peak 10-g averaged SAR
values and increasing frequency.
The maximum 10-g average SAR values of 1.7W/kg and 1.5W/kg for both the
laterally-incident and top-incident plane waves, respectively, is below the 4W/kg
limits of Table II, implying that the method is indeed safe for a plane wave
directed along either the y-axis or the z-axis at the specific frequencies and input
powers investigated.
At very low frequencies, < 1 GHz, the SAR values, for both the unaveraged and
10-g averaged, are lower than would be predicted by the linear regression
analysis. This can be explained by the fact that the exposure limits (i.e. the
maximum input power), as denoted by Table VI, at these frequencies are much
more stringent than at higher frequencies [3]. Hence, since there is a decrease in
the amount of incident power to the breast, there is a decrease in the amount of
absorbed energy.
50 | P a g e
Figure 5.2.6: Peak 10-g averaged SAR values. Simulation results (denoted by ‘o’ and ‘x’ markers) and
corresponding linear fit for both z-directed and y-directed plane waves. Reproduced from [31].
5.3 Antenna Excitation
In this section we explore and expand on the results and methodology initially
presented [47]; in which the breast was illuminated by the broadband antenna
designed in [41], with the antenna placed at locations to match the antenna
positioning suggested in the experimental setup of [9]. In this analysis, the
antenna is placed at four specific locations, while the breast is illuminated by
nine possible frequencies. These nine frequencies are chosen to represent three
distinct pulses (the centre frequency and both the high and low -3 dB
frequencies). Again, the unaveraged local SAR values are calculated along with
the 10-g averaged SAR values.
The simulation environment is completed by embedding the breast model,
with the skin layer and chest wall layer, in a matching medium to improve signal
transmission. The choice of the matching medium was made based on the
51 | P a g e
antenna performance. Simulations were done with the antenna embedded in
three different background mediums: air, canola oil (as per [31]), and a fatmimicking matching medium (similar to the experimental system of [43]). Figure
5.3.1 shows an S11 plot of the antenna embedded in each of the three
background media. At frequencies below 10 GHz the antenna performance is
best when it is embedded in the fat-mimicking material; hence, we chose to use
this material as the matching material for the subsequent simulations [47].
Furthermore, it is clear from this plot (comparing performance in air and in the
fat-mimicking material) that the choice of matching medium plays a large role in
the performance of this antenna.
As mentioned previously, the various breast tissues, and tissue of the chest
wall, along with the matching material are modeled in SEMCAD-X as dispersive
media following a single-pole Debye model.
Figure 5.3.1: A plot of the S11 of the antenna embedded in three separate matching mediums. Reproduced
from [47].
In [30], a safety assessment was carried out in which a breast was illuminated
by an antenna which had been excited by an incident broadband pulse.
52 | P a g e
However, in this study, specific absorption (SA) values are computed to indicate
the amount of energy absorbed by the breast tissue (in Joules) when exposed to
all the frequencies which make up the incident pulse. Unfortunately, due the
highly complex breast model, and the subsequently fine mesh required to
resolve the delicate structures of the antenna, it is impossible to compute the SA
values when using the advanced breast model of [33] in the SEMCAD-X software.
The SA computation requires a recording of the E- and H- fields at each voxel for
every frequency of the pulse, the resulting data file is inexplicably huge (> 500
GB) and attempting to process it often causes the simulation software to fail.
Thus, due to computational limitations the simulations of [47] were conducting
by exciting the antenna with a single frequency and then computing the resulting
SAR values.
The frequencies have been specifically chosen to represent three pulses;
specifically, we are interested in transmitting the centre frequency of each pulse
as well as both the high and low -3 dB frequencies. The first pulse is a Gaussian
modulated pulse, similar to the one used in [30], with a centre frequency of 6.85
GHz. The second pulse is a band-limited pulse with the majority of its power in
the 2 ‒ 4 GHz range. The pulse has been created experimentally by making use of
the pulse shaping technology outlined in [36]. The third pulse matches the pulse
used in the experimental setup of [9].
A time-domain plot of the three pulses is shown below in Figure 5.3.2. Each of
the three plots has been normalized to the peak amplitude of each of the
respective pulses (i.e. they are all of differing input power). The Gaussian plot is
an ideal numerical representation; it has not been reproduced experimentally.
The spectral content of each of the three pulses is shown in Figure 5.3.3 for
comparison purposes.
53 | P a g e
Figure 5.3.2: Time domain representation of each of the three pulse shapes. The pulses have been
normalized to their respective maximum and time-shifted so they can be plotted on the same graph.
Figure 5.3.3: Spectral content of the three pulses analyzed. Reproduced from [47].
54 | P a g e
The complete simulation environment is depicted in Figures 5.3.4. and 5.3.5,
where a side view and front view of the breast model is shown. Figure 5.3.4
serves to illustrate the location and orientation of the antennas in the four
possible locations, whereas Figure 5.3.5 provides an image which clearly discerns
the various breast tissues from each other.
Figure 5.3.4: Side view of the simulation environment showing the four possible antenna locations.
Antennas in position 2 and 4 are cross-polarized in comparison to antennas placed in position 1 and 3.
Reproduced from [47].
55 | P a g e
Figure 5.3.5: Front view of the simulation environment illustrating the various tissues of the breast model.
Note that the breast is in front of the antenna from this viewing angle; the antenna is at no point in
contact with the skin. Reproduced from [47].
The four antenna locations have been chosen to specifically match the
locations of the antenna from the experimental system of [9]. The antennas are
arranged to ensure that they remain, at a minimum, 2.5mm away from the
surface of the breast. The choice of these locations allows us to match the results
of the safety assessment with an experimental system, thus, we are provided
with a guideline for the maximum power to be used in the experimental system.
Antennas placed in the odd positions, antennas in Position 1 and 3, have the
same orientation. These two antennas are referred to as co-polarized; they are
oriented parallel to the x-axis. The antennas in Position 2 and 4 are also in the
same orientation; however, these antennas are cross-polarized when compared
to the antennas in the odd positions. Antennas in Positions 2 and 4 are oriented
with perpendicular to the x-axis. They are referred to as cross-polarized.
Results
The results of [47], as well as some of the unreported results from the FDTD
simulations, are discussed in this section. In particular, the focus is on the peak
56 | P a g e
unaveraged local SAR values and the 10-g averaged SAR values, similar to results
presented in Section 5.2.
Maximum peak unaveraged local SAR values are presented in Table VII for
each of the four antenna positions and for each of the three pulses used to
excite the antenna. For each of the four antenna positions, exciting the antenna
with the highest possible frequency, 8.9 GHz (-3 dBH for the Gaussian pulse)
resulted in the highest peak unaveraged local SAR value. Indicating that at these
high frequencies energy absorption is maximized. Further, evidence for this
trend can be noticed when looking at the results for the other two pulses. In all
but one case (Generic Impulse, Antenna Position 2) the maximum energy
absorption occurs at the -3 dBH frequency at each antenna position. This trend is
in agreement with the results of Figure 5.2.2 [31], where a correlation between
increasing frequency and peak unaveraged local SAR values was demonstrated.
Additionally, for both the Band-Limited Pulse and the Gaussian pulse the peak
SAR value at antenna Position 4 is much greater than the other antenna
positions.
TABLE VII: PEAK UNAVERAGED SAR IN MW/KG, CORRESPONDING EXCITATION FREQUENCY IN
PARENTHESIS
Antenna
Generic Impulse
[mW/kg]
Band-Limited Pulse
[mW/kg]
Gaussian Pulse
[mW/kg]
17.2 (1.5 GHz)
76.1 (3.89 GHz)
177 (8.9 GHz)
27.1 (39 MHz)
29.4 (3.89 GHz)
171 (8.9 GHz)
21.0 (1.5 GHz)
65.4 (3.89 GHz)
248 (8.9 GHz)
20.2 (1.5 GHz)
140 (3.89 GHz)
643 (8.9 GHz)
Position 1
Antenna
Position 2
Antenna
Position 3
Antenna
Position 4
57 | P a g e
A cross-sectional profile of the unaveraged SAR values is shown in Figure
5.3.6(A)-(D), presenting a spatial distribution of the SAR values within the breast
tissue. Each of these images are created by taking a cross-section through the
breast tissue of the plane which contains the peak unaveraged local SAR value,
hence they do not correspond to the same location in space (this explains why
the breast shape is not exactly the same in each image). Each of the figures, (A) (D), corresponds to a specific scenario when the antenna is located at each of the
Positions 1 through 4. For example, Figure 5.3.6(A) corresponds to the scenario
when the breast is illuminated by the antenna in Position 1, Figure 5.3.6(B)
corresponds to illumination of the breast when the antenna is in Position 2, and
so forth. In each of the Figures the frequency of excitation is 8.9 GHz.
The unaveraged local SAR values allows us to get a better understanding of the
locations where the majority of the energy is absorbed. As expected, the skin
regions closest to the radiating antenna absorb the most energy. This is evident
in Figure 5.3.6 where the peak SAR values occur in the skin region closest to the
radiating antenna; as the antenna are shifted through Positions 1 through 4 they
move towards the nipple region, consequently we observe that the bright spots
in the image, denoting peak SAR values, move along with the antenna.
It is interesting to note the regions of high absorbance in the chest wall region
directly below the antenna. As expected, attenuation through the breast
structure greatly diminishes signal strength at regions far from the radiating
antenna, subsequently, these regions of the breast experience minimal heating.
58 | P a g e
A)
B)
59 | P a g e
C)
60 | P a g e
D)
Figure 5.3.6: (A)-(D) Show the distribution of unaveraged SAR values for illumination of the breast at 8.9
GHz with the antenna located in positions 1 through 4 respectively. Each cross-section contains the peak
unaveraged SAR value for each respective scenario. The scale is noted in dB relative to this local peak SAR
value. Note how the antenna position changes the location of the local peak as well as the distribution of
the absorbed energy within the breast. A sample image of the antenna is shown to indicate its
approximate location.
Table VIII compiles the data in regards to the peak 10-g averaged SAR values at
each antenna position for each of the three incident pulses. The results have
been normalized for an input power of 1 mW. The maximum peak 10-g averaged
SAR value is 10.0 mW/kg, approximately 400 times below the limit established in
Table II. Thus, setting the incident power to 400 mW would represent a safe
assumption which adheres to the safety standards established by the IEEE.
Similar to the peak unaveraged SAR values, for each of the four antenna
positions, the maximum 10-g averaged SAR values occur for the highest
excitation frequency. The 10-g average values, much like the peak unaveraged
SAR values, are higher for the antennas located in positions 3 and 4. A high
portion of glandular tissue is located near the nipple region of the breast. The
61 | P a g e
high conductivity of the glandular tissue may cause additional energy to be
absorbed when compared to the surrounding fatty tissue, which has a much
lower conductivity. Since the glandular tissue is located near the skin surface it
causes the average SAR values near the nipple region to be higher. Therefore,
when the antenna is in positions 3 and 4, and the peak SAR values are observed
on the skin surface near the nipple region, we can expect the 10-g average to be
increased when compared to positions 1 and 2 when the peak SAR values are on
the breast surface closer to the chest wall.
TABLE VIII: PEAK 10-G AVERAGED SAR IN MW/KG, CORRESPONDING EXCITATION FREQUENCY IN
PARENTHESIS
Antenna
Generic Impulse
Band-Limited Pulse
Gaussian Pulse
[mW/kg]
[mW/kg]
[mW/kg]
0.69 (1.5 GHz)
0.683 (2.38 GHz)
4.66 (8.9 GHz)
1.11 (39 MHz)
0.488 (2.38 GHz)
4.64 (8.9 GHz)
0.97 (1.5 GHz)
3.61 (3.89 GHz)
6.02 (8.9 GHz)
4.08 (39 MHz)
2.038 (1.72 GHz)
10.0 (8.9 GHz)
Position 1
Antenna
Position 2
Antenna
Position 3
Antenna
Position 4
62 | P a g e
5.4 Conclusions
Based on the peak unaveraged local SAR values we have observed correlation
between increasing frequency and an increase in the amount of energy absorbed
by the breast tissue. This observation was true for illumination of the breast by
both the plane wave and the antenna. We can expect that the breast will absorb
more energy when it is illuminated by higher frequency radiation.
The relatively high conductivity of the skin layer, and its proximity to the
radiating antenna, will cause it to absorb a significant percentage of the incident
energy. The peak unaveraged local SAR values occur in the skin region,
regardless of the excitation frequency and the antenna position. Furthermore,
the location of this peak SAR value is dependent on the location of the radiating
antenna. The peak SAR value occurs in the skin region nearest the radiating
antenna. This observation can easily be explained by the fact that regions further
from the antenna are exposed to a weaker EM signal; the EM wave is attenuated
as it passes through the various breast tissues. This phenomenon is made clear in
Figure 5.3.6, where the regions furthest from the radiating antenna experience
minimal heating.
The 10-g averaged SAR values are used to determine the safety of the imaging
technique by comparing the computed peak values with the maximum limits
established in Table II. For illumination of the breast with the plane wave the
maximum recorded 10-g averaged SAR value was found to be 1.7 W/kg, which
falls well below the maximum of 4.0 W/kg. The results for excitation by the
broadband antenna have been normalized to 1mW of incident power. We found
that the maximum recorded 10-g averaged SAR value in this case, occurring at an
excitation frequency of 8.9 GHz, to be below the maximum limit established by
the IEEE by a factor of 400. Since this study investigated the use of antennas that
are used in an existing imaging system, and placed at the same location from the
breast, it is possible to extrapolate the maximum power (400 mW) of the
incident pulse which adheres to current safety limits.
63 | P a g e
CHAPTER 6: INTEGRATION OF PULSE SHAPING TECHNOLOGY
Based on previous experimental work, [9], [37], we decided that an easily
implementable change to the system was to use a different input signal. We
sought to choose a new pulse shape which improved signal transmission in the
breast by making better use of the antenna efficiency at specific frequencies. It
was decided to implement this new pulse by reshaping the previously generated
impulse signal, in lieu of buying additional software and hardware to create
arbitrary broadband pulses, by making use of the microstrip line technology
described in [34]-[36] in which a transmission line with a specific conductive
trace profile can be used to reshape a specific input into a desired target pulse
shape. We refer to this fabricated transmission line as a synthesized broadband
reflector (SBR).
This section is focused on describing the various components that must be
integrated with the existing experimental system in order to successfully
implement the SBR device, such that the newly shaped pulse matches the
desired target shape. The performance of several of the components are
presented, we categorize changes to the pulse shape (in both the time and
frequency domain) along each intermittent step, and justify why certain pieces
of hardware have been included in the system.
6.1 Pulse-Shaping Circuit
The components for the pulse shaping circuit include the impulse generator, a
4-port directional coupler, the SBR terminated with a matched broadband 50 Ω
load, and the broadband amplifier. A high-level overview of this pulse shaping
scheme, including the various components and the overall signal flow, is
illustrated below in Figure 6.1.1.
64 | P a g e
Figure 6.1.1: An illustration of the pulse shaping circuit and the subsequent components used to achieve
successful integration into the experimental system.
6.2 The SBR Structure
The SBR is implemented in microstrip technology, in which a single conductive
trace runs on a dielectric substrate directly above a ground plane. It is the width
profile of this conductive trace that is directly related to its frequency response.
The methods of [34]-[36] dictate the shape of the microstrip line required to
transform the input signal to the desired target output. The SBR is fabricated
such that the reshaped pulse has the majority of its power concentrated in the 2
– 4 GHz range. We have chosen this frequency band for two reasons: i) at
frequencies above 4 GHz the EM wave experiences increased attenuation when
propagating through the breast tissues, enough to severely dampen the signal
strength, and ii) below 2 GHz antenna performance is compromised (S11 > -5 dB
as per Figure 3.7), limiting the amount of power that is transmitted into the
breast tissue. A plot of the target pulse shape is shown in the time domain and
the frequency domain below in Figures 6.2.1 and 6.2.2, respectively.
65 | P a g e
Figure 6.2.1: Time domain plot of the target pulse shape.
Figure 6.2.2: Spectral content of the target pulse. The majority of the frequency content is in the 2 - 4 GHz
range, with suppression of the outside frequencies.
66 | P a g e
The SBR is a reflection mode device; the input and output are at the same
terminal of the device. The frequencies which make up the target pulse shape
are reflected by the SBR, it is this reflected signal which will be used to excite the
antenna. The fabricated SBR was tested using a VNA (Agilent 8703B Lightwave
Component Analyzer) to measure the S11 profile. The S11 measurement
corresponds to the spectral distribution of the power in the reflected signal;
hence, we expect the measurement to be similar to the frequency content of the
target pulse. A plot of the S11 measurement is shown in Figure 6.2.3, and, as
expected, the SBR reflects nearly all signal content in the 2 – 4 GHz range.
Figure 6.2.3: S11 of fabricated SBR. This plot shows the frequency content which is reflected by the device.
Finally, a photograph of the fabricated SBR is shown below in Figure 6.2.4. The
device is compact, measuring 2.9 cm wide, 14.3 cm long, and 1.9 cm thick
(limited by SMA connector). The 'end' terminal of the device (right of image) is
designed to be terminated with a 50 Ω broadband load to match the resistance
67 | P a g e
of the conductive strip at the terminal and reduce any possible reflections from
the termination.
Figure 6.2.4: Photograph of the fabricated SBR. The dimensions of the device are 2.9cm x 14.3 cm.
6.3 Signal Routing
A directive device is implemented to recover the signal reflected by the SBR,
which contains the reshaped pulse signal, and route it back towards the
transmitting antenna.
Two such devices were tested: i) a 3-port circulator with an operating
frequency of 2 - 4 GHz and 3 dB insertion loss, and ii) a broadband 4-port
directional coupler (Pulsar CS06-09 436/9) with an operating frequency of 2 - 8
GHz, 6 dB coupling loss (between the two coupled channels) and 0.5 dB insertion
loss.
To compare the performance of the two structures a VNA was used to
measure the signal leaving the directional device (i.e. the signal passes through
the device, reflects off the SBR, and this reflected signal is then redirected to the
output by the directive device). In Figure 6.3.1, we compare the spectral profile
of the synthesized pulse with and without (same as figure 6.2.4) passing through
each of the respective devices. These measurements indicate that the 3-port
circulator is not a viable component to be used in the experimental system. From
the plot we notice that the low and high frequencies are not properly
attenuated; the spectral content of the reshaped pulse has changed. However,
68 | P a g e
using the directional coupler, there is minimal change to the spectral profile of
the pulse. Hence, we have chosen to use the directional coupler as part of the
signal routing component of the new experimental system. The directional
coupler introduces additional loss to the system, as evident by from Figure 6.3.1
where the S11 plot for the directional coupler has been shifted down. A
broadband amplifier is used to offset the additional attenuation (approximately
10 dB).
Figure 6.3.1: Comparison of the frequency content of the signal reflected by the SBR after it has passed
through the signal routing components (circulator and coupler). A plot of the S11 for the fabricated SBR is
also included to provide a comparison.
6.4 Pulse Analysis
In this section, we compare the shape of the newly formed pulse after it has
passed through the pulse shaping circuit (the SBR and directive device) and the
impulse created by the impulse generator with the target pulse. This comparison
is made in both the time and frequency domain.
69 | P a g e
The initial pulse, 70 ps at full-width half maximum and -7.5 V amplitude, is
created by the impulse generator (Picosecond Pulse Labs Impulse Generator
Model 3600). This signal is fed into the directional coupler, routed to the SBR,
and the reflected signal, containing the reshaped pulse, is recovered and
redirected to the transmitting antenna.
An oscilloscope (Pico Technology, [42], PC Oscilloscope 9201) is used to
perform the time domain measurements. Data is sampled at an equivalent of 80
GS/sec. The noise floor of the device is 2 mV. The reshaped pulse is recorded
right before it is sent to the transmitting antenna (i.e. after being redirected by
the directional coupler).
A comparison of the time domain plots of the initially generated impulse and
the SBR-shaped pulse is shown in Figure 6.4.1. The shape of the pulse has been
completely transformed; what was once a DC-centric narrow pulse has been
reshaped into a purely AC signal. The amplitude of the pulse has been decreased
in the shaping procedure, due to losses from the directional coupler and the SBR
itself. However, since the pulse now contains only the desired spectrum we can
amplify the signal before it is transmitted into the breast. Such a process would
not be helpful with the generic impulse since we would be increasing the power
only to have it reflected or lost at the transmit antenna for being mostly in the
wrong frequency range.
70 | P a g e
Figure 6.4.1: Time domain comparison of the initial pulse (blue) and the newly reshaped pulse (orange).
The pulse shaping circuit has been successfully implemented; the reshaped
pulse meets our target spectral profile. A comparison of the frequency content
of the reshaped pulse, our desired target spectral profile, as well as the spectral
content of the generic impulse has been plotted in Figure 6.4.2. As desired, the
majority of the signal content has now been shifted into the 2 – 4 GHz range,
while the signal content outside this frequency range is strongly suppressed.
Additionally, we can see how the generic impulse is not well matched for the
antenna performance. The majority of its power resides below 2 GHz, with
additional content above 4 GHz.
71 | P a g e
Figure 6.4.2: Spectral profile of the generic impulse and the reshaped pulse. The target spectral shape is
included to show the successful implementation of the pulse shaping technology.
6.5 Amplification
The loss from the pulse shaping circuit is offset by the use of a broadband
amplifier (Mini-Circuits ZVE-3W-83+) which operates in the 2 – 8 GHz range,
provides a typical gain of 35 dB, and has a maximum output of 33 dBm. We have
chosen to use this amplifier based on several factors: i) the operating frequency
corresponds well with the frequency content of the reshaped pulse, ii) it
provides a fair amount of gain, iii) the large maximum output power is very
favourable as it allows us to maximize the signal to be transmitted by the
antenna, and iv) it is a discrete and cost-efficient component that is easily
integrated with the experimental system.
The newly formed pulse has a peak power of 6 dBm, hence, it is necessary to
attenuate the signal before amplification otherwise we risk entering the non-
72 | P a g e
linear range of operation of the amplifier. Therefore, we include an additional 9
dB of attenuation prior to amplification.
The broadband amplifier is used for the sole purpose of increasing the signal
amplitude; it should not have any effect on the pulse shape by introducing any
unwanted frequencies or filtering out any of the signal content in the 2 – 4 GHz
range. A comparison of the frequency content of the pulse prior to and after
amplification is depicted in Figure 6.5.1. Additionally, Figure 6.5.2 compares the
time domain plots of the reshaped pulse before and after amplification. The
pulse has changed from a 1.2 V peak-to-peak signal to a signal that is now nearly
16 V peak-to-peak. These measurements demonstrate that the broadband
amplifier chosen is well suited for the task of amplifying the reshaped pulse
whilst maintain the pulse shape and spectral profile.
Figure 6.5.1: Frequency domain comparison of the pulse prior to and post amplification.
73 | P a g e
Figure 6.5.2: Amplification of the reshaped pulse from a 1.2 V peak-to-peak signal to a 16 V peak-to-peak
signal.
74 | P a g e
CHAPTER 7: INTEGRATED EXPERIMENTAL SYSTEM DESIGN
In this chapter we provide an overview of the new augmented experimental
system design and a description of the interconnections of the various system
components (which have been previously described in Chapters 3 and 6).
Additionally, we provide details on the measurement procedure; a complete
description, and illustrations, of the four antenna arrangements tested, and the
four breast phantoms used for this analysis, is presented.
The system is driven by the Tektronix clock operating at 25 MHz. This 25 MHz
signal is used to drive both the impulse generator, thus setting the pulse
repetition rate, and to trigger the oscilloscope. The generated impulse is then
sent to the pulse shaping circuit. The signal is routed by the directional coupler,
reflected off of the SBR, and this reflected signal, which is the reshaped pulse, is
re-routed towards the transmit antenna. Before being transmitted, the pulse is
first amplified by a broadband amplifier. The amplifier is powered by a 15 V
DC-power source. Prior to amplification the signal must first be attenuated by
9 dB to ensure that the amplifier is operating in the linear range. After
amplification the signal is then sent to the transmitting antenna which is placed
inside a specific slot of the radome. The signal then propagates through the
radome walls, through the breast phantom, and finally recovered by the
receiving antenna (also placed in a slot of the radome). Depending on the
antenna configuration, we can recover the reflected signal (antennas on the
same side of the radome) or the transmitted signal (antennas on opposite ends
of the radome). The recovered signal is then fed into the oscilloscope and
recorded digitally for processing. A block system diagram of the augmented
system, depicting the various system components and their interconnections, is
illustrated in Figure 7.1. Note that in this diagram the antennas are arranged
such that the recorded signal is representative of the transmission through the
breast phantom.
75 | P a g e
Figure 7.1: A high-level depiction of the complete experimental setup. The setup includes a clock to drive
the pulse generator, a directional coupler to route the signal from the SBR structure towards the antenna,
a radome to house the antennas and the breast phantom, and an oscilloscope to record the time-domain
data. Note that the two antennas are placed within the slots of the radome. In this case the antennas are
arranged such that the transmitted signal will be recorded. To record the reflected signal the antennas
are placed on the same side of the radome.
Each measurement conducted requires a pair of antennas. Each antenna is
placed into one of the 16 antenna slots of the radome. These 16 slots are
organized in a 4 x 4 grid as described in Chapter 3 (see Figures 3.4 and 3.5 for
radome schematic). Recall that in two of the quadrants alternating slots are
cross-polarized (perpendicular orientation with respect to co-polarized slots).
From previous studies , [9], [43], we have observed that there is a significant
difference in the performance of the system when analysing the transmission
(antennas on opposite sides of the radome) scenarios when compared to the
reflective (antennas on same side of the radome) scenarios; the tumour is much
easier to detect for the reflective cases. In this thesis we focus on four specific
antenna arrangements, referred to as Cases 1 through 4. Each of the four cases
are representative of the reflective scenario. In Table IX a top-view illustration,
and a description of the antenna arrangement, of each of the four cases is
76 | P a g e
shown. We investigate two co-polarized arrangements (Case 1 and Case 3) and
two cross-polarized arrangements (Case 2 and Case 4). The location of the
transmitting and receiving antenna is outlined in the Case Description column of
Table IX. The tumour is located half-way between the centre of the phantom and
the radome wall in all four cases.
77 | P a g e
TABLE IX: ILLUSTRATION AND DESCRIPTION OF ANTENNA ARRANGEMENTS
Top-View Illustration
Case Description
Case 1: The two antennas are co-polarized. The
transmitting antenna is placed in the second slot from
the chest wall. The receiving antenna is placed in the
third slot from the chest wall.
Case 2: The two antennas are cross-polarized
(perpendicularly oriented with respected to each
other). The transmitting antenna is placed in third slot
from the chest wall. The receiving antenna is placed
in the second slot from the chest wall. It is the
receiving antenna which is rotated by 90° with
respect to the flat surface of the radome (i.e. this
antenna is rotated with respect to the antennas in
Case 1).
Case 3: The two antennas are co-polarized. The
transmitting antenna is placed in the slot closes to the
chest wall. The receiving antenna is placed in the
second slot from the chest wall.
Case 4: The two antennas are cross-polarized.
The transmitting antenna is placed in second slot
from the chest wall. The receiving antenna is placed
in the third slot from the chest wall. It is the
transmitting antenna which is rotated by 90° with
respect to the flat surface of the radome (i.e. this
antenna is rotated with respect to the antennas in
Case 1).
78 | P a g e
This thesis is focused on analysing the tumour detection improvements of the
augmented system when using these four antenna arrangements tested with
four different breast phantoms. The four breast phantoms are chosen to
represent differing levels of difficulty for tumour detection. The first, and most
simple, breast phantom tested was a 100% fat phantom; the breast phantom is
constructed solely of adipose-mimicking tissue. There is no skin layer or
glandular structures. The second phantom is made up of a 2-mm skin layer
surrounding a homogeneous, 100% fat phantom. Both the third and fourth
breast phantoms are no longer homogeneous; discrete quantities of glandular
cones have been added to the phantom. The third phantom is made up of 50%
glandular content, by volume, whereas the fourth phantom is made up of 70%
glandular content. Each of these phantoms also has a 2-mm thick skin layer that
surrounds the inhomogeneous phantom. Details on breast phantom
construction are provided in Chapter 2.4.
Specifically, these different breast phantoms allow us to investigate the
following questions: i) is tumour detection easiest in a purely adipose
homogeneous breast phantom? ii) does the inclusion of a skin layer negatively
affect tumour detection? iii) what role does an increasing amount of glandular
content play in regards to tumour detection?
79 | P a g e
CHAPTER 8: EXPERIMENTAL RESULTS WITH AUGMENTED SYSTEM
In this chapter we compare the results when using the augmented system,
presented in Chapter 7, in comparison to the experimental system used in [43]
(improved version of the experimental system first developed in [9]), in order to
assess whether the augmented system with the SBR-shaped pulse improves
tumour detection. Time-domain measurements are carried out for both
experimental setups on each of the four breast phantoms from Chapter 7. For
each system, these measurements are carried out on the same breast phantom
(on the same day and with the same antenna) to eliminate any potential bias.
For each of the cases (outlined in Chapter 7) two signals are recorded. The first
measurement is a baseline signal, representative of the healthy breast. For this
measurement there is no tumour; the radome is completely filled with the breast
phantom. For the second measurement a cylindrical tumour phantom is inserted
into the breast phantom. The tumour is always placed half-way between the
phantom centre and the radome wall, approximately 4 cm from the radiating
edge of the antenna. The cylindrical tumour is 3 cm in length with a radius of 0.5
cm.
In order to assess the detection of the tumour several parameters are used;
notably, we are interested in the tumour response signal, the tumour response
signal relative to the input signal, and the tumour response signal relative to the
baseline signal.
The tumour response signal is defined as,
. . =   −  ;
(2)
where T.R. is the tumour response signal. The tumour response signal is the
difference between the baseline signal and the signal recorded when the tumour
is inserted into the phantom.
80 | P a g e
We define the relative tumour response parameter as
 = 20 log
   
  
;
(3)
where  is the relative tumour response parameter. This metric provides
information about the relative signal strength of the tumour response signal
when compared to the input signal to the transmitting antenna. This parameter
is particularly interesting since the two systems, with and without the integration
of the SBR, have input signals of unmatched power.
It is possible that the increase in the tumour response signal may not be
indicative of an easier detection scenario. Since signal transmission is improved
when implementing the SBR-shaped pulse, and thus more power is sent into the
breast, it is possible that both the baseline signal and the tumour response signal
have been increased by the same proportion. Thus we present a metric, the
tumour response to baseline ratio, defined as follows,
/ = 20 log
   
  
;
(4)
where / is the tumour response relative to the baseline signal. The tumour
response to baseline ratio allows us to assess whether there the increase in the
baseline signal is outweighed by gains in the tumour response signal.
The chapter is broken down into five subsections; each of the first four
subsections is devoted to an analysis of the measurement results when testing
each specific breast phantom, respectively. The final section is used to provide
an overall assessment of the improvement from incorporating the SBR with the
experimental system.
81 | P a g e
8.1 Homogeneous Adipose Breast Phantom
A comparison of the recorded baseline signal, for Case 1, when illuminating
the breast with both the generic impulse (original system) and the reshaped
pulse is shown in Figure 8.1.1. This signal is measured at the receiving antenna
after the microwave energy from the transmitting antenna has scattered and
reflected off of the fat phantom. The new system design has greatly improved
the recorded signal strength, increasing the peak-to-peak voltage of the received
signal from approximately 0.3 V to approximately 0.8 V. Furthermore, the new
system retains the original pulse shape after transmission through the dispersive
and lossy breast tissue, in contrast to the original system which introduced
distortion.
Figure 8.1.1: Time-domain measurement of the received signal for the baseline recording of Case 1. The
plot compares the recorded signal when the breast is illuminated by both the generic impulse and the
reshaped pulse when using the SBR.
82 | P a g e
In Figure 8.1.2 we compare the frequency content of the two baseline signals.
These plots prove our assertion that the illumination of the breast phantom with
a generic impulse will introduce various frequencies not initially present in the
incident signal.
The transmission of the input signal, with the original setup, through the
dispersive breast phantom shifts a large portion of the signal power from DC to
higher frequency. The breast phantom itself can be seen to act as a filter. Based
on Figure 8.1.2, and with knowledge of the spectral content of the respective
input signals, we can make an estimate on what the expected frequency
response of the breast phantom would be. It appears that the homogeneous,
purely adipose, breast phantom acts as a passband filter with two separate pass
bands, one centered around 500 MHz and the other in the 2 ‒ 4 GHz range.
Using the newly designed system setup, the frequency content of the
transmitted signal is remains in the 2 ‒ 4 GHz range, matching closely to our
desired pulse shape. A result we expected based on the time-domain plot in
Figure 8.1.1 where the transmitted signal maintained the incident pulse shape.
A comparison of the tumour response signal, for Case 1, for both experimental
systems is shown in Figure 8.1.3. Whereas the tumour response using the
generic impulse seems to be spread out in time (including what appears to be a
“ringing” tail at the end of the signal), the response using the SBR structure
remains a nicely compact signal. Additionally, there is an improved response
when using the SBR-formed pulse.
83 | P a g e
Figure 8.1.2: Frequency content of the recorded baseline signals for Case 1 using both the generic impulse
and newly formed pulse to excite the antenna.
Figure 8.1.3: A plot comparing the tumour response signal, in the time domain, for Case 1, with both
experimental systems.
84 | P a g e
In Figure 8.1.4, we compare the frequency profile of the tumour response for
both setups. For the original system, the tumour response is primarily centered
in the low frequency range, specifically near 1 GHz. Additionally, we note that
there is a jump in the frequency content in the 2 ‒ 5 GHz range as well. This high
percentage of power in the 1 GHz range is apparent in the time domain plot in
Fig. 8.1.3 as it is the cause of the "ringing tail" effect we observe. The presence of
this low frequency content is potentially problematic as it is a limiting factor for
the pulse repetition rate. It becomes necessary to reduce the pulse rate such
that consecutive pulses will not cause the tumour response to overlap. As
expected, based on the time-domain observations, there is minimal distortion
using the SBR-system. The most significant percentage of the power remains in
the 2 ‒ 4 GHz range.
The spectral content of the tumour response signals plotted in Figure 8.1.4
have been normalized to the peak power of the tumour response signal, thus the
noise floor of the system, governed by the 2 mV noise floor of the oscilloscope,
can be expressed in dB relative to this peak power. The noise floor of the system
corresponds to approximately -20 dB and -23 dB for both the original system and
the SBR-system respectively in Figure 8.1.4. The majority of the noise, observed
at higher frequencies, falls within this threshold. Thus we conclude that this high
frequency content has been buried in the noise of the system and there can be
no meaningful data extracted.
85 | P a g e
Figure 8.1.4: A comparison of the spectral content of the tumour response signal, for Case 1, with the
SBR-enabled system and the original system.
Note that Cases 3 and 4 are absent in the subsequent analysis. Since the
homogeneous breast phantom is an unrealistically simple scenario, only limited
resources were used for testing this breast phantom. Thus, we focused on just
one co-polarized and cross-polarized case, respectively. These two cases (Case 1
and Case 2) are sufficient to observe the trends in regards to the both the signal
strength and the frequency content of the transmitted signal when using the
SBR-enabled system.
Table X summarizes the results of our measurements. We present the data of
the comparison of the newly developed SBR-system and the original system. We
present data regarding the maximum received signal and the peak tumour
response. The maximum received signal is defined as the maximum signal
recorded at the receiving antenna. The peak tumour response is defined as the
peak signal level in the tumour response signal. In both cases, the new system
86 | P a g e
design has improved the received signal strength by a factor of more than twofold. The peak tumour response signal is also improved.
TABLE X: MAXIMUM RECEIVED SIGNAL AND PEAK TUMOUR RESPONSE SIGNAL COMPARISON
Received Signal
Tumour Response
[mV]
[mV]
SBR-
Original-
SBR-
Original-
System
System
System
System
Case 1
382.4
177.7
28.2
19.6
Case 2
91.8
40.4
20.2
16.3
In Table XI we present a comparison of the  parameter (See equation (3)) for
both systems and both cases (1 and 2). This parameter helps us assess whether
changes in the tumour response signal is dependent on the changes in the input
signal amplitude. For both cases, using the augmented system provides a
marginal gain in the  value in comparison with the original system. Hence, we
can conclude that even though the amplitude of the input signal is increased
with the augmented system, the amplitude of the tumour response signal is
increased by a greater factor.
TABLE XI: COMPARISON OF THE T PARAMETER
T Parameter (dB)
SBR-
Original-
Gain (in dB) from
System
System
SBR-System
Case 1
-48.82
-50.16
+ 1.34
Case 2
-51.71
-51.74
+0.03
87 | P a g e
Lastly, in Table XII we compare the results of computing the / metric for
both cases and both systems. The tumour response to baseline ratio helps to
identify whether the tumour response signal has increased in regards to the
baseline signal. This parameter gives insight on the difficulty of the imaging
scenario; if the ratio is very low it implies that the tumour response can be
buried beneath the background signal (signal from the healthy tissues). For the
100% fat phantom we note that the / values are higher for the initial
experimental system when compared to the augmented system. This suggests
that using the reshaped pulse, in a purely adipose medium, increases the
baseline signal by a larger factor than the tumour response signal in comparison
with the generic impulse. This outcome can potentially make tumour detection
more difficult when using image reconstruction algorithms, since there is an
increase in the power of the baseline signal and a subsequent decrease in the
relative power in the tumour response.
TABLE XII: COMPARISON OF THE T/B PARAMETER FOR BOTH INPUT SIGNALS
T/B Parameter (dB)
SBR-
Original-
Gain (in dB) from
System
System
SBR-System
Case 1
-22.65
-19.15
-3.50
Case 2
-13.028
-7.88
-5.15
88 | P a g e
8.2 Fat and Skin Breast Phantom
In this section we present the results when testing the fat and skin breast
phantom; a 2mm-thick skin layer surrounds the homogeneous, purely adipose,
breast phantom. The presented results follow the sequence of data presented in
Chapter 8.1.
A comparison of the recorded baseline time domain signals, for Case 1, is
shown in Figure 8.2.1. As with the 100% fat phantom, the use of the SBR to
reshape the input pulse improves the amplitude of the transmitted signal, and
the transmitted pulse is now a compact signal with a well defined profile.
Figure 8.2.1: A comparison of the recorded baseline signal for both experimental systems when the
antennas are arranged as per Case 1.
A comparison of the spectral content of both baseline signals, using the SBRsystem and the original experimental system, is shown in Figure 8.2.2. Similarly
to the 100% fat phantom case, the peak power in the baseline signal when using
89 | P a g e
the generic impulse as input to the system is centred around 1 GHz, whereas the
peak power for the SBR-system remains, once again, restricted to the 2 – 4 GHz
range. From Figure 8.2.2 we can make assumptions on the frequency response of
the breast phantom based on the spectral content of the baseline signals. In
certain frequency ranges we observe relative peaks for both systems, this
suggests that these are possible pass bands of the breast phantom (since the
both systems have very different input spectrum). Specifically, the pass bands at
1 GHz and from 2 – 4 GHz appear to have the least attenuation.
Figure 8.2.2: A comparison of the frequency content of the baseline signal for Case 1 when the breast
phantom is illuminated by the generic impulse and the reshaped pulse.
The tumour response signal, for Case 1, for both experimental systems is
plotted in Figure 8.2.3. The amplitude of the tumour response signal is increased
almost three fold by using the new augmented experimental system setup.
Additionally, the tumour response signal from the original system is not well
90 | P a g e
defined in time; there is a low-frequency ringing tail which extends for nearly
10ns.
Figure 8.2.3: A comparison of the tumour response signal, for Case 1, for both experimental systems.
In Figure 8.2.4, we compare the frequency profile of the tumour response for
both setups. For the original system, the spectral content of the tumour
response signal is restricted to the low frequency range, specifically near 500
MHz. This low frequency content explains the ringing tail observed in the tumour
response signal. For the SBR-system, the majority of the power remains
restricted to the 2 – 4 GHz range. However, there also appears to be a peak in
the spectral content aligned with the 500 MHz peak when using the original
system; this suggests that the response of the breast phantom and tumour has
potential pass bands at 500 MHz and in the 2 – 4 GHz range.
For the plots in Figure 8.2.4, the noise of the system is at -20 dB and -30 dB for
the original and augmented systems, respectively. Any content below these
thresholds should be considered to be buried within the noise. This confirms the
assumption that the tumour response signal is restricted around 500 MHz for the
91 | P a g e
original experimental system, and to the 2 – 4 GHz range for the augmented
system.
Figure 8.2.4: Frequency response of the tumour response signal for both experimental system when the
antennas are arranged in Case 1.
In Table XIII we compare the recorded transmitted signal, and the computed
tumour response signal, for all four antenna arrangements when the breast
phantom is illuminated by both input signals. The use of the augmented
experimental system leads to an increase in the amplitude of the received signal
regardless of the antenna positioning used. Furthermore, as expected, there is a
decrease in the signal amplitude when compared to measurement with the
purely adipose breast phantom. This additional loss can be attributed to the high
conductivity of the skin region.
The use of the SBR for pulse shaping improves the tumour response signal in
three of the four cases. Additionally, we observe that the SBR-system performs
best when the antennas are arranged in a co-polarized fashion (Case 1 and Case
3).
92 | P a g e
TABLE XIII: MAXIMUM RECEIVED SIGNAL AND PEAK TUMOUR RESPONSE SIGNAL COMPARISON
Received Signal
Tumour Response
[mV]
[mV]
SBR-
Original-
SBR-
Original-
System
System
System
System
Case 1
287.1
109.3
63.8
20.8
Case 2
99.7
52.8
10.7
9.2
Case 3
290.0
120.7
15.7
10.9
Case 4
84.6
54.0
13.9
36.7
A comparison of the T parameter, for both systems and each of the four cases,
is shown below in Table XIV. For the two co-polarized cases the augmented
system has better performance; the tumour response signal is increased by a
greater factor than the increase of the input signal. Additionally, based on the
optimal antenna arrangement (Case where T is maximized), corresponding to
Case 1 for the SBR-system and Case 4 for the original system, the SBR-system
performs better.
TABLE XIV: COMPARISON OF THE T PARAMETER
T Parameter (dB)
SBR-
Original-
Gain (in dB) from
System
System
SBR-System
Case 1
-41.72
-49.62
+ 7.90
Case 2
-57.23
-56.71
- 0.52
Case 3
-53.90
-55.24
+1.34
Case 4
-54.96
-44.69
- 10.3
93 | P a g e
Lastly, in Table XV, we compare the computed T/B metric. In all but Case 1,
based on the T/B, there is no improvement on the performance of the system
when implementing the SBR pulse shaping schematic (in fact performance is
worse). However, we suggest that caution be taken when trying to reach
conclusions based on this metric alone. The T/B parameter is dependent on the
amplitude of the baseline signal. With the original system the baseline signal
itself is very weak; furthermore, the tumour response signal is much lower, and
begins to approach the noise floor of the system. Hence, due to the relatively
weak baseline signal, even a weak tumour response signal will lead to a higher
T/B value.
We also note that the measurement with the original experimental system for
Case 4 seems to be an anomaly. The tumour response signal is abnormally high,
and, in fact, is almost of the same magnitude of the received signal. It is possible
that some small changes to antenna positioning between two measurements
(baseline and tumour signal) is responsible. These small changes in the antenna
location can lead to a misalignment of the time-domain signals. This offset in
time can cause the two signals to be shifted with respect to each other, and thus,
when we subtract the two signals from each other to compute the tumour
response signal, we get an erroneous value which appears larger than
anticipated.
TABLE XV: COMPARISON OF THE T/B PARAMETER FOR BOTH INPUT SIGNALS
T/B Parameter (dB)
SBR-
Original-
Gain (in dB) from
System
System
SBR-System
Case 1
-13.1
-14.1
+ 1.0
Case 2
-19.4
-15.1
- 4.3
Case 3
-25.3
-20.2
- 5.13
Case 4
-15.7
-3.36
- 12.34
94 | P a g e
8.3 50% Gland Breast Phantom
Figure 8.3.1 compares the recorded baseline signal for Case 3 when
illuminating the breast with both the generic impulse (original system) and the
reshaped pulse (SBR-enabled system). Using the SBR to reshape the input signal
increases the peak-to-peak voltage of the baseline signal from approximately 0.2
V to approximately 0.6 V. As per the previous two sections, despite the addition
of the glandular content in the breast phantom, the shape of the newly formed
pulse has been preserved after transmission through the breast structure.
Figure 8.3.1: Time-domain measurement of the received signal for the baseline recording of Case 3. We
compare the recorded signal for both the experimental system outlined in our previous work and the
newly designed system. Reproduced from [48].
Figure 8.3.2 compares the frequency content of these two recorded baseline
signals. Similar to the observations made with the 100% fat and the fat + skin
phantoms, we observe what appears to be a passband at 0.5 GHz (first small
peak for both systems), followed almost immediately by a stop band appearing
to be centered at ~ 1 GHz (first valley evident in both systems), and then
95 | P a g e
followed by an additional passband in the 2 – 4 GHz range. The addition of the
glandular content to the breast phantom does not modify the location of the
passbands for transmission through the breast phantom, suggesting that
changing the breast phantom composition from purely adipose tissue to a 50/50
blend of fat and gland tissue has minimal effect on the response of the phantom.
However, there is apparent attenuation at higher frequencies, beginning as early
as 3 GHz (affecting the spectral content of the signal in the 2 -4 GHz passband).
The high electrical conducitvity of the glandular content at these high
frequencies explains this additional attenuation.
Figure 8.3.2: Frequency content of the recorded baseline signal for Case 3 using both the generic impulse
and the newly formed pulse to excite the antenna. Reproduced from [48].
A comparison of the computed tumour response signal for Case 3 for both
experimental systems is plotted in Figure 8.3.3. Whereas the tumour response
using the generic impulse is difficult to distinguish, the response using the SBR
structure is a localized, easily discernible signal. The amplitude of the tumour
96 | P a g e
response signal is greatly increased when implementing the augmented system
in comparison to the original system.
Figure 8.3.3: Time-domain measurement of the tumour response for Case 3. We compare the tumour
response signal for both experimental prototypes. Reproduced from [48].
In Figure 8.3.4 we compare the frequency profile of the tumour responses
signal, for Case 3, for both setups. The peak power of the tumour response signal
for the two systems is very distinct. As we have previously reported, there
appears to be two distinct pass bands one centred at 0.5 GHz and the other in
the 2 – 4 GHz range. However, the addition of the glandular content attenuates
the signal in the 2 – 4 GHz range. Specifically, it is the high conductivity of the
glandular tissues that attenuates the higher frequency signals.
The noise floor, delimited by the 2 mV noise floor of the oscilloscope,
corresponds to approximately -15 dB and -29 dB for both the original system and
the SBR-system respectively in Figure 8.3.4. Thus the noise observed in Figure
97 | P a g e
8.3.4, mainly at higher frequencies, is mostly below this threshold, thus we
conclude that the noise in the plot is due to the limitations of the oscilloscope.
Figure 8.3.4: A comparison of the frequency content of the tumour response signal for Case 3, for both
experimental systems. Reproduced fro [48].
Table XVI summarizes the results of our measurements. We present two
data of interest: the maximum (peak) baseline signal and the maximum (peak) of
the measured tumour response signal for both experimental systems.
In all four cases, the new system design has increased the amplitude of the
baseline signal and the peak tumour response signal. This improvement in the
measured voltage of the transmitted signal and the tumour response signal
improves our ability to perform time-domain recordings since the new signal is
well above the noise floor of the oscilloscope. The most drastic improvements
occur for Case 1 and Case 3, the co-polarized cases.
98 | P a g e
TABLE XVI: MAXIMUM RECEIVED SIGNAL AND PEAK TUMOUR RESPONSE COMPARISON
Received Signal
Tumour Response
[mV]
[mV]
SBR-
Original-
SBR-
Original-
System
System
System
System
Case 1
285.8
95.7
48.8
10.6
Case 2
82.6
35.9
33.1
9.5
Case 3
303.8
124.8
55.6
8.4
Case 4
99.5
49.1
53.4
12.3
In Table XVII a comparison of the computed T parameter for each of the four
cases and for both of the experimental systems is presented. In all four cases the
use of the SBR for reshaping the input pulse results in improved performance.
The co-polarized cases (Case 1 and Case 3) perform best. Based on the value of
the T parameter, the optimal antenna arrangement corresponds to Case 3 for
the SBR-system and Case 4 for the original system.
TABLE XVII: COMPARISON OF T PARAMETER
T Parameter (dB)
SBR-
Original-
Gain (in dB) from
System
System
SBR-system
Case 1
-44.05
-55.48
+11.43
Case 2
-47.42
-56.43
+9.01
Case 3
-42.92
-57.50
+14.58
Case 4
-43.27
-54.19
+10.92
Table XVIII demonstrates that the use of the SBR for pulse shaping improves
the tumour detection capabilities of the experimental system by increasing the
99 | P a g e
T/B metric indicating that there is an increase in the power of the tumour
response signal in comparison to the baseline signal. This improvement is
observed across all four cases, with no significant bias for the co-polarized cases,
with the optimum antenna arrangement (maximum T/B value) being Case 3.
TABLE XVIII: COMPARISON OF THE T/B PARAMETER FOR BOTH INPUT SIGNALS
T/B Parameter (dB)
SBR-
Original-
Gain (in dB) from
System
System
SBR-system
Case 1
-15.4
-19.1
+3.7
Case 2
-7.9
-11.6
+3.7
Case 3
-14.8
-23.4
+8.6
Case 4
-5.4
-12.0
+6.6
100 | P a g e
8.4 70% Gland Breast Phantom
The recorded baseline signals for the original and augmented experimental
systems, for antennas arranged as per Case 3, when imaging a breast phantom
that is composed of 70% glandular tissue by volume is plotted in Figure 8.4.1.
Using the frequency limited pulse formed by the SBR improves signal
transmission through the breast in comparison to the recorded baseline when
the breast is illuminated by the generic impulse. The reshaped pulse retains its
shape after transmission through the breast phantom whereas the generic
impulse has become distorted.
Figure 8.4.1: A comparison of the time domain recordings of the baseline signal for Case 3 with both
experimental systems.
In Figure 8.4.2 a plot comparing the frequency content of the recorded
baseline signals using both input signals, for Case 3, is shown. Unlike all the
previous results (Sections 8.1, 8.2, and 8.3) the low frequency pass band (~ 0.5
101 | P a g e
GHz) seems to have been eliminated; there is no distinct peak in the power of
either signal at this frequency. The frequency content of the baseline signal,
when using the original experimental system, does not appear to be focused in
any one frequency range. However, for the augmented experimental system, the
spectral content of the baseline signal remains restricted to the 2 – 4 GHz.
As was observed in Chapter 8.3 with the 50% glandular phantom, the high
conductivity of the glandular tissues causes additional attenuation as the
frequency of excitation is increase. This effect is noticeable from approximately 3
GHz (decrease in power for the augmented system).
Figure 8.4.2: A comparison of the spectral content of the recorded baseline signal for both input signals.
The antennas are arranged as per Case 3.
A comparison of the computed tumour response signal for both the original
and augmented experimental system, with antennas arranged as per Case 3, is
shown in Figure 8.4.3. With the 70% gland breast phantom the tumour response
signal is significantly dampened; in fact, using the original experimental system,
102 | P a g e
the tumour response signal is barely discernible from the noise of the system.
The integration of the SBR with the experimental system increases the tumour
response signal, for Case 3, over 5 fold.
Figure 8.4.3: Tumour response signal comparison for both experimental systems. The antennas are
arranged as per Case 3.
In Figure 8.4.4, we compare the frequency content of the tumour response
signal for both experimental systems. For the original experimental system there
is no discernible frequency range with meaningful signal content; the tumour
response signal is buried in the noise of the system. For the SBR-enabled system
the peak power occurs at approximately 3 GHz, with the majority of the signal
content within the 2 – 4 GHz range. The 70% gland phantom is the most difficult
of the imaging scenarios presented. Using the original experimental system
makes detection nearly impossible (tumour response signal barely above the
noise floor), however, the use of the reshaped pulse for antenna excitation
greatly improve the tumour detection abilities of the system.
103 | P a g e
Figure 8.4.4: A comparison of the frequency content of the tumour response signal between the two
experimental systems. The antennas are arranged as per Case 3. The power from the signal computed
using the original system is buried within the noise of the system.
In Table XIX we present a summary of the measured data for the 70% (by
volume) glandular breast phantom. The table compares the peak baseline
(received) signal and the peak tumour response signal for all the antenna
arrangements and each of the experimental systems. In comparison with the
three other breast phantoms tested, the baseline signals for the 70% glandular
phantom are the lowest. This result is expected; the glandular tissues have a
higher loss than the adipose tissue, thus, as the glandular content is increased
we expect the propagating signal to undergo additional attenuation.
For the augmented system the co-polarized orientation of the antennas
greatly improves both the transmitted and the tumour response signal. The
improvements in the cross-polarized cases are minimal in comparison.
Additionally, we note two anomalies. In Case 2, for the SBR-system, the baseline
signal is very low in comparison to the other cases (less than half). In Case 4, for
104 | P a g e
the original system, the tumour response signal is nearly the same amplitude as
the baseline signal. A potential cause for this anomaly is the possibility of
antenna movement during measurement (as was explained in Section 8.2).
TABLE XIX: MAXIMUM RECEIVED SIGNAL AND PEAK TUMOUR RESPONSE COMPARISON
Received Signal
Tumour Response
[mV]
[mV]
SBR-
Original-
SBR-
Original-
System
System
System
System
Case 1
207.1
101.7
46.8
7.4
Case 2
28.9
38.9
23.6
15.7
Case 3
197.7
103.4
37.5
6.6
Case 4
60.0
44.3
10.8
27.9
In Table XX we compare the values of the T parameter across all antenna
arrangements for both experimental setups. In 3 of the 4 cases there is an
improvement by integrating the SBR into the system. Furthermore, the use of
the co-polarized antenna arrangements leads to the biggest improvement in the
T value. The optimal antenna arrangement, based on the maximum value of the
T parameter, is Case 1 for the augmented system and Case 4 for the original
system.
Lastly, we present the computed T/B parameter values for the 70% glandular
phantom in Table XXI. We note that the anomalies in the measurement data
mentioned previously, Case 2 for the SBR-system (very weak baseline signal), and
Case 4 for the original system (very large tumour response signal potentially due
to antenna movement), have a very prominent effect on the T/B values.
However, we do observe that for the co-polarized cases the tumour detection
capabilities of the experimental system are increased by the implementation of
the pulse shaping circuit. Based on the increase T/B values, there is an increase
105 | P a g e
in the power of the tumour response signal relative to the power in the baseline
signal.
TABLE XX: COMPARISON OF T PARAMETER
T Parameter (dB)
SBR-
Original-
Gain (in dB) from
System
System
SBR-system
Case 1
-44.42
-58.60
+14.18
Case 2
-50.36
-52.07
+1.71
Case 3
-46.34
-59.60
+13.26
Case 4
-57.15
-47.07
-10.08
TABLE XXI: COMPARISON OF THE T/B PARAMETER FOR BOTH INPUT SIGNALS
T/B Parameter (dB)
SBR-
Original-
Gain (in dB) from
System
System
SBR-system
Case 1
-12.91
-22.80
+ 9.89
Case 2
-1.76
-7.91
+ 6.15
Case 3
-14.40
-23.90
+ 9.5
Case 4
-14.90
-4.02
- 10.88
106 | P a g e
8.5 System Performance Assessment and Comparison
Time domain measurements were carried out on each of the four breast
phantoms described (100% fat, 100% fat +skin, 50% gland, and 70% gland) when
using both the original and the augmented experimental systems. Four antenna
arrangements, Cases 1 through 4, were tested on each phantom. In this section
we will compare and assess the overall performance of the newly designed
experimental system, with respect to the original experimental system, across
the four phantoms tested and for each of the four antenna arrangements.
Table XXII, [48], compares the maximum peak tumour response signal, in mV,
for both systems, across all four cases and for each of the breast phantoms
under investigation. Furthermore, we include the gain, in mV, obtained by
incorporating the SBR into the experimental system. The pre-shaping of the
input pulse with the SBR has led to an increase in the recorded tumour response
signal in each of the four breast phantoms created. Included in parenthesis is the
antenna arrangement that corresponds to the given signal recording.
In each of the four phantom types, using a co-polarized antenna arrangement
(Case 1 and 3) increases the amplitude of the tumour response signal when
imaging the breast with the augmented experimental system. Hence, we can
conclude that when using the SBR-enabled system it is ideal to arrange the
antennas so that they are co-polarized. For the original system, such a claim
cannot be made. In fact, as was presented in [43], for cross-polarized antennas
an increase in the tumour response signal is observed.
The gain in the tumour response signal when using the SBR-system compared
to the original system, is more pronounced in the complex heterogeneous breast
phantoms. These cases correspond to the most life-like phantoms and, as such,
are the most difficult to image. The original system struggles to detect tumours
in these complex models; the incorporation of the pulse shaping circuitry greatly
phantoms.
107 | P a g e
TABLE XXII: MAXIMUM TUMOUR RESPONSE SIGNAL COMPARISON (IN MV) ACROSS TRIALS FOR EACH
PHANTOM TYPE AND EXPERIMENTAL SYSTEM, [48]
Tumour Response Signal [mV]
Phantom
Type
SBR-
Original-
Gain from
System
System
SBR-system [mV]
Fat
28.2 (case 1)
19.6 (case 1)
+ 8.6
Fat +Skin
63.8 (case 1)
36.7 (case 4)
+ 27.1
50% Gland
55.6 (case 1)
12.3 (case 4)
+ 43.3
70% Gland
46.8 (case 3)
15.7 (case 4)
+ 31.1
The T parameter is used to provide an understanding of the strength of the
tumour response signal relative to the input signal of the system.
Table XIII compares the best  value, expressed in dB, for each phantom,
across the four antenna arrangements, between the original system and the
augmented system. The input signal, for each system respectively, is
independent to the specific trial, thus the antenna arrangement which led to the
maximum tumour response signal from Table XXII is identical to the antenna
arrangement which leads to the maximum T value.
The integration of the SBR within the experimental system has improved the 
parameter in each of the four breast phantoms. Similarly, the improvement in
tumour detection is most apparent in the more complex heterogeneous breast
phantoms composed of 50% and 70% gland. Furthermore, while the original
system has more difficulty detecting tumours in the heterogeneous breast
phantoms when compared to the homogeneous 100% fat phantom, the use of
the SBR has the opposite effect; tumour detection is easier, as denoted by a
larger  parameter value, for the heterogeneous breast phantoms when
compared to the 100% fat phantom.
108 | P a g e
TABLE XXIII: COMPARISON OF T PARAMETER (DB) ACROSS TRIALS FOR EACH PHANTOM TYPE AND
EXPERIMENTAL SYSTEM, [48]
Tumour Response Parameter,  [dB]
Phantom
Type
SBR-
Original-
Gain from
System
System
SBR-system [dB]
Fat
-48.82
-50.16
+ 1.3
Fat +Skin
-41.72
-44.69
+ 3.0
50% Gland
-42.92
-56.02
+ 13.1
70% Gland
-44.42
-52.08
+ 7.7
In regards to the T/B parameter, which indicates the amount of power in the
tumour response signal relative to the baseline, similar trends have been
observed: i) the use of the augmented system led to a greater improvement
when imaging the complex heterogeneous breast phantoms, and ii) when using
the SBR-enabled system the co-polarized antenna arrangements are ideal for
tumour detection.
Based on the frequency analysis of both the measured baseline signals and the
compute tumour response signals we can conclude that the response of the
various breast phantoms can be approximated by a pass band filer. Specifically, it
appears that there are two major pass bands, once centred near 0.5 GHz and the
other existing in the 2 – 4 GHz range.
A possible explanation for the pass band at lower frequencies can be related to
attenuation characteristics of the tissues of the breast phantom. Electrical
conductivity is a function of frequency; as the frequency increase so too does the
conductivity. This increased conductivity causes an increase in attenuation for
signals propagating through the breast structure at these higher frequencies.
109 | P a g e
Thus, the low frequency content experiences minimal attenuation in
comparison.
The presence of this low frequency content can limit the performance of the
system in two ways: i) it is a limiting factor for the pulse repetition rate (prevent
signal overlap) ii) at these low frequencies, the maximum amount of energy
which can be safely deposited into human tissue, based on safety standards
established by the IEEE, is much lower [8], [27].
We note that as the glandular content is increased the attenuation of the
signal content at higher frequencies is more apparent. The glandular tissue has
very high conductivity in comparison to the adipose tissues, thus, as the
frequency of the input signal is increased, the conductivity of the breast
phantom is sharply increased as well, leading to additional losses for the signal
propagating within the breast phantom.
110 | P a g e
Chapter 9: Summary and Conclusion
A first generation experimental system for time domain microwave imaging
was initially developed by Porter et al. and was tested on various breast
phantom tissues to assess the ability of the system to detect the presence of
tumours. In this thesis we present numerical models used to mesh simulations to
this experimental setup. We present new numerical models to represent the
various breast phantoms that were created experimentally. These models have
been created to have dielectric properties which behave in accordance to the
measured data of the fabricated breast phantoms.
A FDTD analysis was performed in order to determine the maximum power
that can be safely input to the time domain imaging system. We conclude that as
the frequency is increased the amount of energy locally absorbed by regions
within the breast structure is also increased. Furthermore, we demonstrate that
this relationship between increasing frequency and increasing peak local SAR
values is well represented by a linear relationship. We also demonstrate that
when an antenna is used to illuminate the breast the location of maximum
absorbance is dependent on antenna location. In all cases the peak SAR value
occurs in the skin region, where conductivity is highest. We determined that for
a breast illuminated by a broadband antenna located in the near field, the
maximum 10-g averaged SAR value, normalized to 1 mW input power, is
approximately 1/400 of the maximum permissible value.
A passive microstrip line, with a specific conductive profile width, is fabricated
to reshape a generic pulse. The newly formed pulse has spectral content
restricted to the 2 – 4 GHz range. This fabricated device, referred to as a SBR, is
then integrated with the experimental system developed by Porter et al. with
some pulse shaping circuitry. We assessed whether the use of this SBR to form
an augmented system improves the tumour detection capabilities of the system
111 | P a g e
in comparison to the original experimental system. Four antenna arrangements
are tested on four different breast phantoms.
The use of the reshaped pulse in the augmented system serves to increase the
tumour detection capabilities of the system, based on an increase in the tumour
response signal, the T parameter, and the T/B parameter. These improvements
are most notable in the heterogeneous breast phantoms composed of glandular
tissue and adipose tissue. Arranging the antennas so that they are co-polarized
optimizes the tumour detection capabilities of the system.
112 | P a g e
REFERENCES
[1] Canadian Cancer Society. (2010, August 17). What is breast cancer? [Online]. Available:
http://www.cancer.ca/
[2] Canadian Cancer Society’s Steering Committee on Cancer Statistics. (2011, May). Canadian
Cancer Statistics 2011. Toronto, ON: Canadian Cancer Society.
[3] Canadian Cancer Encyclopedia. (2012, Jan 25). [Online]. Available: http://info.cancer.ca/cceecc/
[4] C. Kuhl, et al. “Mammography, Breast Ultrasound and Magnetic Resonance Imaging for
Surveillance of Women at High Risk of Familial Breast Cancer. ” Journal of Clinical Oncology.
vol. 23, no.33, pp. 8469-8476, Nov. 2005.
[5] M. Kriege, et al. "Efficacy of MRI and mammography for breast cancer screening in women
with a familial or genetic predisposition". N Engl J Med 351:427-437, 2004
[6] E.C. Fear, P.M. Meaney, M.A. Stuchly, "Microwaves for breast cancer detection?" IEEE
Potentials, pp.12-18, February/March 2003.
[7] W.M.A. Ibrahim and H.M. Algabroun, "Family Tree of Breast Microwave Imaging
Techniques," in Biomed 2008, Proceedings 21, 2008, pp. 258-261.
[8] Health Canada, “Limits of Human Exposure to Radiofrequency Electromagnetic Fields in the
Frequency Range from 3 kHz to 300 GHz: Safety Code 6 ,” Health Canada, 2009.
[9] E. Porter, A. Santorelli, M. Coates, M. Popovid, “An experimental system for time-domain
th
microwave breast imaging,” in Proc. 5 European Conference on Antennas and Propagation
(EUCAP 2011), Rome, Italy, April 11-15, 2011.
[10] eHealthMD.
(2010)
Breast
Anatomy.
http://www.ehealthmd.com/library/breastcancer/BRC_anatomy.html
[Online].
[11] Mayo Clinic Staff, Mayo Clinic. (Oct. 19, 2011). ‘Femal Breast Anatomy’ *Online+. Available:
http://www.mayoclinic.com/health/breast-cancer-early-stage/BC00001
[12] V. Kumar, A. Abbas, N. Fausto and J. Aster, Robbins &Cotran Pathological Basis of Disease
(Chapter 7: Neoplasia), 8th ed., pp.259-330, United States: Elsevier, 2010.
[13] M. Lazebnik et al., "A large-scale study of the ultrawideband microwave dielectric properties
of normal breast tissue obtained from reduction surgeries," Phys. Med. Biol., vol. 52, pp.
2637-2656, 2007.
[14] M. Lazebnik et al., "A large-scale study of the ultrawideband microwave dielectric properties
of normal, benign and malignant breast tissue obtained from cancer surgeries," Phys. Med.
Biol., vol. 52, pp. 6093-6115, 2007.
[15] M. Lazebnik, M. Okoniewski, J. H. Booske, S. C. Hagness, "Highly Accurate Debye Models for
Normal and Malignant Breast Tissue Dielectric Properties at Microwave Frequencies," IEEE
Microwave and Wireless Components Letters, vol. 17, no. 12, pp. 822-824, December 2007.
[16] Craddock, I. J.; Preece, A.; Leendertz, J.; Klemm, M.; Nilavalan, R.; Benjamin, R.; ,
"Development of a hemi-spherical wideband antenna array for breast cancer imaging,"
Antennas and Propagation, 2006. EuCAP 2006. First European Conference on , vol., no., pp.15, 6-10 Nov. 2006
[17] Craddock, I.J.; Klemm, M.; Leendertz, J.; Preece, A.W.; Benjamin, R.; , "An Improved
Hemispeherical Antenna Array Design for Breast Imaging," Antennas and Propagation, 2007.
EuCAP 2007. The Second European Conference on , vol., no., pp.1-5, 11-16 Nov. 2007.
113 | P a g e
[18] Klemm, M.; Craddock, I.J.; Leendertz, J.; Preece, A.W.; Benjamin, R.; , "Breast Cancer
Detection using Symmetrical Antenna Array," Antennas and Propagation, 2007. EuCAP 2007.
The Second European Conference on , vol., no., pp.1-5, 11-16 Nov. 2007.
[19] Klemm, M.; Craddock, I.J.; Leendertz, J.A.; Preece, A.; Benjamin, R.; , "Radar-Based Breast
Cancer Detection Using a Hemispherical Antenna Array—Experimental Results," Antennas
and Propagation, IEEE Transactions on , vol.57, no.6, pp.1692-1704, June 2009.
[20] Klemm, M.; Craddock, I.J.; Leendertz, J.A.; Preece, A.; Gibbins, D.R.; Shere, M.; Benjamin, R.; ,
"Clinical trials of a UWB imaging radar for breast cancer," Antennas and Propagation
(EuCAP), 2010 Proceedings of the Fourth European Conference on , vol., no., pp.1-4, 12-16
April 2010.
[21] P.M. Meaney, M.W. Fanning, L. Dun, S.P. Poplack, and K.D. Paulsen, "A clinical prototype for
active microwave imaging of the breast," IEEE Transactions on Microwave Theory and
Techniques, vol. 48, no. 11, pp. 1841-1853, 2000.
[22] P.M. Meaney et al., "Clinical microwave breast imaging – 2D results and the evolution to 3D,"
in International Conference on Electromagnetics in Advanced Applications, 2009. ICEAA 2009,
2009, pp. 881-884.
[23] M. Lazebnik, E. Madsen, G. Frank, S. Hagness, "Tissue-mimicking phantom materials for
narrowband and ultrawideband microwave applications," Phys. Med. Biol., vol. 50, pp. 42454258, 2005.
[24] E. Porter, J. Fakhoury, R. Oprisor, M. Coates, M. Popovid, "Improved tissue phantoms for
th
experimental validation of microwave breast cancer detection," in Proc. 4
European
Conference on Antennas and Propagation (EuCAP), 2010, vol., no., pp.1-5, 12-16 April 2010.
[25] E. Porter, A. Santorelli, A. Bourdon, D. Coulibaly, M. Coates, M. Popovid, “Time-Domain
Microwave Breast Cancer Detection: Experiments with Comprehensive Glandular
Phantoms”, in Proc. 2011 Asia Pacific Microwave Conference, Melbourne, Australia
December 5-8, 2011, in press.
[26] E. Porter, A. Santorelli, D. Coulibaly, M. Coates , M. Popovid, “Time-Domain Microwave
Breast Screening System: Testing with Advanced Realistic Breast Phantoms”, in Proc. 2012
6th European Conference on Antennas and Propagation (EUCAP 2012), Prague, Czech
Republic, March 26-31, 2012, in press.
[27] IEEE Std C95.1—1999, IEEE Standard for Safety Levels With Respect to Human Exposure to
Radio Frequency Electromagnetic Fields, 3 kHz to 300 GHz. Piscataway, NJ: IEEE Press, 1999.
[28] P. Bernardi, M. Cavagnaro, S. Pisa, E. Piuzzi, “Specific Absorption Rate and Temperature
Elevation in a Subject Exposed in the Far-Field of Radio-Frequency Sources operating in the
10-900-Mhz Range," IEEE Trans. Biomed. Eng., vol. 50, pp. 295–304, March 2003.
[29] E. Zastrow, S.K. Davis, and S.C. Hagness, “Safety assessment of breast cancer detection via
ultrawideband microwave radar opoerating in swept frequency mode,” in Proc. of IEEE
International Symposium on Antennas and Propagation, Albuquerque, NM, July 2006.
[30] E. Zastrow, S. K. Davis, and S. C. Hagness, (2007), “Safety assessment of breast cancer
detection via ultrawideband microwave radar operating in pulsed-radiation mode,”
Microwave and Optical Technology Letters, 49: 221–225.
[31] A. Santorelli, M.V. Schueren and M. Popovid, "SAR Levels in Microwave Breast Imaging: 3-D
Safety Assessment with Plane-Wave Illumination," in Proc. The Asia Pacific Microwave
Conference (APMC 2011), Melbourne, Australia, December 5-8, 2011.
[32] E. Zastrow, S. Davis, M. Lazebnik, F. Kelcz, B. V. Veen, and S. Hagness, “Development of
anatomically realistic numerical breast phantoms with accurate dielectric properties for
114 | P a g e
modeling microwave interactions with the human breast,” IEEE Trans. Biomed. Eng., vol. 55,
no. 12, pp. 2792–2800, Dec. 2008.
[33] G. Zhu, B. Oreshkin, E. Porter, M. Coates, M. Popovid “Numerical Breast Models for
Commercial FDTD Simulators”, in Proc. European Conference on Antennas and Propagation
(Eu-CAP'09), Berlin, Germany, Mar. 2009.
[34] I. Arnedo, M. A. G. Laso, F. Falcone, D. Benito, T. Lopetegi, “A series solution for the single
mode synthesis problem based on the coupled mode theory,” Microwave Theory and
Techniques, IEEE Transactions on, vol. 56, no. 2, pp. 457–466, Feb. 2008.
[35] M. Chudzik, I. Arnedo, I. Arregui, A. Lujambio, M.A.G. Laso, D. Benito, T. Lopetegi, “Synthesis
technique for microwave circuits based on inverse scattering: Efficient algorithm
implementation and application”, International Journal of RF and Microwave ComputerAided Engineering, vol.21, no.2, pp. 163-174, 2011.
[36] I. Arnedo, J.D. Schwartz, M.A.G. Laso, T. Lopetegi, D.V. Plant, J. Azaa, "Passive microwave
planar circuits for arbitrary UWB pulse shaping," Microwave and Wireless Components
Letters, IEEE , vol.18, no.7, pp.452-454, July 2008
[37] J.M Sill, E.C. Fear , "Tissue sensing adaptive radar for breast cancer detection - experimental
investigation of simple tumor models," Microwave Theory and Techniques, IEEE Transactions
on, vol.53, no.11, pp. 3312- 3319, Nov. 2005
[38] Tektronix. (2011, August 15). [Online]. Available: http://www.tek.com/
[39] Picosecond Pulse Labs. (2011, August 15). [Online]. Available: http://www.picosecond.com/
[40] Friatec. (2011, August 15). [Online]. Available:http://www.friatec.com/
[41] H. Kanj and M. Popovid, "A Novel Ultra-Compact Broadband Antenna for Microwave Breast
Tumor Detection,.” Progress in Electromagnetics Research, vol. 86, pp. 169-198, 2008.
[42] Pico Technology Limited. (2011, August 15). [Online]. Available:http://www.picotech.com
[43] E. Porter, A. Santorelli, M. Coates, M. Popovid, “Microwave breast imaging: time-domain
experiments on tissue phantoms,” in Proc. 2011 IEEE International Symposium on Antennas
and Propagation (AP-S 2011), Spokane, Washington, U. S. A, July 3-8, 2011.
[44] SPEAG, Schmid & Partner Engineering AG. (2012, February 3). [Online].Available:
http://www.speag.com/
[45] ITU-R (2001). Article 5 of the Radio Regulations.
http://users.ictp.it/~puboff/lectures/lns016/Vol16Annex.pdf
[Online]
Available:
[46] J. Bourqui, M. Okoniewski, E. C. Fear, “Balanced Antipodal Vivaldi Antenna With Dielectric
Director for Near-Field Microwave Imaging”, IEEE Trans. on Antennas and Prop., VOL. 58, NO.
7, July 2010.
[47] A. Santorelli, M. Popovid, "SAR distribution in microwave breast screening: Results with
TWTLTLA wideband antenna," Intelligent Sensors, Sensor Networks and Information
Processing (ISSNIP), 2011 Seventh International Conference on , vol., no., pp.11-16, 6-9 Dec.
2011.
[48] A. Santorelli, E. Porter, M. Popovid, J. Schwartz “Pulse Shaping for Time Domain Microwave
Breast Tumour Detection: Experiments with Realistic Tissue Phantoms,” in Proc. 2012 6th
European Conference on Antennas and Propagation (EUCAP 2012), Prague, Czech Republic,
March 26-31, 2012, in press.
115 | P a g e
Документ
Категория
Без категории
Просмотров
0
Размер файла
3 912 Кб
Теги
sdewsdweddes
1/--страниц
Пожаловаться на содержимое документа