‘gjjg'i, 43%;? 73.1%..

:3.
M

A;
[5

; $3

’4
an-‘ _

; v. -{ '
4:":
‘3’,

:Nn-J.‘
a” .

{on
A“

: , .1"?
131%: 3939,3931 ’i.‘ '7’
.s 3435’*::2;m‘3« ‘- 333
v:.=‘:£"’ ' '

§
3

I. > _ , .
g- 4131*. ,ew ‘
a‘

.‘s

unlv
T

.M

:1" f’
”39.. .3:

 

-a
\g.

5’ p
D
*J

This is to certify that the
dissertation entitled

Hyperthermia treatment of Breast Cancer with RF phased
array applicator and RF/U S hybrid applicator

presented by

LIYONG WU

has been accepted towards fulfillment
of the requirements for the

 

 

 

 

Ph.D degree in Electrical & Computer
EngineerinL
Majcfi Professors Signaggi
Obi- 2— 7 L 2.00 6
Date

MSU is an Affinnative Action/Equal Opportunity Institution

W

Michigan Stale
University

 

PLACE IN RETURN BOX to remove this checkout from your record.
To AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

6/07 p:/CIRC/DateDue.indd-p.1

Hyperthermia treatment of Breast Cancer with RF phased

array applicator and RF/U S hybrid applicator

By

Liyong Wu

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY
Department of Electrical and Computer Engineering

2006

ABSTRACT

Hyperthermia treatment of Breast Cancer with RF

phased array applicator and RF/U S hybrid applicator

By
Liyong Wu

Several RF phased array applicators have been designed and constructed for
hyperthermia treatments in the intact breast. These RF phased arrays consist of
four or five antennas mounted on a Iexan water tank, and geometric focusing is
employed so that each antenna points in the direction of the intended target. The
operating frequency for these phased arrays ranges from 130-160 MHz. The RF
arrays have been characterized both by electric field measurements in a water
tank and by electric field simulations using the finite element method. The finite
element simulations are performed includes the geometry of the tank enclosure,
antennas and 3D patient model extracted from medical images. The electric (B)
field, specific absorbing ratio (SAR) and temperature in the breast region by the
four antenna applicator are calculated with SAR Optimization. In order to
improve the heating performance in deep region, an radio frequency/ultrasound
(RF/US) hybrid applicator is designed and a square shape ultrasound phased
array is mounted on one side wall and/or the bottom wall. The ultrasound
phased array delivers energy into the deep region of the tumor, where RF can
not reach and the combined heating has better performance than RF alone and

US alone. The temperature distribution generated by the hybrid applicator shows

that the character of RF and US source improves the heating performance for
regional hyperthermia. The effect of the E field polarization is also examined
and a cross antenna, radiated B field in the circular polarization mode, is
designed to remove the hot spot caused by B field linear polarization of the dual
dipole antenna. Several new applicators based on the dipole antenna, U-shaped
and cross antenna are designed for improving the heating performance and

steering ability of the RF hyperthennia applicator.

Copyright © by
LIYONG WU
2006

For my wife, Xiangrong and my parents

ACKNOWLEDGMENTS

There are many peOple who deserve acknowledgement for both the work contained
in this thesis, and the countless other things that I was able to be involved in during
my time at Michigan State. First and foremost, a special thanks to Dr. Robert
McGough for letting me walk my own paths, while always being there to point to
things on the map. I have truly enjoyed my time as your student and feel fortunate
to have been able to study under your guidance. Thanks to Dr. Shanker B. and Dr.
Leo Kempel for teaching me some computational EM stuff that will help me for years
to come. A thank you is also in order for Dr. Guowei Wei and Dr. Ed Rothwell, for
agreeing to serve on my Ph.D. committee.

Thanks to James Kelly for research related talk, which inspires me new ideas.
Thanks to Xiaozheng and Chen Duo for cooperation on the research. which helped
me to keep tip with publication activities.

A special acknowledgement to my parents for their love and support always.
Thank you for that. My deepest gratitude is reserved for wife, Xiangrong, who
has supported me throughout my studies and will give me a beautiful baby. Without
your help, all this would not have been possible. I love you and our future baby with

all my heart and soul.

vi

TABLE OF CONTENTS

LIST OF FIGURES ................................ xi

LIST OF TABLES ................................. xvii
CHAPTER 1

Introduction ..................................... 1

1.1 The cellular and molecular basis for hyperthermia ........... 1

1.2 The breast cancer ............................. 2

1.2.1 Early or operable breast cancer ................. 2

1.2.2 Locally advanced breast cancer ................. 3

1.2.3 Advanced breast cancer ..................... 3

1.3 The RF/ Microwave modality for hyperthermia ............. 3

1.4 The ultrasound phased array modality for hyperthermia ....... 4
CHAPTER 2

Simulation Methods and Routines ......................... 6

2.1 Electromagnetic numerical calculation ................. 6

2.1.1 The Finite Element Method (FEM) ............... 6

2.1.2 Finite Difference in Time Domain (FDTD) ........... 10

2.2 Ultrasound pressure calculation ..................... 11

2.2.1 The Fast near field method (FNM) for calculating pressure field 11

2.2.2 Angular spectrum approach (ASA) ............... 11

2.3 Thermal analysis ............................. 12

2.4 E-field calculation with HFSS ...................... 14

2.5 Focusing strategy ............................. 14
CHAPTER 3

Four channel RF phased array applicator .................... 17

3.1 RF phased array applicator and E field measurement system ..... 18

3.1.1 Applicator geometry ....................... 18

3.1.2 Amplifier system ......................... 21

3.1.3 Measurement system ....................... 21

3.2 Simulation Methods ............................ 23

3.3 Results ................................... 25

3.3.1 E-field measurements in the water tank ............. 25

3.3.2 E—field simulations in the water tank .............. 27

3.3.3 Breast model simulations ..................... 30

3.4 Discussion ................................. 34

vii

3.4.1 Measured and simulated E-fields in the water tank ...... 34

3.4.2 Focusing strategy ......................... 39

3.4.3 Simulated E-fields evaluated in the breast model ........ 39

3.5 Conclusion ................................. 43
CHAPTER. 4

The Five RF dual—dipole antenna applicator ................... 45

4.1 The five antenna MRI con‘ipatible applicator .............. 45

4.2 Phantom and Patient ........................... 47

4.2.1 Homogeneous Gel Phantom ................... 49

4.2.2 Fat/ Tumor Heterogeneous Breast Phantom ........... 49

4.2.3 The patient and patient model .................. 49

4.3 Simulation model and method ...................... 52

4.4 Simulation and Experiment Results ................... 52

4.4.1 E field and Focus steering in water ............... 53

4.4.2 Focus Steering in Homogeneous Gel Phantom ......... 53

4.4.3 Heat Focus Steering in Fat / Tumor Heterogeneous Phantom . 56

4.4.4 Heat the real 3D breast model .................. 60

4.5 Discussion ................................. 62

4.5.1 Focus steering in water and homogeneous gel phantom . . . . 66

4.5.2 Heat focus steering in fat / tumor heterogeneous phantom . . 68

4.5.3 Heat the real 3D breast model .................. 69

4.6 Conclusion ................................. 70
CHAPTER. 5

Hybrid RF/ US phased array applicator (1) .................... 72

5.1 Motivation ................................. 72

5.2 Applicator and 3D patient model .................... 73

5.2.1 Planar ultrasound phased array ................. 73

5.2.2 RF phased array and applicator ................. 73

5.2.3 Patient model .......................... 74

5.3 Methodology ............................... 76

5.3.1 Pressure field calculation and mode scanning techniques . . . 76

5.3.2 E field computation ........................ 78

5.3.3 SAR and Temperature Optimization .............. 78

5.4 Results ................................... 79

5.5 Discussion ................................. 85

5.6 Conclusions ................................ 93

CHAPTER 6

Sector-vortex scanning for a large square US phased array aperture ......
6.1 Introduction ................................
6.2 Phasing Scheme ..............................
6.3 Methods ..................................

6.3.1 Pressure field calculation .....................
6.3.2 Patient model and thermal calculation .............
6.4 Results ...................................
6.4.1 Pressure field patterns ......................
6.4.2 Temperature distribution .....................
6.4.3 Frequency penetration study ...................
6.5 Discussion .................................
6.6 Conclusion .................................

CHAPTER. 7

Hybrid RF/US Phased Array Applicator (2) for Small Breast .........
7.1 Introduction ................................
7.2 Applicator ................................

7.2.1 RF phased array and applicator .................
7.2.2 Planar ultrasound phased array .................
7.3 Methodology ...............................
7.3.1 Pressure field calculation .....................
7.3.2 Temperature Optimization ...................
7.4 Results ...................................
7.5 Discussion .................................
7.6 Conclusions ................................

CHAPTER 8

E field polarization and Cross antenna ......................
8.1 Boundary condition ............................
8.2 Cross antenna and applicator ......................
8.3 Results ...................................

8.3.1 Electromagnetic wave propagation in antenna .........
8.3.2 E field distribution for one antenna ...............
8.3.3 Single cross antenna applicator at 915MHz with oil bolus . . .
8.3.4 E field distribution in four antenna applicator .........
8.4 Discussion .................................
8.4.1 The delay line of the cross antenna ...............

ix

94
94
95

97
97
97

101
101
103
103

106
109

110
110

111
111
112

114
114
115

115
125
128

129
130
131

135
136
136
138
141

146
146

8.4.2 E field distribution in breast model ............... 148

8.4.3 Single cross antenna applicator at 915MHz .......... 149

8.4.4 E field distribution in four antenna applicator ......... 150

8.5 Conclusion ................................. 151
CHAPTER 9

Summary and future research ........................... 152

BIBLIOGRAPHY ................................. 155

Figure 3.1

Figure 3.2

Figure 3.3

Figure 3.4

Figure 3.5

Figure 3.6

Figure 3.7

LIST OF FIGURES

RF phased array prototype with four antennas designed for hyper-
thermia treatments in the intact breast. ..............

RF phased array amplifier system. This rack-mounted system con-
sists of a signal generator, a vector voltmeter, a multiplexer / switch
box, and four RF power amplifiers. The signal source generates a
common excitation frequency for each of the amplifiers, the vec~
tor voltmeter provides phase and power feedback, and the multi-
plexer/ switch box combination controls the inputs to the RF am-
plifiers. In turn, the RF amplifiers drive the individual antennas
in the phased array applicator. ...................

Three E-field array probes are attached to a Plexiglas rod for mea-
surements of the electric field produced by the RF applicator. This
probe arrangement is scanned across a rectilinear grid within the
water tank by a computer-controlled positioning system. The mea-
surements obtained with these scans characterize the E-field dis-
tribution generated by the RF phased array. ...........

Simulation model for the four antenna RF phased array applicator.
The geometric model, which has the same dimensions as the ap-
plicator in Fig. Figure 3.1, defines the input parameters for finite
element simulations. .........................

Schematic of the breast model defined for FEM simulations. The
breast is modeled by a hemisphere with a 75mm radius, and a
spherical tumor model with a 25mm radius is located inside the
breast. The hemispherical breast model is attached to a 5mm thick

skin layer, a 25mm thick fat layer, and a 42mm thick muscle layer.

E-field distributions generated by the RF phased array depicted
in Fig. Figure 3.1. The applicator prototype operates at 140MHz,
producing a focus in the center of a tank filled with deionized wa-
ter. The E-field measurements are performed by the apparatus
depicted in Fig. Figure 3.1, and the E-field is computed with the
finite element method for uniform phase and amplitude inputs.

Examples of measured (a) and simulated (b) 140MHz E—fields in
the xy plane achieved through electronic steering. Although some
differences appear near the far corner of the grid, the shapes of
these E-field meshes are quite similar, particularly in the region
near the peak. In (a) and (b), the E-field is measured 3cm below
the water surface (z *- -3cm). ....................

xi

19

20

22

24

26

28

31

Figure 3.8

Figure 3.9

Figure 3.10

Figure 4.1

Figure 4.2
Figure 4.3
Figure 4.4

Figure 4.5
Figure 4.6

Figure 4.7

Figure 4.8
Figure 4.9
Figure 4.10
Figure 4.11
Figure 4.12

E field distribution in the water tank. the input power applied
to each channel is 1W, and the phase value is 0° for each input
channel. a) E field in xy plane with z '— -3cm, b) E field in xz
plane with y - 0cm and b) E field in yz plane with x -= 0cm.

Simulated E-field, SAR, and temperature distributions generated
by the RF phased array applicator and evaluated in the x = 0 plane
of the breast tumor model, where the white contours indicate the
external outlines of the breast and tumor. These simulation results
show that, in the :1: = 0 plane, the E-field peaks in water are near
the source antennas, the E-field peaks in tissue are near the skin
surface, the peak SAR values are within the tumor model and near
the skin surface, and the peak temperature in the :c = 0 plane is
within the tumor boundary. ....................

Simulated E-field, SAR, and temperature distributions generated
by the RF phased array applicator and evaluated in the y =
—16mm plane of the breast tumor model. In the y = —16mm
plane, E-field peaks in tissue appear in fat near the tumor inter-
face, peak SAR values are in fat near the tumor interface and in
the tumor proximal to the applicator, and the peak temperature
in the y = 0 plane is located in fat near the tumor interface.

The thermal therapy applicator and MRI integrate system. The
MRI compatible applicator is mounted in the bench. The patient
lies on the bench with face down. This integrate system can heat
patient and monitor the temperature distribution simultaneously.

The five antenna applicator .....................
Focus Steering in Homogeneous Gel Phantom ...........

Tumor and fat phantom Fat/Thmor Heterogeneous Breast Phan-
tom for breast. The size of the steak is about 5 cm ........

One slice of MRI imageszpatient with five antenna applicator.

B field distribution in water and 3 cm below the water surface. The
E field is focused at the center of the applicator ...........

E field distribution in water and 3 cm below the water surface. The
E field focus is steered to (y r 10cm) in the applicator .......

Focus steering in experiment and simulation ............
MR image of the heterogeneous phantom ..............
Heat Focus Steering in Fat / Tumor Heterogeneous Phantom . . .
Temperature varying with time at sample point ..........

Temperature 3D distribution without optimization ........

xii

32

35

36

46
48
50

51
52

54

55
57
58
59
61
62

Figure 4.13
Figure 4.14

Figure 4.15

Figure 4.16

Figure 4.17

Figure 5.1

Figure 5.2

Figure 5.3

Figure 5.4
Figure 5.5
Figure 5.6
Figure 5.7
Figure 5.8
Figure 5.9
Figure 5.10
Figure 5.11

No optimization ............................ 63

Temperature 3D distribution with the Wiersma method. The av-
erage SAR ratio (the average SAR in tumor to the average SAR
in health tissue) is maximized and the optimized phases for five
channels are 212°, 108°, 18°, -90°and 65° ............... 64

With the W iersma method. The average SAR ratio (the average
SAR in tumor to the average SAR in health tissue) is maximized
and the optimized phases for five channels are 212°, 108°, 18°, -
90°and 65°. .............................. 65

Temperature 3D distribution with the point phase method, the E
field value at specified location [-2, 2, -5]cm is maximized and the
optimized phases for five channels are -153°, 146°, 176°, 17°and 0°. 66

With the point phase method, the E field value at specified loca-
tion [-2, 2, -5|cm is maximized and the optimized phases for five
channels are -153°, 146°, 176°, 17°and 0° ............... 67

The hybrid applicator with 4 U-shaped antennas mounted on the
four sides of the plastic tank, a planar ultrasound phased array
mounted on a side panel. ...................... 74

One slice of MRI images of the patient with contours. Those MR
images were obtained when a patient lay prone to the hyperthermia
applicator mounted in a medical couch. The contours are extracted
manually with the Matlab based program ’CT_con’. ....... 75

Phase a scheme for four focus mode scan and focusing strategy for
planar ultrasound phased array ................... 77

hybrid applicator including 4 RF antenna and US phased array . 80
Temperature results in the xz plane (y r 10mm) with RF only . . 82
Temperature results in the xz plane (y — 10mm) with US only . . 83
Temperature results in the xz plane (y - 10mm) with hybrid RF/ US 84
Temperature results in the yz plane (x T 0) with RF only . . . . 85
Temperature results in the yz plane (x *- 0) with US only . . . . 86
Temperature results in the yz plane (x :- 0) with hybrid RF/ US . 87

3D view of the temperature distribution with optimized RF source
only. The solid surface represents the 3D region over 42°C tem-
perature heating by the RF only. The external mesh contour is
the tumor. The power input magnitude for four channels are 1,
1, 1, 1 and the phase are 70.40, -63.8°, 113.5°, 0°. The SAR ratio
(tumor/ normal) is about 0.45. ................... 88

xiii

Figure 5.12

Figure 5.13

Figure 6.1
Figure 6.2
Figure 6.3

Figure 6.4

Figure 6.5

Figure 6.6

Figure 6.7

Figure 6.8

3D view of the temperature distribution with only US source, the
solid contour represents over 42°C temperature contour.The exter-
nal mesh contour is the tumor. The US phased array is located
at(.r,y,z) = [—11,0,—8.2]cm; The center of the foci ring is at
(31:, y, z) = [3.5,1,—4]cm. ......................

3D view of the temperature distribution with RF/ US hybrid
source, the solid contour represents over 42°C temperature con-
tour and the external mesh contour is the tumor. The over 42°C
region covers more of the tumor region than that with ultrasound
only and that with RF only. The total SAR of hybrid heating is
the sum of 70.14% of the SAR by E field and 67.82% of the SAR
by ultrasound pressure field. ....................

Schematics of planar array with rectangular element. .......
Focusing strategy and steering the focus ..............

A cross section of the thermal model used for temperature calcu-
lation. The material properties is listed in Table. Table 6.1. The
tumor is a sphere with 4cm diameter, 6.2 cm deep from the skin
and on the z axis. The thickness of the skin layer is 0.4cm, the fat
layer is 1.5cm, the muscle layer is 1.5cm and the viscera layer is
7.5cm. ................................

Pressure field in the focal plane and the depth plane for different
modes. The focus centers of these modes are all at [0,0,12] cm
(with a = 1). Top row is the depth plane and the bottom row is
the focal plane. From the left to the right, they are mode 4, mode
8 and mode 12. ............................

Steering the focus of mode 4 off axis and along axis, the focus
center from [0, 0, 12] cm to [10, 0, 13.5] cm (with a = 1) ......

a steering for Mode 8. The plot of a = lis shown in the center
column of Fig. Figure 6.4 .......................

Temperature distribution for Mode 12, 8 and 4 (with a = 1). The
mesh plots are the tumor model and the black solid surfaces are
the isothermal surface of 42°C. The peak temperature are 44°C
in each calculation. a) Mode 12 with focus located at [0,0,12] cm,
b) Mode 8 with focus located at [0,0,11] cm, c) Mode 4 with focus
located at [0,0,10.5] cm. .......................

The heating pattern for different frequency range from lMHz to
1.5MHz. The array size is 12cm x 12cm, the gap between array
and breast is 3cm and the focus is 11cm from the array. The top
two mesh contours are two ribs, the large external mesh contour
is the breast, the center mesh contour is the tumor model and the
solid contour are the 3D temperature iso-surface at 42°C ......

xiv

89

90
96
98

99

100

102

104

105

107

Figure 7.1

Figure 7.2
Figure 7.3

Figure 7.4
Figure 7.5
Figure 7.6
Figure 7.7
Figure 7.8

Figure 7.9

Figure 7.10
Figure 7.11

Figure 8.1
Figure 8.2

Figure 8.3
Figure 8.4

Figure 8.5

Figure 8.6
Figure 8.7

The hybrid applicator with 4 U-shaped antennas mounted on the
four sides of the plastic tank, a planar ultrasound phased array
mounted on a side panel.

Schematic of a planar array with rectangular elements. ......

Phase a scheme for four focus mode scan and focusing strategy for
planar ultrasound phased array ...................

hybrid applicator including 4 RF antenna and US phased array
Temperature results with RF focus at location 1 (0,10,0) .....
Temperature results with with RF focus at location 2: (5,0,0)
Temperature results with with RF focus at location 2: (5,0,0)

Temperature results with the combination of above 3 RF heating
with weight [0.5, 0.3, 0.5] .......................

Temperature results with US mode 8 ................
Temperature results with hybrid RF/US ..............

3 cut view of the temperature distribution with RF/US hybrid
source.

E field polarization and boundary condition ............

a) Original two dipole antenna which are perpendicular to each
other, b) Cross antenna with A/4 phase delay line .........

A cross section of one antenna applicator with breast model

3D view of the four cross antenna applicator. Each antenna is
rotated 45 degree relate to the edges of each panel and mounted
on the inner side of the panels.

A short Gaussian pulse is an input and the E field distribution on
the antenna plane are recorded in different time steps. a) t - 0, b)
t —.: T/8, c) t — T/4, d) t - T/2(1/T — 140 MHz) .........

Cross antenna model and B field distribution in the breast model

Single cross antenna applicator driven at 915MHz, the applicator
is filled with mineral oil, whose relative permittivity is about 3.
The enclosure of the applicator is same at that of the five antenna
applicator and the cross antenna is mounted at the center of the
bottom panel. The breast model is the simple breast model which
is similar to that in Chapter 3 and the four cylinders behind the
breast are the four ribs.

XV

112
113

116
117
119
120
121

122
123
124

126
131

133
134

137
139

Figure 8.8

Figure 8.9

E field distribution in the breast model and water tank. a) shows
the E field on the xz plane with y :— 0, b) shows the E field on
the yz plane with x ,— 0; There are strong E field close the an-
tenna region. They also show E field decays with increasing the
penetration depth. ..........................

SAR distribution in xz and yz plane ................

Figure 8.10 Temperature increase distribution in xz and yz plane. The peak

Figure 8.11

temperature is 44°C inside the tumor and the oil bolus temperature
is body temperature 37°C. .....................

Temperature increase distribution. for the breast model with tumor
off axis shifting -1 cm along 2: axis ..................

Figure 8.12 E field distribution in the four cross antenna applicator ......

xvi

145
147

LIST OF TABLES

Table 5.1 Material properties for the breast model used in paper. The values
of e, and a are obtained from [1] and the values of cp, n, and W
are obtained from [2] .......................... 76

Table 6.1 Material property for human tissue and the blood specific heat is
Cb = 4000J/kg/°C .......................... 99

xvii

CHAPTER 1

INTRODUCTION

1.1 The cellular and molecular basis for hyperthermia

Hyperthermia (also called thermal therapy or thermotherapy) is a type of cancer
treatment in which body tissue is exposed to high temperatures. High temperatures
can damage and kill cancer cells, usually with minimal injury to normal tissues [3]. By
killing cancer cells and damaging proteins and structures within cells [4], hyperther-
mia can shrink tumors. Hyperthermic heating of the tumor to 43°C combined with
radiation and/or chemotherapy is a proven treatment for malignant tumors [3, 5].
Hyperthermia makes some cancer cells more sensitive to radiation and harms other
cancer cells that radiation cannot damage. Hyperthermia can also enhance the ef-
fects of certain anticancer drugs. The effectiveness of hyperthermia treatment is a
function of the temperature achieved during the treatment, as well as the length of
treatment and cell and tissue characteristics [3, 4]. To ensure that the desired tem-
perature is reached, but not exceeded, the temperature of the tumor and surrounding
tissue is monitored throughout the hyperthermia treatment [5]. Various modalities
are available to heat tissue in vivo. The selected method depends on the location
and the volume to be treated. There are two ways to heat the tumor: interstitial
heating and external heating. The interstitial heating can be obtained by radio-
frequency/ microwave, ultrasound electrodes or introducing ferro-magnetic seeds in a
tumor. The external heating can be obtained by ultrasound applicators (ultrasound

array) or radio—frequent / microwave electromagnetic radiation.

Hyperthermia has a cytotoxic effect on cells and the highest heat sensitivity is
observed during the mitotic phase (M-phase). There is a great diversity of molecular

mechanisms of cell death following hyperthermia after being exposed to tempera-

ture greater than 43°C. Cells can develop thermotolerance against hyperthermic cell
death. This thermotolerance might be attenuated under some conditions such as
lowered intracellular pH and some forms of drug resistance [4].

Hyperthermia temperature over 42°C have some special features in tissue and cells
in viva that induce alterations of tumor blood flow and tumor micro environment.
Hyperthermia enhances the cytotoxicity of various antineoplastic agents (thermal
chemosentization). Hyperthermia also has effects on the cell membranes, cytoskele-

ton, cellular proteins, nucleic acids, and cellular immune response.

1.2 The breast cancer

Breast cancer is the most prevalent cancer among women and affects approximately
one million women worldwide. In the United States, among women, breast cancer is
the most common malignancy diagnosed and the second leading cause of death from
the cancer. Breast cancer accounts for 30 percent of all female cancers in the UK
and approximately 1 in 9 women in the UK will get breast cancer sometime during
their life (http://www.who.int). Of all women diagnosed with breast cancer, 20%
have locally advanced breast cancer (LABC). Women with locally advanced cancer of
the breast who are exposed to standard treatments experience a 5 year survival rate
of 50%, and this rate declines to 30% after 10 years. The treatment of the disease
depends on the tumour type and the stage of disease -— how far the cancer has spread
to involve either lymph glands or other organs in the body. There are various ways
a cancer can be staged and classified. Stages or classifications of breast cancer are
divided into three groups (http://www.breastcancersource.com/): Early or Operable

breast cancer, Locally advanced breast cancer and Advanced breast cancer.
1.2.1 Early or operable breast cancer

Early breast cancer refers to cancer that is confined to the breast and/or the lymph

glands in the axilla (arm pit) on the same side of the body. At this stage, the tumor

size is usually smaller than 2 cm. The stage tumor can be removed by the surgery

1.2.2 Locally advanced breast cancer

Locally advanced breast cancer has spread beyond the breast and axillary lymph
glands and involves the skin or the chest wall of the breast. These cancers tend to
have a worse outlook than early breast cancer and are usually best initially treated by
drug therapy or radiotherapy rather than surgery. In locally advanced breast cancer,
the skin of the breast can either be directly involved by cancer or it is swollen or
red. These changes occur because cancer cells enter the fluid channels that drain the
breast (lymphatics) and block them, which causes the skin of the breast to be swollen
and look like the skin of an orange. Locally advanced breast cancers were initially
treated with surgery but this treatment was successful in only about 30 percent of
patients. In the remainder, the cancer recurred in the areas immediately next to

where the surgery was performed

1.2.3 Advanced breast cancer

In this statge, breast cancer has spread beyond the breast to other parts or organs
of the human body such as lymph glands in the neck, bone, lungs, liver and even in

brain.

1.3 The RF / Microwave modality for hyperthermia

Radio frequency (RF) radiation is one of the most versatile hyperthermia modalities;
absorption of the electromagnetic energy in tissue causes a temperature elevation. The
spatial control of microwave and radio frequency systems is limited due to physics.
At higher frequencies some control is possible but penetration depth is extremely
limited; while at lower frequencies with high penetration depths, spatial control is
limited due to the significant wavelength. The wavelength in the muscle tissue at

140 MHz is about 22 cm. RF phased arrays for hyperthermia are now commercially

available at several institutions, such as BSD—2000 system. Currently, there are no
commercial systems particular for breast cancer with external, radiative RF arrays.
In order to obtain high performance of breast tumor treatment, RF arrays need to

be placed around and close to the tumorous breast.

1.4 The ultrasound phased array modality for hyperthermia

Ultrasound has numerous applications in medicine including ultrasound imaging and
therapy. The therapeutical applications of ultrasound includes high intensity focused
ultrasound (HIFU) and ultrasound hyperthermia. These applications are based on
the absorption of acoustic wave. HIFU usually heats the tissue heated over 50°C and
destructs the tumor cells directly. In ultrasound hyperthermia treatment, tissue is
usually heated to temperature between 41745°C for 30‘120 mins.

Among different ultrasound therapy system, the ultrasound array system utilize
electronic and/or geometric focusing to increase the intensity gain and improve the
penetration depth. Ultrasound systems generate precise focal spots, usually much
smaller than ordinary tumor dimension (Hunt, 1990). To expand the heated region,
mechanical scanning or electronic scanning are utilized to improve the tumor cover-
age. But ultrasound phased array posses greater potential for improving the heating
performance(Hynynen and Lulu 1990) than mechanical scanning. There are various
ultrasound phased array prototypes including sector-vortex, NxN square, cylindrical
section and spherical section.

The research in this dissertation is focus on computational simulations of breast
cancer hyperthermia treatments of breast cancer with RF phase array and RF/ US
phased array applicator. The electromagnetic field distribution and properties are
investigated to help the treament planning. The hybrid applicator, which combines
RF modality and ultrasound modality togeter, takes advantage from both modalities

and is proposed here to be the future clinical hyperthermia system. A new ultrasound

focuing method, which combines sector-vortex method and phase delay method, is
designed here for 2D planar array to generate large focus for ultraound heating pur-
pose. This dissertation is organized into 9 chapters. Chapter 2 introduces the main
simulation methods used in this whole dissertation. Chapters 3 and 4 describe the
inestigation on electromagnetic field distribution and properties in four antenna RF
applicator and five antenna RF applicator. Chapters 5 and 7 describe two designs of
RF/ US hybrid applicator. Chapter 6 discusses the new focusing method for planar
array to generate large focus. A new type antenna is dicussed in Chapter 8, which
can generate circular polarized wave in tumor region. Finally, Chapter 9 summarizes
the conclusions drawn from this work and provides some suggestions for the future

work.

CHAPTER 2

SIMULATION METHODS AND ROUTINES

Computer simulations and treatment planning routines require several computational
procedures for field calculation and optimization. The field calculations simulate the
eletromagnetic and ultrasound power depositions for each antenna or ultrasound el-
ement and optimization routines maximize the tumor heating by selecting the mag-
nitude and phase for each channel. In this thesis, the finite element method (FEM)
and the finite difference time domain (FDTD) method calculate the electric fields
radiated by the antennas, the fast near field method (FNM) and angular spectrum
approach (ASA) calculate the acoustic pressure field, and the bio-heat transfer equa-

tion computes the temperature distribution.

2.1 Electromagnetic numerical calculation

2.1.1 The Finite Element Method (FEM)

The finite element method is one of the most successful frequency domain computa-
tional methods for electromagnetic simulations. It combines geometrical adaptability
and material generality for modeling arbitrary geometries and materials of any com-
position [6]. The latter is particularly important to current project since the human
body is composed of fat, ribs, heart, lung, muscle, etc. which have different material
properties; In addition, the applicator consists of copper antennas, a plastic enclosure,
and water. The problem of electromagnetic analysis is actually a problem of solving a
set of Maxwell’s equations subject to boundary conditions. The electric field satisfies

the vector wave equation with a differential form given by [7]:

vx <—VXE) —k3€.E= (2.1)

Where E is the vector of the electric field, [so is the wave number in free space, It.
is the relative permeability and c. is the relative permittivity. A boundary condition
must also be specified for a unique solution. Boundary conditions include Dirichlet
condition, Neumann condition, impedance condition, radiation condition and higher-
order conditions. The principle of the FE method is to replace an entire continuous
domain by a number of sub-domains in which the unknown function is represented by
simple interpolation functions with unknown coefficients [7]. The system of equations
is formulated by the Ritz method (a variational method) or the Galerkin’s Method
(a weighted residual method), the latter is presently more popular.

The finite element analysis includes the following basic steps:

1. Subdivision of the domain,
2. Selection of the interpolation functions,
3. Formulation of the system of equations,

4. Solution of the system of algebraic equations.

The radiation boundary condition is also essential in this problem. Because generating
a mesh or grid in an infinite region is impossible, truncating the problem into a finite
region is necessary, and the truncated boundary is the radiation boundary.

Most computations of the electric field (E-field) generated by radio-frequency (RF)
electromagnetic devices designed for hyperthermia presently utilize either the finite
difference time domain (FDTD) method [8] or the finite element method (FEM)
[9, 10, 11]. The finite element method provides an advantage in E—field modeling by
facilitating the geometric adaptability of the computational mesh. The finite element

method solves the weak form of the vector wave equation [7]

/'[#i(v x E) - (v x W) —— rat-.12: - W]dV (2.2)

2. W-(ithxEMS— f-WdV.

. so . v

This expression describes the E-field throughout the volume V subject to a given set
of boundary conditions on the surface 30 enclosing the volume V. In Eq. 2.2, u, is the
relative permeability, W is a weighting function, e, is the relative permittivity, k0 is

the wavenumber in free space, and f is the electromagnetic source function given by

1
Hr

In Eq. 2.3, Z0 is the free space wave impedance, J is the electric current density,
and M is the magnetic current density. The first item in the right hand side (RHS)
of Eq. 2.2 is the boundary integral, which vanishes on a perfect electrical conductor
(PEC). This term represents the absorbing boundary condition on a truncated surface

for an Open structure [7].

Finite element simulation results are generated by HFSS Version 8.0, which is a
commercial software package produced by Ansoft Corp. (Pittsburgh, PA). In HFSS
simulations of the electric field generated by the phased array applicator, the compu-
tational model consists of a three dimensional mesh of tetrahedral elements. HFSS
provides a convenient interface for defining geometric structures and assigning mate-
rial properties of each structure, then HFSS automates the generation of the mesh
based on the structural properties of the overall system. In each tetrahedral element,
the initial lengths of the edges are determined from the material properties, and the
mesh is automatically refined based on the results of the computed E—field. The con-

venient user interface provided by HFSS allows an experienced operator to quickly

define a new applicator structure and phantom geometry, generate the corresponding

mesh, and solve for the resulting E—fields.

Finite element simulations of the phased array applicator truncate the boundary
of the defined mesh with an absorbing boundary condition (ABC). The ABC for these
simulations is implemented as a second order radiation boundary condition located
in the air surrounding the outside of the applicator. In HFSS, the ABC elements are
cuboid that surround the computational volume. The absorbing boundary is located
roughly /\ / 10 away from the applicator, and since the antennas radiate primarily into
the water tank, the E—field decays significantly before reaching the ABC. Results (not
shown) were also obtained with an absorbing boundary located A from the applicator.
The difference between the computed fields obtained with the boundaries at A/ 10 and
/\ is approximately 1.24%, therefore the distance from the water tank to the absorbing
boundary is A/10 for all simulations. By placing the ABC closer to the applicator,
the size of the mesh is reduced, which in turn diminishes the computer memory

requirements and the total computation time.

The E-field distribution is computed for each antenna separately, and then the

results are superposed. The magnitude of the total E-field is computed according to

N
]E(:r,y,z)| = ZUn(:1:,-y,z)1,, , (2.4)
n=1

 

 

where 1,, is a complex number representing the amplitude and the phase of the n-
th antenna input, and the vector I steers and focuses the E-field. In Eq. 8.2, Un
represents the electric field contribution produced by the n—th antenna for a unit
input excitation (i.e., 1,, 2 140°), N = 4 is the number of the antennas in the phased
array applicator, E represents the total electric field, and (1:,y,z) represents the
Cartesian coordinates of the simulated E-field. In the simulation model, the voltage

excitation is located in the center of the antenna, and each antenna input is modeled

as a lumped gap source located at the center of each input port with an impedance
of 509. Once the 3D E-field is computed, the SAR is calculated from the amplitude

of the total E-field according to

a(.r, y, z)

|E($, y, Z)|2, (2-5)

where 0 represents the conductivity and ,0 represents the density for each tissue type.

2.1.2 Finite Difference in Time Domain (FDTD)

FDTD is another numerical method solving the Maxwell Equation. The FDTD
method is a numerical solution for Maxwell’s curl equations, which are based upon
volumetric sampling of the unknown electric field and magnetic field within and sur-

rounding the structure of interest over a period of time.

The FDTD method is efficient and accurate for electromagnetic wave interaction
problems in unbounded regions. For our problems, and absorbing boundary condition
(ABC) must be introduced at the outer lattice boundary to simulate the extension of
the lattice to infinite. This is achieved for FDTD simulations by the perfectly matched
layer(PML). Any plane waves of arbitrary incidence, polarization and frequency are

matched at the boundary and are absorbed by the PML.

In this thesis, the E field propagations in and out of antenna in the time domain
are calculated by the commercial software XFDTD (Remcom, Inc. State College,
PA). XFDTD can handle 3D models of antennas and applicators, by subdividing
these into equal-space meshes with 1 mm grid size (about 0.005 @ 140MHz in water)

and automatically adding a PML boundary to the region of interest.

10

2.2 Ultrasound pressure calculation

2.2.1 The Fast near field method (FNM) for calculating pressure field

There are several different methods for calculating ultrasound fields generated by rec-
tangle or circular piston. The most widely used approach for diagnostic imaging is
the impulse response method derived by Oberhettinger and Stepanishen and Lock-
wood and W illette. This approach computes the pressure quickly but some numerical
problems near the edge of the transducer. Other methods include the point source
superposition methods (Zemanek) and the rectangular radiator method (Ocheltree).
These two methods compute the field generated by circular and rectangular pistons
by dividing these into small sub-elements. Both methods generate relatively large

errors, and the computation times are relatively long.

The fast near field method (FNM) is derived directly from the impulse response
method. The fast near field method eliminates the singularities that are produced by
the impulse response method, thus the fast nearfield method (McGough) reduces the

computation time and the peak normalized error.

2.2.2 Angular spectrum approach (ASA)

The angular spectrum approach computes acoustic pressures in a 3D volume in par-
allel planes. This approach is based on the theory that waves diffracted from finite
apertures can be approximated by a sum of plane waves traveling in different direc-
tions [12]. The two dimensional (2D) Fourier transform decomposes the diffracted
wave into transverse components, each of which has a vector angular frequency asso-
ciated with the direction cosines. The ASA is essentially a Green’s function approach
that computes the field by multiplying the source and the Green’s function (propa-
gator) in the spectral domain. The accuracy and numerical efficiency of this method
has been studied widely. For example, Christopher [13] described a ray theory trun-

cation method to reduce the errors for radially symmetric fields. Wu]14, 15, 16, 17]

11

discussed the optimal angular ranges in terms of the spatial aliasing errors.

The simulation stage often involves a large number of pressure field calculations,
which are extremely time consuming using conventional analytical methods. However,

much faster calculations can be achieved using the angular spectrum approach.

2.3 Thermal analysis

The temperature distribution is determined from the computed SAR. distribution

using the steady—state Bio-Heat Transfer Equation (BHTE) ,

0 = HVQT + cbI-i«"b(T — Tb) + SARp. (2.6)

In Equation 2.6, H is the thermal conductivity (W / m/°C), T is the tissue tempera-
ture (°C), 0,, is the specific heat of blood (J/kg/°C), Wb is the blood perfusion rate
(kg/m3/s), Tb is the temperature of blood (°C), a is the electrical conductivity, SAR
represents the specific absorption rate and pis the density of the tissue (kg/m3). In
these simulations, the steady-state finite difference solution of Eq. 2.6 is obtained
for the resulting tissue temperature T. The finite difference calculations maintain
the water surrounding the breast at a constant temperature, and a radiation bound-
ary condition is enforced at the remaining interfaces. Subject to these boundary
conditions, the finite difference model is evaluated within a computational grid that
contains the entire 3D model of the breast depicted in Fig. Figure 8.3. The grid
elements are cuboid, and each grid location contains SAR values obtained from the

truncated finite element mesh defined for E—field calculations.

The temperature distribution is computed from the SAR. distribution using the
Bio-Heat Transfer Equation (BHTE). The static Pennes’ bioheat equation (Eq.2.7)

are given by:

12

dT

ppCpEt— = V(I{.p v T) + CbI’I"yb(Ta — T) + SAR ' pp (2.7)
0 = v(rtp v T) + Obi/VAT, — T) + SAR - pp (2.8)

In Equation 2.8 and 2.7, T is the tissue temperature, t is the time, CI, is the tissue
specific heat (J/kg/°C), 5,, is the tissue thermal conductivity (watt/m/°C), Cb is the
specific heat of blood (J/lrg/°C), Wb is the blood perfusion rate (kg/m3/s), Tb is the
temperature of blood (°C), [2,, is the tissue density (kg/m3), and SAR represents the

specific absorption rate.

The final temperature can be computed with a steady-state finite difference
method. The 3D discretization manipulation of Eq. 2.8 is shown in Eq. 2.9. In
order to speed up the calculation, successive over-relaxation is utilized in the simula-
tion. The time-varying temperature distribution can be calculated by the discretized
form of Eq. 2.7 with it’s discretization form in Eq. 2.10. Since the water is main-
tained a constant temperature, the thermal computational region is cuboid truncated
from the FE region and this cuboid region contains the whole breast model and some
water around the breast model. The six boundary faces of the cuboid region are set

as a thermal radiation boundary condition for BHTE calculations.

 

7".“ = ”P
I’J’k 61%,, + ((25wabe
(62:)21012 SA m Tm I‘m Tm Tm m
X _K Rwyk + i+1.j.k + i—1.j,k + i,j+1.k + 131—1,1: + i.j.k+1 + mule—1
t

(2.9)

13

_ 6,11"be _ 66m, X T" k + 115.43.. _k + art/,0,
MN .3.

 

 

T..’=1 Tb

1, ,k r
J ppCp ppCp°zf CP 10120!)
(SCH!) n n
——p C 62 [Trude + Tardy}.- + Ti?j+l,k + TIE—Lie + 73x“ + flak—1:] (2-10)
P P I

In these two equations, m indicates the iteration step, n is the time step, 6,, is the
step in space, and (it is the time step. 773,, is the temperature at step m for the

spatial point [2' - 61.,j - (5,, k - 6,].

2.4 E—field calculation with HFSS

The FEM calculation of the E-field in the simulation model is performed by HFSS.
Modeling the patient in HFSS is the most time-consuming step. The contours, which
are extracted from MRI or CT images, are drawn in HFSS and connected into solids
(by macro commands including connect, coversheets, and stitch). Sometimes the
connect command does not work very well, when connecting two contours into a solid
object, which is caused by the irregular point distribution along the contours. The
average or maximum edge length of the mesh should be kept smaller than a tenth
of a wavelength in that material. The lumped gap source is the power input for
the antenna with the rectangular shape and the calibration line and impedance line
are assigned inside the port and attached to the antenna. The external radiation
boundary is ideally at least A/ 10 from the antenna. After the E field is calculated,

the result is exported to another program for thermal analysis.

2.5 Focusing strategy

The E—field is typically focused by optimizing the SAR [18, 2, 19, 20], and focusing
is also achieved by directly optimizing the temperature [21]. Here, the phased array

is focused by maximizing the constructive interference produced by the individual E-

14

field components at a single point. A focus is generated when the magnitude squared

of the total E-field

4 4

IEI° = Z Z (Umlm) (U..1..)* (2.11)

m=1 n=1
is maximized relative to the sum of the squared input magnitudes Zi=1i1ni2- Defining

the individual components of E through a matrix-vector product yields [21]

 

 

 

 

f , ‘ ' . . . . ,2 11
EA UIA U; U3\ U;
12
13” = U,” U," U3” If . (2-12)
13
E2 LU? U23 U32 0,2
_ _I a 1’4

 

 

which is equivalent to _E = g1, where a single underbar indicates a vector quantity
and a double underbar indicates a matrix quantity. After this matrix-vector notation
is applied to Eq. 2.11 and the result is normalized with respect to the sum of the

squared input magnitudes (I, I) = 1*‘1, the optimization problem becomes

(E, E) lttgttgl
1*‘1 '

 

||
:5
p—4
"-th
X

(2.13)

The extrema of the quadratic form in Eq. 2.13 are the eigenvalues of g"g, and the

maximum value Am” of Eq. 2.13 is achieved by the eigenvector Ima of g"g [21]

1:

that satisfies

gig; =/\ I (2.14)

max mart: —ma.r '

The input In,“ maximizes the power deposition at a single coordinate in the tumor
target. As a result, constructive interference is also achieved in nearby regions, and

the surrounding tumor tissue is also heated. The optimal excitation vector Imam

15

is selected such that the peak power deposition is located within the tumor in a
location proximal to the applicator. Focusing in this location reduces hot spots in
normal tissues and other problems related to penetration depth.

The expression in Eq. 2.14 represents the Optimal solution when the magnitudes
and phases of the excitation are unconstrained. If the magnitudes of the entries in
I are specified in advance, then the expression in Eq. 2.11 is a nonlinear function of
the phase variables arg (In) 2 on. Defining the arguments of the antenna outputs for
unit input excitations as arg (Uji) = 0;, arg (Uzi) = 63;, and arg (U5) 2 6% yields

an equivalent objective function

3

2
Z Z |1,,| um) {|U;}'| [Ugfl cos (9;? —— 6;}; + 3,, — as...)
n=1

m=n+1

+ lU.’.’| lUr’élCOS (93.” -6.’.. +¢>n - a...) + lUf| lUiICOSWf — 05. +¢n —¢...)}

3
+ 211.1114 {lUfl IUfICOSWff - 9f + en)

n=1

+ lUi'l lUI'Icos(6i ~63” w.) + IUE

 

 

U42] cos (OZ —- 942 + (1)")} (2.15)

71

when the reference phase for the fourth input is 3., = 0°. The expression in Eq. 2.15
is readily maximized by iterative solvers such as the Matlab function fmz'nvnc, which
quickly converges to the optimal values of 3,, gig, and (53 that produce a global
maximum. Solutions to both the linear and nonlinear objective functions in Eqs. 2.14

and 2.15, respectively, generate viable focal patterns for thermal therapy.

16

CHAPTER 3

FOUR CHANNEL RF PHASED ARRAY APPLICATOR

External heating devices appropriate for deep hyperthermia in the intact breast in-
clude ultrasound phased arrays [22] and radio-frequency (RF) electromagnetic phased
arrays [23, 24, 25, 26]. Ultrasound is an appropriate local modality for heating small
targets in the breast (up to about 2cm diameter [27]), whereas heat generated by RF
electromagnetic devices is delivered regionally across a much larger area. RF phased
arrays have been developed previously for deep hyperthermia in the pelvis [28] and
in the extremities [29, 23], respectively. These arrays apply RF frequencies in the
60-140MHz range for increased penetration while delivering heat to deep targets. A
microwave phased array system has also been constructed for thermal therapy in the
breast [30]. The microwave phased array system requires breast compression because

of the shallow penetration achieved at 915MHz.

To exploit the greater penetration depths afforded by RF applicators, a four—
channel RF phased array applicator has been developed for hyperthermia cancer
treatments in the intact breast. This phased array operates at 140MHz, which in-
creases the penetration depth substantially over that achieved by microwaves. This
RF phased array has been characterized with measurements of the electric field, and
these measurements are consistent with the fields predicted by finite element simula-
tions. Additional finite element and bio-heat transfer modeling results suggest that
this array is capable of delivering therapeutic heat in an idealized model of the breast.
These measurement and simulation results show that this RF phased array system
can focus the electric field within a water tank and in simulated tumor targets in the

breast.

17

3.1 RF phased array applicator and E field measurement system

3.1.1 Applicator geometry

A photograph of the prototype four-antenna RF phased array applicator is shown
in Fig Figure 3.1. This design is an adaptation of previous cylindrical phased ar-
ray geometries [31, 23], where only the portion of the array that directs RF energy
to the breast is retained. The applicator consists of a water tank enclosure with
four pairs of end-loaded dipole antennas mounted on the inner surface of the water
tank. The tank enclosure consists of 6 solid Lexan (GE Polymerland, North America:
www.gepolymerland.com) side panels and a Lexan top piece with a large opening. RF
antennas are mounted on four of the rectangular side panels, and the two remaining
side panels are irregular pentagons with one axis of symmetry. The four rectangu-
lar side panels are 12.8cm by 21.5cm, and the pentagonal side panels are 12.8cm by
12.8cm by 12.8cm by 12.8cm by 33.2cm. In each pentagonal side panel, the obtuse
angle measured at the lowest point on the tank is 112°, the two obtuse angles mea-
sured at adjacent vertices are 153°, and the two remaining acute angles are 61°. Each
Lexan panel is approximately 5mm thick. The top opening is 33.2cm by 20.50m, the
overall size of the tank enclosure in the x, y, and 2 directions is 21.5cm by 33.80m
by 18.4cm, respectively. For electric field measurements and patient treatments, the

water tank is filled with deionized water.

The RF phased array applicator consists of four end-loaded dipole antennas that
are composed of attached segments of 0.03mm thick copper foil [23]. This antenna
geometry forms a transmission line with the two horizontal segments, and these in
turn guide the RF energy to the vertical sections. For each of the antennas shown in
Fig. Figure 3.1, the RF voltage input is applied at the center of the end-loaded dipole.
Impedance matching is implemented with a coaxial cable stub, so each antenna is

driven at 140MHz with a low return loss.

18

 

Figure 3.1. RF phased array prototype with four antennas designed for hyperthermia
treatments in the intact breast.

19

Signal —>
Generator

t.
V? I: Mr 4"
‘Joltmet er

<f— PC

/ r .1 -‘ 4 . ‘ /
I‘-‘II11I1}IIIBXE‘I‘," ’ . a], i i I
Stretch Box I ‘ Y

PJ’IIIZIIITIBI‘S l4)

 

Figure 3.2. RF phased array amplifier system. This rack-mounted system consists
of a signal generator, a vector voltmeter, a multiplexer/switch box, and four RF
power amplifiers. The signal source generates a common excitation frequency for
each of the amplifiers, the vector voltmeter provides phase and power feedback, and
the multiplexer/switch box combination controls the inputs to the RF amplifiers. In
turn, the RF amplifiers drive the individual antennas in the phased array applicator.

20

3.1.2 Amplifier system

The main components of the RF phased array amplifier system include a signal gen-
erator, a vector voltmeter, a multiplexer, a switch box, and four RF power amplifiers.
The signal source for all channels is an HP (Palo Alto, CA) 8647A signal generator.
An HP (Palo Alto, CA) 8508A vector voltmeter measures the output phase and power
as well as the reflected power. These measurements deliver phase and power feedback
to the controlling computer via GPIB. The multiplexer/ switch box combination spec-
ifies the amplitude and phase of the input to the RF power amplifiers with analog
attenuators and phase shifters through a Kontron (San Diego, CA) AOB6-P digital
to analog converter (DAG) card that is installed in the controlling computer. The
RF power amplifiers are LCF Enterprises (Post Falls, ID) model B020 rack-mount
amplifier systems that operate in the 120-170MHz frequency range with a maximum
power output of 150W. The output of each RF power amplifier is attached to a cir-
culator, a low-pass filter, and a directional coupler that provides a direct connection
to the antenna and a feedback connection to the vector voltmeter. The controlling
computer is a Pentium II system with 128MB of RAM running the Windows 98 op-
erating system, and the system control software is written in Visual Basic (Microsoft

Corporation; Redmond, WA).

3.1.3 Measurement system

The electric field is measured with an E—field probe attached to a 3-axis positioning
system. The measurement system hardware consists of 1) a LinTech (Monrovia, CA)
M1-102412 3-axis positioning system, 2) three EP-400 E-field array probes from BSD
Medical Corporation (Salt Lake City, UT), 3) an HP / Agilent (Palo Alto, CA) 34970A
Data Acquisition/ Switch Unit, 4) three Digiplan (Poole, Dorset, United Kingdom)
PDX15 motor drivers (one for each axis), 5) a custom water degasser, and 6) a

personal computer (PC). The E-field mapping software is controlled through a Matlab

21

 

Figure 3.3. Three E-field array probes are attached to a Pleadglas rod for measure-
ments of the electric field produced by the RF applicator. This probe arrangement
is scanned across a rectilinear grid within the water tank by a computer-controlled
positioning system. The measurements obtained with these scans characterize the
E-field distribution generated by the RF phased array.

22

graphical user interface Operating on a Windows 98 platform. Low-level interface
routines are provided by the Matlab Instrument Control toolbox. The EP-400 E—field
probes are connected to the HP 34970A, which transfers measured voltage values to
the PC via GPIB. The stepper motor positioning system is controlled by the Digiplan
motor drivers, and these communicate with the PC through a daisy-chained serial port
connection.

The 3-axis probe combines three separate EP—400 E-field array probes attached to
a Plexiglas rod in the orthogonal arrangement shown in Fig. Figure 3.3. Each probe
contains three miniature diodes, so the 3-axis E-field probe consists of nine miniature
diodes. This arrangement permits averaging of the individual E-field components,
which improves the signal-to-noise ratio for each E-field measurement. Each probe is
offset from the center of the Plexiglas rod, and the control software compensates for

the offset distance before the signals are averaged.

3.2 Simulation Methods

Fig. Figure 3.4 contains a geometric description of the simulation model for the
four antenna phased array applicator. In the finite element model, the size of the
Lexan tank matches the dimensions of the prototype applicator. The Lexan tank
is surrounded by air and filled with deionized water. In these simulations, the per-
mittivity of Lexan is c, = 2.9, the conductivity of Lexan is 0 = 0, the relative
permittivity of deionized water is 6,. = 76.5, and the conductivity of deionized water
is a = 0.00018/m. The four end-loaded dipole antennas, which are attached to the
Lexan tank, are c0pper strips excited by a lumped gap source. The resulting E-field
is evaluated in the water tank and in a tissue phantom model suspended within the
water tank.

The simulated tissue phantom in these finite element simulations is the simplified

3D breast model shown in Fig. Figure 3.5. This 3D model, which was in part

23

  
  
 
    

g-‘IOT
V-15r ‘i
N :
-20~2‘ _
-25jijii

X(cm) - '20

 

 

Figure 3.4. Simulation model for the four antenna RF phased array applicator. The
geometric model, which has the same dimensions as the applicator in Fig. Figure 3.1,
defines the input parameters for finite element simulations.

24

motivated by the axis-asymmetric model in [32], consists of an extended hemisphere
and a spherical tumor mass that is offset somewhat with respect to the central axis
of the breast model. The model in Fig. Figure 3.5 includes a skin layer, a fat layer,
and a muscle layer. In Fig. Figure 3.5, the radius of the hemispherical breast is
7.5cm, the radius of the simulated tumor mass is 2.5cm, the thickness of the skin
layer is 0.5 cm, the muscle layer is about 4.2cm thick, and the fat layer is 1cm thick.
In this model, skin, breast, and fat share the same material property values [33].
The material properties for each tissue type are listed in Table Table 5.1. In FEM
simulations, the applicator in Fig. Figure 3.4 is filled with deionized water, and the
breast model is suspended in deionized water. The E-field is then computed in the
breast and all surrounding regions, including the deionized water, the lexan tank, and

the surrounding air space.

3.3 Results

3.3.1 E-field measurements in the water tank

The electric field is focused in the water tank with an appropriate selection of phases
and amplitudes. A focus is generated by iteratively modifying the phase, and the
combination of phase values that produce the maximum peak magnitude of the mea-
sured electric field in the desired location is selected as the input. This focusing
approach is required because of the phase errors that are typically present in RF
phased array systems [34]. These phase errors are in part caused by the unavoidable
crosstalk between antennas that are coupled to a resonant cavity (the water tank)
filled with low-loss material (deionized water) and in part caused by errors inherent in
the controlling hardware [34]. After the desired combination of input phases and am-
plitudes is determined, the resulting E-field distribution is measured with the E-field
probe depicted in Fig. Figure 3.3.

A focus was achieved in the center of the water tank with input phase settings of

25

   
   

. . .Sklni
Fat layer.
Muscle.
Breast
.Tumor

.20. mm'i

Figure 3.5. Schematic of the breast model defined for FEM simulations. The breast
is modeled by a hemisphere with a 75mm radius, and a spherical tumor model with a
25mm radius is located inside the breast. The hemispherical breast model is attached
to a 5mm thick skin layer, a 25mm thick fat layer, and a 42mm thick muscle layer.

26

0°, —60°, —30°, and —90° applied to channels 1, 2, 3, and 4, respectively. Forward
power settings of 40W were applied to each channel, and for the centered focus,
the measured values of the reflected powers were 3.4W, 3.9W, 10.4W, and 1.7W
for channels 1, 2, 3, and 4, respectively. The measured E-field distribution for the
centered focus is shown in Fig. Figure 3.6. In each figure, the total E—field is the
square root of the sum of the squares of the individual component magnitudes. The
measured E-field values in Fig. Figure 3.6 are indicated by a star (‘*’) symbol, and the
intermediate values indicated by dashed lines are obtained from the results of cubic
spline interpolation. Fig. Figure 3.6a contains the measured E-field along the x-axis,
Fig. Figure 3.6b shows the measured E-field along the y-axis, and Fig. Figure 3.6c
demonstrates the measured E-field along the z-axis. Figs. Figure 3.6a and Figure 3.6b
show that the desired peak is located near the center of the water tank. Fig. Figure
3.6c indicates that the maximum measured value along the z-axis is located 3cm
below the water surface and the minimum measured value is located 12cm below the
water surface.

A steered focus was generated in the water tank with input power values of 50W
for channels 1 and 2 and 25W for channels 3 and 4. The phase settings for the steered
focus were 0°, 0°, —130°, and —130° and the measured reflected powers were 65W,
9W, 0.4W, and 14W for channels 1, 2, 3, and 4, respectively. These reflected power
values emphasize the need for circulators in the amplifier system while indicating
that focusing is preferentially achieved in the center of the water tank. The measured
E-field distribution is shown in Fig. Figure 3.7a, which shows that the peak is located
at (0,4, —3)cm.

3.3.2 E—field simulations in the water tank

Simulation results for specified input settings are obtained by superposing the contri-
butions from each antenna. In these simulations, the E-field is computed separately

for each antenna using a common finite element mesh. This mesh is defined by the

27

_A
N

 

 

 

 

 

 

 

 

 

1.2
1 1.
08 '/*\
0.8 i
:72 " E
u'i 0‘ :35 0.6]
3 04 i E 0.4)
E 7:
£02 r: games 1 § oz) r———~——- ,
'_ c ' +- Experiment
‘ l _Si'flu'mm L———_S_l4rnulatlon ]
1’4 -2 o 2 3 6 lie —5 3 5 10
X(cm). . Y(cm)
(a) Measured (dashed line) and sunu- (b) Measured (dashed line) and Simu-

lated (solid line) E-field values in the lated (solid line) E-field values in the
water tank plotted with respect to the x water tank plotted with respect to the y

 

 

 

 

 

 

 

 

 

 

coordinate where y = 0 and z = —3cm.. coordinate where :1: == 0 and z = —3cm.
1.2 . g
- s- Experiment =
1 _ —Simulation 1
0.8* i
E
.2
"i'
m 0.6—
N
a 0.4"
E
b
O
C 02*
-Q5 -10 -5 0 5

(c) Measured (dashed line) and simulated (solid line) E-field val-
ues in the water tank plotted with respect. to the z coordinate
where :t' = 0 and y 2 00m.

Figure 3.6. E—field distributions generated by the RF phased array depicted in
Fig. Figure 3.1. The applicator prototype operates at 140MHz, producing a focus
in the center of a tank filled with deionized water. The E-field measurements are
performed by the apparatus depicted in Fig. Figure 3.1, and the E—field is computed
with the finite element method for uniform phase and amplitude inputs.

28

model geometry and the material properties of water, Lexan, copper, and air, and for
the breast model, material prOperties of different tissue types are also included. The
finite element mesh is initially generated by HF SS and then automatically refined.
Additional manual mesh manual refinement is applied as necessary near material in-
terfaces, in regions where the electric field is changing rapidly, and in regions where
node spacings are otherwise too large. After the mesh generation is completed, the
average distance between adjacent nodes is between A/ 10 and A/ 75, depending on
the model geometry and material properties, with higher node densities near radiat-
ing sources and material interfaces. The finite element mesh includes an absorbing
boundary condition, where the shortest distance in air from the edge of the water
tank to the absorbing boundary is 20cm in the x-direction, 18cm in the y-direction,
20cm measured from the top of the water tank in the z-direction, and 12cm measured

from the bottom of the water tank in the z-direction.

A typical "finite element. simulation consists of an initial automated mesh genera-
tion, 11 automated mesh refinements, and one or more additional manual refinements
resulting in a mesh containing up to 73,000 tetrahedra. After each mesh refinement,
the E-field is computed with the finite element method using an edge element formu-
lation. The computation time associated with this sequence of calculations is about
8 hours on a 2.4GHz Pentium 4 personal computer with 1GB RAM running the

Windows XP operating system.

A simulated focus in the center of the water tank is generated with uniform power
and phase inputs. The simulated inputs are normalized with respect to the peak input
power, so the input power applied to each channel is 1W, and the phase value is 0°
for each input channel, shown in Fig. Figure 3.8. The simulated E-field distribution
for the centered focus is shown in Fig. Figure 3.6. Each plot is also normalized, which
facilitates comparisons between simulated and measured values. The simulated E-

field values in Fig. Figure 3.6 are indicated by solid lines, where Fig. Figure 3.6a

29

contains the simulated E-field along the x-axis, Fig. Figure 3.6b shows the simulated
E-field along the y-axis, and Fig. Figure 3.6c demonstrates the simulated E-field along
the z-axis. Figs. Figure 3.6a and Figure 3.6b show that the simulated peak is again
near the center of the water tank. Fig. Figure 3.6c indicates that the simulated peak
value along the z-axis is also located about 3cm below the water surface.

The simulated focus is steered to the right with input powers of 0.5W, 0.5W, 1W,
and 1W and input phases of —87°, —91°, —-66°, and 0° applied to channels 1, 2, 3,
and 4, respectively. The input power values in these simulations are proportional
to the input powers that generated the measured E«field distribution depicted in
Fig. Figure 3.7a, and the input phases maximize the objective function in Eq. 2.15 for
a focus selected at (0, 4, —3)cm. This target represents the location of the peak E-field
magnitude and the maximum constructive interference. The phases that generate this
result were obtained from the solution to Eq. 2.15, which converged to the optimal

solution after 170 iterations.
3.3.3 Breast model simulations

Simulations that combine the applicator model in Fig. Figure 3.4 with the breast
tumor model in Fig. Figure 8.3 demonstrate an example of RF energy focused within
a tumor target. In these simulations, the breast model is suspended in the water tank
such that the skin layer indicated in Fig. Figure 3.5 is parallel to the flange on the top
of the water tank, and then the breast model is centered within the opening on top of
the tank. The E-field is focused on the tumor in the breast model with the solution to
the eigenvalue problem in Eq. 2.14, and the resulting E-field, SAR, and temperature
response are computed in 3D using the material properties in Table Table 5.1. The
ABC boundaries for the breast model are the same as those defined earlier for E—field
simulations in the water tank. These dimensions encompass the simulation model that
combines the applicator in Fig. Figure 3.4 and the breast tumor model in Fig. Figure

3 .5. After the E-field is computed, the SAR is calculated with Eq. 2.5. To reduce the

30

           
   
  
   

 
 

   

  

  
     
        
      
        

  

 
  
    
    
    

  

-452Wta“
0 8 11.1"" c 'o’s’e‘ \‘\\\\\‘\x
. \9'0'0 Dbos‘\‘\
"‘ 0.4 WWW
3 \\ ”0: l:,'h.-.~.~\~
°.. "' ""o;:.t.-“
5 \ 90". III/f"
a 0.2 “ "'I,Il//
E II’
0
o
C 5
10
' -5
X (cm) —5 -10 V (cm)
(3.) Measured E—field for a steered focus.
1 -
o a 3“‘\‘\\\‘\\\\
' \\
= 0.6 ‘ ‘\\‘\\\\\\\%\\\\
I §§\\
"‘ 0.4
8
g 0.2
a
g o
I: 5 10

0
' -5
X (cm) -5 -10 V (cm)

(b) Simulated E—field for a. steered focus.

Figure 3.7. Examples of measured (a) and simulated (b) 140MHz E-fields in the xy
plane achieved through electronic steering. Although some differences appear near the
far corner of the grid, the shapes of these E—field meshes are quite similar, particularly
in the region near the peak. In (a) and (b), the E-field is measured 3cm below the

water surface (2 : -30m).

31

Y (mm)

 

 

Figure 3.8. E field distribution in the water tank. the input power applied to each
channel is 1W, and the phase value is 0° for each input channel. a) E field in xy plane
with z : -3cm, b) B field in xz plane with y — 0cm and b) E field in yz plane with x
: 0cm.

32

memory and time required for temperature calculations, SAR values are stored within
a restricted volume that contains the entire breast and some connected soft tissues.
For SAR and temperature calculations only, the x values are limited to the range
from ——81mm to 81mm, the y values are limited to the range from —81mm to 81mm,
and the 2 values are limited to the range from —110mm to 110mm. Since there is
very little penetration beyond the breast at 140MHz, the SAR distribution in these
restricted coordinates produces approximately the same temperature distribution in
the breast as the larger grid of SAR values in much less time. The temperatures in
the breast are then calculated with a steady-state finite difference implementation of
Eq. 2.6 that maintains a boundary condition of 39°C in the water tank, where this
boundary condition simulates the effect of a circulating water bath during a patient

treatment.

The results of these E-field, SAR, and temperature calculations are shown in
Figs. Figure 3.9 and Figure 3.10 for a focus located at (22,3), 2:) = (0,—15,—67)mm,
which is within the tumor model near the tumor boundary and proximal to the
applicator. The computed values for the E-field, SAR, and temperature in Fig. Figure
3.9 are demonstrated in the central portion of the breast model evaluated in the
a: = 0 plane. The E—field, SAR, and temperature values in Fig. Figure 3.10 are
evaluated in the y = —16mm plane, where the peak SAR and temperature values
occur. This plane is near the center of the spherical tumor model, which is located
at (11:, y, z) = (0, -12, —50)mm in the applicator coordinate frame. The center of the
breast is located at (:r,y) = (0,0)mm in this coordinate frame, with the surface of

the skin layer defined in Fig. Figure 8.3 coincident with the z = 0 plane.

The antenna phases that generate these results are 61 = 26.76°, 02 = 724°,
03 = 60.2°, and 64 = 00°, and the corresponding magnitudes of each excitation are
[11] = 0.65, [12] = 1.0, [13] = 0.85, and [I4] = 0.57. The E—field and SAR values are

scaled such that the peak overall output temperature is equal to 436°C. The com-

33

puted fields shown in Figs. Figure 3.9 and Figure 3.10 are obtained after the scale
factor is applied. Fig. Figure 3.9a shows that the largest E-field values in water are
located near the individual antennas, whereas in tissue, the peak E-field values in the
x = 0 plane are near the skin surface. In Fig. Figure 3.10a (which is orthogonal to the
result shown in Fig. Figure 3.9a), the peak E—field values are located near the skin
surface and in fat near the tumor interface. The corresponding SAR distributions
are depicted in Figs. Figure 3.9b and Figure 3.10b. The computed SAR distributions
indicate that multiplying the squared E-field by higher values of conductivity in the
tumor model shifts the locations of some peaks into the tumor, although significant
SAR. contributions remain near the skin in Fig. Figure 3.9b and in fat near the tu-
mor in Fig. Figure 3.10b. Overall, the computed temperatures in Figs. Figure 3.90
and Figure 3.10c follow trends similar to those established for the SAR distributions
in Figs. Figure 3.9b and Figure 3.10b. In particular, the temperature peak occurs
within the tumor model in Fig. Figure 3.9c, and the peak temperatures occur in
fat near the tumor interface in Fig. Figure 3.10c. The overall temperature peak is
436°C, which is achieved in Fig. Figure 3.10c at (113,2) = (—27mm,—62mm) and
(:r, z) = (27, —62mm). These simulation results demonstrate an example of the re-
gional heating that is produced by this RF breast applicator. Although a focus is
generated within the tumor, boundary conditions and material properties strongly

influence the locations of the SAR and temperature peaks.

3.4 Discussion

3.4.1 Measured and simulated E-fields in the water tank

Figs. Figure 3.6 and Figure 3.7 show that finite element simulations predict the ap-
proximate shape and location of the focus generated by the RF phased array ap-
plicator. Figs. Figure 3.6a and Figure 3.6b demonstrate that, for a focus in the

center of the water tank, the measured and simulated E-field distributions are closely

34

 

[Vlm]

771500

A 1000

E

5

N 500
o

 

-80 —40 80

0
Y (mm)

(a) Simulated E-field in the :r = 0 plane.

[Wlkg]
rj 60

2 (mm)

 

—80 -40 BO

0
Y (mm)

(b) Simulated SAR in the :r. = 0 plane.

[°C]
r144
42
E
E
N 40
38

 

Y(mm)
(c) Computed temperature distribution in the
z = 0 plane. The peak temperature in this
plane is 425°C.

Figure 3.9. Simulated E—field, SAR, and temperature distributions generated by the
RF phased array applicator and evaluated in the :r = 0 plane of the breast tumor
model, where the white contours indicate the external outlines of the breast and
tumor. These simulation results show that, in the :r = 0 plane, the E-field peaks
in water are near the source antennas, the E-field peaks in tissue are near the skin
surface, the peak SAR values are within the tumor model and near the skin surface,
and the peak temperature in the :r = 0 plane is within the tumor boundary.

35

 

40
X(mm)
(a) Simulated E—field in the y = -16mm
plane.

 

—80 —40 0
X (mm)

(b) Simulated SAR in the y = —16mm plane.

 

80

O
x (mm)

(c) Computed temperature distribution in the
y = -16mm plane. The peak temperature
in this figure, which is also the peak overall
temperature, is 436°C.

Figure 3.10. Simulated E—field, SAR, and temperature distributions generated by the
RF phased array applicator and evaluated in the y = —16mm plane of the breast
tumor model. In the y = —16mm plane, E—field peaks in tissue appear in fat near
the tumor interface, peak SAR values are in fat near the tumor interface and in the
tumor proximal to the applicator, and the peak temperature in the y = 0 plane is

located in fat near the tumor interface.

36

aligned along the x- and y-axes, respectively, for magnitudes evaluated at z = —30m.
Fig. Figure 3.6c also indicates that the finite element results successfully predict the
locations of the minimum and maximum values along the z-axis. Likewise, simula-
tions and measurements indicate that the E—field distributions are also similar for a
steered focus as shown in Fig. Figure 3.7. These figures demonstrate that the focus
generated by this RF phased array is quite broad, which suggests that this device is

appropriate for regional heating.

Some differences between the measured and simulated E-field distributions are also
evident in these figures. For example, simulated and measured values along the z-axis
in Fig. Figure 3.6c diverge outside of the water tank. In particular, the measured E-
field drops off rapidly in air beyond the water interface, whereas the simulated E-field
decays slowly. Although the simulated E-field is similar to the measured E—field near
the focus in Fig. Figure 3.7, the differences between simulated and measured E-fields

increase with distance from the focal peak.

These differences are attributed to simulation and measurement errors. Finer
meshes will reduce the numerical error where the E-field is changing rapidly, and
memory limitations also restrict the total number of tetrahedrons in the finite element
mesh. For these simulations of the E-field produced in the water tank, the average
distance between adjacent nodes is 0.0306/\ in air, 0.0134A in lexan, and 0.05A in
water. Nodes in the finite element mesh are more closely spaced near interfaces,
where the E-field changes rapidly, and where the magnitude of the E-field is relatively
large, whereas the distance between adjacent nodes increases where relatively small
E—field magnitudes are changing slowly. The error is introduced by the absorbing
boundary is relatively small, as indicated by simulations that compare results for
absorbing boundaries located /\/10 and A from the water tank. Errors in the tank
dimensions, antenna dimensions, and antenna locations likewise contribute to errors

in the numerical model.

37

Errors are also introduced by misalignment between the simulation model and the
measurement grid. Another source of error is the inherent misalignment between the
three E-field probes. These E-field probes were applied to each measurement, and
then the results were averaged to improve the signal to noise ratio. The relative probe
locations were carefully measured before and after these measurements; however,
probe locations changed after the probes were repeatedly inserted into the water
tank and extracted from the water tank. In addition, some coupling between the
computer-controlled 3-axis positioning system and the RF applicator was observed
during these measurements. As a result, the probes were mounted on a polycarbonate
rod, which was then attached to the 3-axis positioner. Additional shielding was added
as necessary. Although this reduced the noise somewhat, the coupling between the
applicator and the positioner nevertheless persisted, and this coupling influenced the

measurement results.

Coupling between antennas is also evident in these measurements. This coupling
is in part caused by standing waves within the water tank. The standing wave pat-
tern changes as the input phases and amplitudes are varied, which also changes the
coupling between antennas. This coupling also changes depending on the load in the
Water tank. In particular, the coupling changes for different water levels and for dif-
ferent phantom materials placed in a plastic cup on the water surface. This coupling
iS responsible in part for the differences between the measured and simulated phase
and amplitude settings required for a focus in the center of the tank and for a steered
f0Cus. These effects, which are commonly encountered in RF phased array systems
dESigned for hyperthermia [?, 34, 7], are reduced by focusing the breast applicator in
center of the water tank so that the reflected powers are minimized. This defines the

location where preferential focusing is achieved for this RF phased array applicator.

38

3.4.2 Focusing strategy

The simulated focus was generated in the water tank and in the breast model with
a modified version of an approach that was defined previously for temperature opti-
mization [21]. This approach, which is similar to a method that focuses ultrasound
therapy arrays [7], maximizes the gain from the aperture to the focus. Eqs. 2.14 and
2.15 provide an alternative to more complicated methods that maximize the total SAR
in the entire tumor relative to the total SAR generated in normal tissues [35, 10, 19].
With this simplified focusing approach, the results presented in Figs. Figure 3.9 and
Figure 3.10 are obtained for a focus that is generated along the edge of the tumor
in a location that is proximal to the applicator. This location minimizes the path
to the focus through the tumor, thus reducing the effects of penetration depth on
the magnitude of the simulated focus. Similar results (not shown) are obtained for
this applicator geometry and breast model with the optimization method outlined in
[19]. With the focusing approach of Eqs. 2.14 and 2.15, additional optimization is

facilitated through control of the focus and the total input power.

3.4.3 Simulated E-fields evaluated in the breast model

The simulation results depicted in Figs. Figure 3.9a and Figure 3.10a show the E-field
within a restricted field of view for the 3D breast model. Although the E—field was
computed for a much larger range of coordinate values, the E-field results are demon-
strated in the restricted field of view for two reasons. First, these results follow the
same range of coordinate values shown in Figs. Figure 3.9b and Figure 3.10b for the
SAR and in Figs. Figure 3.9c and Figure 3.10c for the temperature. Plotting these
results on the same axes promotes comparisons of the similarities and differences be-
tween the computed E-field, SAR, and temperature distributions. Second, restricting
the field of view centers attention on the computed E-field in soft tissues. Including

the fields near the antenna obscures the fine structure of the E—field distribution in

39

soft tissues as demonstrated in Fig. Figure 3.9a, where the magnitude of the E-field
immediately adjacent to the RF antennas is significantly greater than in soft tissue.
In planes that avoid the near field of the individual source antennas, more detail is
clearly evident, as demonstrated in Fig. Figure 3.10a. This figure, which reduces
the dynamic range by roughly a factor of two relative to Fig. Figure 3.9a, shows the
full range of computed values in the y 2 —160m plane. By restricting the range of

coordinate values in this plot, the computed values in tumor and fat are emphasized.

For the computed E-field evaluated across the entire range of coordinate values
(not shown), the largest magnitudes are generated near the antennas, and intermedi-
ate values are indicated in water and within the soft tissue target. These simulations
also show that the E-field decays rapidly in air. This result suggests that the RF
phased array applicator operates as a resonant cavity that produces standing waves
within the coupling medium. These standing waves, which are manipulated by the
phase and amplitude inputs of the RF phased array applicator, produce peaks and
valleys as indicated in Fig. Figure 3.6. The peaks and valleys produced by the stand-
ing wave patterns are manipulated through scanning and focusing, and these standing
wave patterns also influence the reflected powers that are measured at the antenna
inputs. In order to avoid unfocused near field components, the breast is positioned
some distance away from the individual antennas, preferably at or near the geomet-
ric focus of the resonant cavity. Although the E-field values are relatively small in
air, these values also influence the computed field through the boundary conditions.
Simulation results (not shown) indicate that the material behind the breast also in-
fluences the distribution of the E-field in the breast and in the water tank at 140MHz.
This suggests a need for more detailed patient models in these simulations, which are
presently under development. These simulation results also depend on the distance
between adjacent nodes in the finite element mesh. In simulations of the E-field gen-

erated in the breast model, the average distance between adjacent. nodes is 0.0466A in

40

air, 0.0213/\ in lexan, 0.0852/\ in water, 0.0701). in fat, and 0.0311A in the simulated
tumor. As in previous simulations, the nodes are more closely spaced near interfaces,
where the E-field changes rapidly, and where the magnitude of the E-field is relatively
large, and the distance between adjacent nodes increases for relatively small E-field

magnitudes that are changing slowly.

Figs. Figure 3.9a and Figure 3.10a show that peak magnitude of the E-field is
located outside of the tumor boundary, and similar effects are also observed for peak
SAR and temperature values. In Fig. Figure 3.9a, the peak magnitude of the E—field
in water is located near one of the antennas, and the peak magnitude of the E-field in
soft tissue is located just beneath the skin surface. In Fig. Figure 3.10a, peak E-field
magnitudes are adjacent to the tumor boundary in fat. Although the constructive in-
terference is maximized at the focus within the tumor, the magnitude of the E—field is
significantly greater at the peak outside of the tumor in Fig. Figure 3.10a. This effect
is a consequence of the boundary conditions at the tumor/ fat interface and the orien-
tation of the E-field generated by the RF phased array applicator. This combination
maximizes the E-field outside of the tumor model, and a severe impedance mismatch

at the boundary reduces the peak magnitude of the E-field within the tumor target.

A similar trend is observed for the SAR distribution in each plane. In Fig. Figure
3.9b, the peak SAR value within the tumor is approximately the same as the peak SAR
value located just beneath the skin surface. In Fig. Figure 3.10b, the peak SAR within
the tumor also approaches the peak SAR achieved along the tumor boundary. Thus,
multiplying the magnitude squared E-field values increases the relative contribution
to the SAR within the spherical tumor model.

Peak temperatures follow the trend of the SAR away from the skin surface, which
is maintained at 39°C by the circulating water bath. A water bath temperature above
body temperature is afforded by the 140MHz excitation frequency and large aperture

of the RF phased array applicator. This combination achieves heating within the

41

breast without overheating the skin; therefore, skin cooling is not required for this
system. In Fig. Figure 3.9c, a peak temperature of 425°C is indicated within the
tumor model. However, a larger peak temperatures of 436°C is encountered outside
of the tumor model in Fig. Figure 3.10c. Although the E-field is focused within the
spherical tumor model and SAR results show that the peak SAR value in the tu—
mor model is comparable to the peak SAR outside of the tumor, the overall peak
temperatures in the breast tumor model are nevertheless generated in fat. These
results demonstrate that the boundary conditions, particularly the orientation of the
E-field vector within the breast and tumor regions, strongly influence the resulting
temperature distribution. Other results (not shown) demonstrate that the peak tem-
perature increases in fat and decreases in the tumor as the focus moves away from
the applicator and deeper into the tumor. This suggests that the penetration depth,
combined with an impedance mismatch at the fat/tumor boundary, strongly influ-
ences the ability of RF devices to heat deep tumors in the breast. The results shown
in Figs. Figure 3.9 and Figure 3.10 also suggest that both SAR and temperature
calculations are required for these evaluations. The heat generated along the tumor
boundary is expected to contribute to regional heating in the breast, increasing the
temperature of arterial blood and thereby improving the temperature distribution in
the tumor. This result also demonstrates the need for noninvasive magnetic resonance
(MR) thermometry [36] that provides valuable feedback for this regional heating de-
vice. Noninvasive thermometry will verify the temperature output generated by this
RF phased array applicator, which changes based on the load that is presented to the
applicator (i.e., for different patients). Temperature feedback will facilitate compen-
sation for load changes and cross-coupling through phase and amplitude adjustments
applied to the controlling amplifier outputs. This feedback will then enable targeted

heating of breast tumors with this RF phased array device.

These results suggest that conformal heating of LABC tumors will be enhanced

42

with an approach that incorporates an ultrasound (US) component within a hybrid
RF/ US phased array device. Measurement and simulation results show that the RF
phased array is a regional heating device that beats a large portion of the intact
breast, and adding local heat produced by an ultrasound phased array is expected to
improve the temperature distribution in the distal portion of the tumor. The hybrid
approach will exploit the regional heat generated by the RF applicator that increases
the temperature of the arterial blood supply, and then the ultrasound component
will locally raise the temperature as needed within the target volume. The existing
RF phased array structure, which occupies four of the six flat panels in the lexan
structure, is ideal for incorporating additional ultrasound applicators. Furthermore,
the large heated region created by the RF phased array combined with the small
focal spots generated by the ultrasound phased array provides multiple Opportunities
for Optimization of the temperature distribution produced by hybrid RF/ US phased
array devices. This RF phased array design, which fills four lexan panels with end-
loaded dipole antennas, is ideal for hybrid structures that incorporate additional
ultrasound applicators into either of the remaining side panels. Other new geometries
that populate fill all of the panels with end-loaded dipole antennas are also under

development.

3.5 Conclusion

E-fields generated by a 4-channel RF phased array applicator designed for hyperther-
mia cancer therapy in the intact breast were measured in a tank filled with deionized
water. This geometry was simulated with the finite element method, and the results
indicate excellent agreement between the measured and simulated E-field magnitudes.
The applicator generates a geometric focus in the center of the water tank, and mea-
surements and simulations show that, through the selection of appropriate phases and

amplitudes applied to each of the antennas, this focus can also be steered off-axis.

43

Some differences between the measured and simulated E-field magnitudes are evident
in the geometric focus and the steered focus, and these are attributed to simula—
tion errors and measurement errors. Overall, comparisons show that finite element
simulations successfully predict the focal pattern generated by the RF phased array
applicator within the water tank.

Finite element simulations were also evaluated in a simplified breast model that
contains a spherical tumor volume. These simulations show that, although the RF
phased array is focused within the spherical tumor model, the peak temperatures
occur outside of the tumor. These temperature peaks, which are a consequence of the
material properties of fat and tumor and the orientation of the incident E-field, are
immediately adjacent to the tumor, so these peaks enhance regional heating generated
by the RF phased array applicator. The simulated E—field, SAR, and temperature dis-
tributions suggest that this RF phased array applicator effectively generates effective
regional heat for hyperthermia treatments of locally advanced breast cancer. Fu-
ture evaluations of this RF phased array applicator will include realistic 3D anatomy

derived from CT and / or MR images.

44

CHAPTER 4

THE FIVE RF DUAL-DIPOLE ANTENNA APPLICATOR

4.1 The five antenna MRI compatible applicator

In recent years, increased interest in thermal therapies, such as hyperthermia and
thermal ablation, has occurred due to the ability of medical imaging techniques to
guide the treatments. Imaging can be used to target diseased tissue, to visualize the
heated area and to image the result after therapy. The use of imaging methods to
measure temperature changes has been tested with ultrasound, CT and MRI. But the
only currently available in vivo temperature imaging method is measuring the water
proton resonant frequency (PRF) shift with MRI. So one key point of designing
the thermal therapy applicator is MRI compatible and there is no electronic and

electromagnetic interference between MRI and the therapy system.

A new MRI compatible applicator is designed and integrated with the MRI system
(1.5T GE Signa Infinity with Excite magnet) to monitor the temperature distribution
in the patient breast in real time. The whole system shown in Fig. Figure 4.1. To
make the whole system compatible, there must be no electrical IOOps involved in the
chamber of the MRI system and distort the RF signal and the magnetic field. Instead
of the common used LC matching circuits, the single open-end transmission stub with
length 22cm (one wavelength at 140 MHz) is used for impedance matching of each

antenna and the connect point is at the feed point of the antenna.

The previous design of applicator is investigated in [37], four U-shaped antenna
parallel mounted the inner side of the plastic tank. This applicator has limitation
in two ways: one is that it lacks the ability to steer the focus in y direction (the
coordinate is shown in Fig. Figure 4.2b) because all four antenna is in zz plane, the

other limitation is E field linear polarization problem[38] which will cause hot spots

 

Figure 4.1. The thermal therapy applicator and MRI integrate system. The MRI
compatible applicator is mounted in the bench. The patient lies on the bench with
face down. This integrate system can heat patient and monitor the temperature
distribution simultaneously.

46

in the E field direction (which will be explained in detail in later chapter). The new
design of the five antenna applicator is shown in Fig. Figure 4.2. Four antennas
are mounted around the breast and the fifth one is mounted on the bottom panel.
With the arrangement of the antennas, the applicator can steer the focus in three
dimensions.

The size of the top panel is about 33.6cm x 22.5cm. The bottom plane is 18.8cm x
22.5cm. Both the top panel and the bottom panel are parallel to the my plane and the
depth of the of the is about 12.0cm. In the two hexagonal side panels, which are both
parallel to the yz plane and symmetric to the xz plane, the top edge is 33.6 cm long,
the bottom edge is 18.8 cm long, higher edges are 3.3 cm, lower edges are 15.3 cm, the
two lowest obtuse angles are 150°, the two top angles are right angles, and the other
two obtuse angles are 120°. Except the top plane, other planes are covered by plastic
(Lexan) with 0.5cm thickness. Five U-shaped antenna are mounted on five sides of
the applicator shown in Fig. Figure 4.2. The arms of three antenna are parallel to
the y axis. The arms of the other two antenna on the hexagonal panel are parallel to
the 2: axis. During the treatment, the applicator is filled with deionized water with
a constant temperature 39°C. During the treatment, the patient lies down prone to
the applicator and the body axis is parallel to the y axis and there is 1-2 cm air gap

between the chest wall and the water surface.

4.2 Phantom and Patient

In this chapter, several phantoms are made for testing and evaluating the applicator.
The polyacrylamide gel (http://www.bergen.org/AAST/Projects/Gel/index.html) is
made to mimic the breast fat tissue. The polyacrylamide gel is composed of 84%
eucerin and 14% mineral oil. To reach the target material property[1], the dialectic
property of the gel phantom is calibrated by adjusting the proportion of these two

materials with the network analyzer[1]. The conductivity and permittivity of the

47

X (m)

 

(b) Simulation model of the 5 antenna applicator

Figure 4.2. The five antenna applicator

48

tumor of breast cancer are similar to the muscle. The tumor phantom is a piece of

beef steak is in the experiment.

4.2.1 Homogeneous Gel Phantom

The hole on the applicator cover, which is left for breast to fit in, is covered with soft
plastic film; The polyacrylamide gel is placed on the film and in the hole; The film
with the gel gets into the hole and water to form a breast phantom. The homogeneous
polyacrylamide gel phantom is made to demonstrate the steering capabilities of the
breast applicator (developed at Hyperthermia, Radiation Oncology, Duke University
Medical Center). The side view and top view of this phantom and the film is shown
Fig 7? and a photo of this phantom is shown in Fig. Figure 4.4b (not including the
beef steak).

One slice of the homogeneous gel phantom and the applicator filled with water is
shown in Fig Figure 4.3, which is under and close the water surface and the phantom

is center circle in this MRI slice.

4.2.2 Fat / Tumor Heterogeneous Breast Phantom

The heterogeneous breast phantom is a two material model: fat-equivalent region
and tumor-equivalent region. The tumor is surrounded by fat, which is common in
Locally Advanced Breast Cancer (LABC). The location of the tumor in this phantom,
indicated in Fig Figure 4.4a, is not at the center and about 2-4cm offset from the
center. The tumor-equivalent phantom is a piece of steak with 4 cm length, shown
in Fig. Figure 4.4b. Three temperature probes were inserted in the phantom (2 in

tumor, 1 in fat) to monitor temperature change during heat up by applicator.

4.2.3 The patient and patient model

The applicator is mounted on the clinical bed, the top of the applicator is kept the
same level at the surface of the bed shown in Fig.Figure 4.1 and the tOp of the

applicator is covered with plastic cover with circle hole left open for patient breast.

49

wcoronal vieWs

~Cf polyacrylamide
Gel Phantom

 

Figure 4.3. Focus Steering in Homogeneous Gel Phantom

50

Side View

 

 

Top \fiew

(a) Drawing of the breast phantom

4

Cat No 09-0”
1]

 

(c) Fat/tumor phantom

Figure 4.4. Tumor and fat phantom Fat/Tumor Heterogeneous Breast Phantom for
breast. The size of the steak is about 5 cm

51

 

Figure 4.5. One slice of MRI imageszpatient with five antenna applicator.

The patient lies on the bed with face down and the breast with tumor is put into the
applicator through the opening. The images of the patient and applicator are scanned
by MRI scanner for treatment planning purpose before the treatment. During the

treatment, the temperature distribution

4.3 Simulation model and method

The applicator, phantom and patient are modeled and E field is calculated with
HFSS and exported in the region of interest (ROI). The total E field and the SAR,
are optimized to achieve good heating performance in desired region or tumor and

temperature increased in ROI are calculated from the total SAR with BHTE equation

4.4 Simulation and Experiment Results

The amplifier and control system and the E field measurement system are the same
as that for four antenna applicator in previous chapter. The measurement of E field

in applicator was conducted without phantom and patient in water with B field probe

52

and the positioning system outside the magnetic chamber. The temperature of the
phantom is monitored with the applicator and phantom inside the MRI system. The
temperature increase in three planes, which are picked by experience, are evaluated

in every 2 mins by measuring the water PRFS.

4.4.1 E field and Focus steering in water

The E field is focused in the water tank with phases and amplitudes by the same
method used in the previous chapter. A focus was achieved at the center of the water
tank with input phase settings of 0°, -170°, -30°, -50°and 0°applied to channels 1, 2,
3, 4and 5, respectively. Forward power settings of 40W is applied to each channel, the
measured reflected powers were 14.6W, 7.4W, 14.4W, 34.3W and 3.5W for channels
1, 2, 3, 4and 5, respectively. The measured E field distribution on the xy plane about
3 cm under the water surface is shown in Fig. Figure 4.6 and Fig. Figure 4.6 gives
four different views of the mesh plot.

A back steered focus was generated in the water tank with input power value 50W
for channel 1, 40W for channel 2 and 3 and CW for channel 4 and 5 (in other words
channel 4 and 5 are closed). Phases for five channels are 0°, -60°, 110°, 0°, 0° and the
reflected power are 29W 25W 26W for channel 1, 2 and 3. The measured E field on

the xy plane about 3 cm below the water is shown in Fig. Figure 4.7.

4.4.2 Focus Steering in Homogeneous Gel Phantom

The MRI-compatible breast applicator has also undergone extensive experimentation
to demonstrate the control it has over the focus of heat. Here a homogeneous poly-
acrylamide gel phantom is used to demonstrate the steering capabilities of the breast
applicator.The MRI coronal view of the polyacrylamide gel phantom with applicator
is shown in Fig. Figure 4.3. Fig Figure 4.8 show temperature increase with different

focus location in experiment and simulation.

Center focus heating is shown in Fig Figure 4.8 a and b. E field at the phantom

53

 

  

 

 

 

 

 

 

6 \ if/ 12 14
aio‘Td'z 4 5 a 10
X(cm) 0 “cm,
ToleeuiZ)

14 . f

12 ,

10
E” a l——
; 6' W:

‘ a ‘

2

o

 

 

4 6
X(cm) X(cm)

Figure 4.6. E field distribution in water and 3 cm below the water surface. The E
field is focused at the center of the applicator.

54

Power Distribution in S-Art. RF Cavity. 6cm from Top: BACK Lateral View (Y)

0.8]

 

U
i 2
20.5) no.5]
N '0
.§
E04»
2 o
202]

    

e ""6 We”; 12 14

 

5‘3“] "4 o '
X(cm) 1002 02466101214
Y(cm)
Y(em)
Top Wew (Z) Lateral \fiew (X)

    

Normalized Power

 

Figure 4.7. E field distribution in water and 3 cm below the water surface. The E
field focus is steered to (y —. 10cm) in the applicator.

55

center and 3cm below the top phantom surface is maximized with focusng strategy
(in Chapter: Four antenna applicator) in the simulation. Because of phase errors
present in RF phased array systems, the focusing and focus steering mainly depends
on the B field focus experience in water. With the same combination of input phases
and amplitudes as center focus in water, the temperature increase at center region
of the phantom is observed which is indicated in Fig Figure 4.8a with a circle, while
the around phantom region is lower temperature. The high temperature region in
experiment and simulation have the same shape and the shape is not circle because
two antenna parallel to yz plane are closer to the phantom than other antennas.
Based on the experience of E field focus steering in water, the heating region can
be steered off the center, shown in Fig Figure 4.8c and e. With maximizing the E
field at different location, the heating region can move to other location, shown in
Fig Figure 4.8d with focus at [0,-5,-3]cm and Fig Figure 4.8f with focus at [5,0,-3]cm.
The peak temperature region in experiment and simulation have similar contours
for the bottom steering case. They also have similar contours for the right steering
case. While the gradient of temperature and the size of the peak temperature are
different for the bottom steering case (Fig Figure 4.8c and d) and the right steering
case (Fig Figure 4.8e and f). The right steering case has smaller peak temperature
region and higher temperature gradient than that in the bottom case, because two

antenna parallel to yz plane are closer to the phantom than other antennas.
4.4.3 Heat Focus Steering in Fat / Tumor Heterogeneous Phantom

One slice MR image of the heterogeneous phantom with applicator is shown in Fig.
Figure 4.9. The white region close the center of the phantom is the tumor phantom
(beef steak) and dark region around the tumor is the gel phantom. Outside the gel
phantom is water.

The temperature change in the phantom was recorded in three parallel planes

with different depth. One of them is shown in Fig. Figure 4.10a at time turning

56

i ’

'L‘itxw: 10111113. 40 ::

 

Y (mm)

 

Figure 4.8. Focus steering in experiment and simulation

57

 

#0)

Ti?

 

Figure 4.9. MR. image of the heterogeneous phantom

off the applicator and temperature reaching the peak. In Fig.Figure 4.10a, the peak
temperature distributes at the tumor/ fat boundary and not inside the tumor, while
some part of the tumor/ fat boundary is not heated as much as that in temperature
peak region. The same phenomena happens in simulation results. The same shape
tumor in the breast model is simulated and E field is maximized in and around the
tumor region. Because of the impedance mismatch between fat and tumor, E field is
reflected back at tumor/fat boundary. Because of four of five antenna in xy plane,
the E field is linearly polarized in xy plane and those fat/tumor boundary vertical to

the polarization direction are prone to be well heated.

The temperature change in time domain is also studied in experiment and simula-
t ion. FigFigure 4.11a shows the measured temperature change at peak temperature

region [3,2,-3]cm. and Fig Figure 4.11b shows the simulated temperature change at

58

X (mm)

 

-60 -30 0 30 60
Y(mm)

(b)

Figure 4.10. Heat Focus Steering in Fat ,1" Tumor Heterogeneous Phantom

the same location. The applicator is turned on for 23 mins and then turned off after
that. The curves match each other when the phantom cooling down, while simulated
temperature increases faster than the measured temperature. It could be that air

around the phantom takes away the heat, which is not counted in the simulation.
4.4.4 Heat the real 3D breast model

The 3D breast model, extracted from the patient MR images, is treated with the five
antenna applicator in simulation. The electric field distribution in the breast model,
water tank and air is calculated with HFSS and the E field for each antenna are
calculated separately. The SAR and temperature are evaluated in 14cmx14cmx 10cm
region. The temperature is calculated from the electric field SAR with the static
BHTE equation. The water is keep at 37°C in all temperature calculation and the
maximum temperature rise in the whole region is set as 6°C. The material property
and blood perfusion are shown in previous chapters. To deliver maximum power from
the power amplifier, the power magnitude of each channel is usually the maximum
available power. In the Optimization, the magnitude for each channel is set unit 1
and only phases of five channel are optimized.

Without phase optimization, the phases for five channel are all 0. The temper-
ature distribution with 44°C up limit is shown in Fig. Figure 4.12. The SAR. and
temperature distributions in three cut planes are shown in Fig. Figure 4.13.

With the Wiersma method, The average SAR ratio (the average SAR in tumor
to the average SAR. in health tissue) is maximized and the optimized phases for five
channels are 212°, 108°, 18°, -90°and 65°. The temperature distribution with 44°C
up limit is shown in Fig. Figure 4.14. The SAR and temperature distributions in
three cut planes are shown in Fig. Figure 4.15.

With the point phase method, the E field value at specified location [-2, 2, -5]cm
is maximized and the optimized phases for five channels are -153°, 146°, 176°, 17°and

0°. The temperature distribution with 44°C up limit is shown in Fig. Figure 4.16.

60

 

 

 

 

10

8 . ,
H 6 '-
0
‘L.
.- 4 . >

2 _

o 1 1 1 1 1

0 5 10 15 20 25 30
t (min)
(a) Experiment result
6 a 4

 

 

 

 

 

o 5 10 1s 20 2‘5 30
t(mins)

(b) Simulation result.

Figure 4.11. Temperature varying with time at sample point

61

 

0

Figure 4.12. Temperature 3D distribution without optimization

The SAR and temperature distributions in three cut planes are shown in Fig. Figure
4.17. some other location, such as [0, 0, -5]cm and [-4, 4, -4]cm, are also optimized

and the 3D temperature contours look similar to the contour in Fig. Figure 4.16.

4.5 Discussion

There exist a lot of sources of uncertainty that can occur in MRI-based temperature
measurements, including noise, uncertainty in the baseline temperature, the effects of
averaging the temperature over several MR voxels, uncertainty in the calibration of
the PRF shift in viva, drift of the static magnetic field, artifacts induced by motion or
tissue swelling and the other reported errors in the temperature mapping. However,
even with the uncertainties described above, a growing number of animal and clini-
cal studies have been performed that demonstrated a good agreement between MRI
temperature mapping and the heating results and showed MRI temperature mapping

can be online control of thermal therapy and/ or ablation.

62

X (mm)

 

SAR in XY plane with z = 4cm T in XY plane with z = 4cm

      

 

[100
80
so a
. e
40 x
20
-50 0 50 0 -60 -30 0 30
Y (mm) Y (mm)
SAR in XZ plane with y = 0cm T in X2 plane with y I 0cm
0 100 0
‘ 50
'80 -50 0 50 0 -40 o
X (mm) x (mm)
T in XY plane with z = 40m
SAR in Y2 plane with x a 0cm
0 [5.100

X (mm)

       

Figure 4.13. No optimization

63

           

o.
A ' 5.x;
Bl ‘ .-
é 4o ‘ 'g:%: ‘-,0
v e '
~ -60~ 15a:

 

Figure 4.14. Temperature 3D distribution with the Wiersma method. The aver-
age SAR ratio (the average SAR in tumor to the average SAR in health tissue) is
maximized and the optimized phases for five channels are 212°, 108°, 18°, -90°and
65°.

In this hyperthermia MRI integration system, the RF applicator also disturb the
81 field of the RF coil and it will also introduce artifact to the imaging although
the RF frequency 140MHz is different from the Larmor frequency. The current RF
coil is the single loop RF coil (transmitter and receiver) for breast imaging. The
electromagnetic interference between the RF coil and the RF antenna will be an issue
in this integrated system. The high power hyperthermia RF power could disturb the
receiver of the RF coil and the metal surface of the RF coil could reflect and disturb
the electric field distribution radiated by RF antenna array, which is not addressed by
simulation. When MRI scanner are acquiring temperature mapping, the RF power

of this applicator need to be blank to keep the receiver safety.

64

SAR in XY plane with z = 4cm T in XY plane with z = 4cm

     

:6
E :4
E
><
2
-50 0 50 -60 -3O 0 30 60 0
Y(mm) Y(mm)

SAR in X2 plane with y = 0cm T in X2 plane with y a Dem

     

0 -40 0 40 0
X (mm) X (mm)
T in XY plane with z = 4cm
T in Y2 plane with x a Dem

X (mm)

     

060

 

-60 .30 0 3
Y(mm)

Figure 4.15. With the Wiersma method. The average SAR ratio (the average SAR in
tumor to the average SAR in health tissue) is maximized and the optimized phases

for five channels are 212°, 108°, 18°, -90°and 65°.

65

 

Figure 4.16. Temperature 3D distribution with the point phase method, the E field
value at specified location [-2, 2, —5]cm is maximized and the optimized phases for
five channels are -153°, 146°, 176°, 17°and 0°.

4.5.1 Focus steering in water and homogeneous gel phantom

The four antenna in previous chapter can only steer the E field and SAR focus in $2
plane (the cartesian coordinate system is shown in Fig. Figure 4.2b), while this five
antenna applicator can steer the E field and SAR focus not only in the 2:2 plane but
also in the my plane. Most of the tumor is not located at the center of the breast, so

the steering ability in my plane is necessary for breast applicator.

In Figs. Figure 4.6 and Figure 4.7, the size of the peak is about 6cm diameter
in water. The wavelength in water is about 24cm at 140MHz and the size of the E
field peak is about a quarter of the wavelength. In both figures, E field peak has
more gradient along y direction, in other words E field is steeper along y direction.
Comparing the peak center in Figs. Figure 4.6 and Figure 4.7, the E field peak moves

about 5 cm along y direction.

By changing the five input magnitudes and phases, the E field and SAR can also

66

T in XY plane with z = 4cm

X (mm)
X (mm)

     

   

60 0 50 -60 -30 0 30 60
Y (mm) Y (mm)
SAR in X2 plane with y 8 Gem T in X2 plane with y 8 Gem
20 100 20 r—[
l ] .6
5O '4 4
2
0 ° .40 0 °
X (mm) X (mm)
SAR in Y2 plane with x = 0cm T in Y2 plane with x I 0cm
0 2

   

Figure 4.17. With the point phase method, the E field value at specified location
[-2, 2, —5]cm is maximized and the optimized phases for five channels are -153°, 146°,
176°, 17°and 0°.

67

be steered in the homogeneous gel phantom. Figs Figure 4.8 shows three steering
case of MRI temperature mapping and simulated temperature distribution in about
a plane about 3 cm below the water surface. These three case show great agreement
between MRI measurement and simulation results. The center focus in Figs Figure
4.8a and b shows temperature peak is not a circle and the peak has an extension
along y direction, because the two antenna on y axis are closer than that two in x
axis. Comparing the bottom focus and right focus, the right focus has higher gradient

in temperature, because the E field focus has the same trend in water
4.5.2 Heat focus steering in fat / tumor heterogeneous phantom

The fat / tumor heterogeneous phantom is the phantom model of breast with tumor.
Fig. Figure 4.9 is one slice of the heterogeneous phantom and the tumor is in white
color and around the white tumor is the fat in black-grey color. There are two gaps
(in deep black) in the fat phantom below the tumor. This kind of gap does not exist
in real breast. Because the single loop RF coil is used here, so only the region inside
and close to the loop can be viewed, some region in water is imaged in bright color.

The thick line contours in Fig. Figure 4% shows the phantom simulation model
in this plane about 3 cm under the water surface. The simulation model is drawn
in HFSS according to the MRI images of the heterogeneous phantom and E field is
maximized around the tumor region. Whatever the E field maximization location
or power magnitude and phase for each channel are chosen, the peak temperature
is outside the tumor region and close to the tumor/ fat boundary in best case. This
is also proven in the measurement with MRI temperature mapping. In Fig. Figure
4.9a, the left and right boundary of the tumor are best heated and heat transfers
into the tumor region from the boundary until reaching the balance. The location of
the temperature peaks on the boundary is determined by the E field polarization in
tumor region, which will be talked in later chapters.

Temperature increasing with time are also monitored with MRI every two mins

68

and simulated with BHTE with time term. The curves for peak temperature location
are shown in Fig.Figure 4.11. With certain SAR level, both the simulation and
experiment reach the peak temperature after 22"24mins after the power is on. After
the power is turn off, the experiment and simulated curves go down with almost the
same trend and gradient. But the experiment and simulated curves show a little bit
different in the gradient of the temperature increase between time 0 mins and time

20 mins.
4.5.3 Heat the real 3D breast model

With the real 3D breast. model extracted from MRI images (one of them is shown
in Fig. Figure 4.5), the temperature simulation with some level optimization will
be the treatment planning for breast cancer. Three different power inputs give huge
difference on the final temperature distributions. Without any Optimization method,
each antenna has the same power level and zero phase and the temperature over 42°C
region is totally outside of the tumor. The Wiersma method maximizes the average
tumor/ normal SAR ratio and neglects the geometry shape of the tumor and breast.
This method works well in a lot of cases, for example the simple breast model in
previous chapter. However there could be a SAR peak with small size somewhere
in the health tissue, which can be caused by the irregular tumor shape. The point
maximization method works very well for this case and the temperature over 42°C
contour cover most of the tumor in Fig. Figure 4.16. The results are not sensitive to
the location of the maximized point, for instance, the 3D temperature contours are
almost the same with naked eye for points [0, 0, -5]cm ]-2, 2, -5]cm and [-4, 4, -4]cm.
The common character among these points are that they are all inside tumor and
close to the tumor/ fat boundary.

Different optimized magnitude and phase settings give different heating contour.
Some tumor region heated in Fig. Figure 4.12, however, is not well heated in Fig.

Figure 4.16. Even different point location in point maximization method can give

69

different heating results. So combining those heating pattern in time sequence can
give even better result than anyone individual. This topic will be talked in later
chapters.

After the magnitudes and phases are chosen from treatment planning or opti-
mization, the total E field inside breast and water is linear polarized in most of the
case. If four of the five antennas except the bottom one is driven with phase 0°,-
90°,180°and 90° individually, the E field inside the breast will be circular polarized
and the SAR. distribution around the tumor will be smoother than that with linear

polarized settings. This topic will be talked in later chapters.

4.6 Conclusion

The only currently available in viva temperature imaging method is measuring the wa-
ter proton resonant frequency (PRF) shift with MRI. A new five dual-dipole antenna
applicator is designed for hyperthermia treatment of breast cancer and integration
with MRI system for treatment planning and online therapy guidance. E field inside
the water tank without phantom was measured and simulated, there is B field focus
at the center region of the applicator and the E field focus can be steered by changing
the input magnitude and phase for five antenna. Temperature distribution of homo-
geneous gel phantom in one plane were measured by MRI scanner with five different
focus locations and the simulated temperature distributions match the temperature
map for the five focus locations. The fat/ tumor heterogeneous phantom was made
with gel and steak and also heated with this applicator with temperature monitor
by MRI. The measured and simulated temperature both show the peak temperature
region locating at the fat/ tumor boundary and heat transfers into the tumor region
from the boundary until reaching the balance. Simulation of the treatment on the
patient shows treatment planning can dramatically improve the applicator heating

performance on the tumor. The E field point maximization method gave better re-

70

sults than others and the results is robust with the point location selection.

71

CHAPTER 5

HYBRID RF / US PHASED ARRAY APPLICATOR (1)

5.1 Motivation

External ultrasound (US) phased arrays produce small focal spots while achieving
high temperatures in deep-seated tumors, but these devices heat relatively small vol-
umes. Ultrasound has good penetration in the ductile tissue in the breast because
of almost no reflection on the tumor/ normal tissue boundary and weak reflection on
the water/ tissue boundary. For heating breast regions, RF devices have limited pen-
etration and temperature peaks are generated outside of the tumor[37] due to the
reflection on the tumor/ normal tissue boundary, which is caused by an impedance
mismatch. Measurement and simulation results show that the RF phased array is a
regional heating device that beats a large portion of the intact breast. Adding local
heat produced by an ultrasound phased array is expected to improve the tempera-
ture distribution within the tumor. The hybrid approach will exploit the regional
heat generated by the RF applicator that increases the temperature of the arterial
blood supply, thereby reducing the power requirements for the ultrasound compo-
nent. The large heated region created by the RF phased array combined with small
focal spots generated by the ultrasound phased array provides multiple opportunities
for optimization of the temperature distribution produced by hybrid RF/ US phased

array devices.

When these two modalities are combined, significantly improved temperature dis—
tributions are observed in simulated hyperthermia treatments of LABC. Power dis-
tributions are simulated for a hybrid applicator consisting of an RF phased array
and a planar square US phased array, and temperature responses are evaluated for

hyperthermia treatments in the intact breast. In this paper, a novel RF/ US hybrid

72

device is proposed for hyperthermia treatment of LABC. This device include a four—
antenna RF phased array and a 2D planar US array with 5,000 circular elements.
The simulation results suggest that this hybrid device deliver more therapeutic heat

into tumor than RF device and US device used along.

5.2 Applicator and 3D patient model

5.2.1 Planar ultrasound phased array

The planar US phased array, driven by lMHz time—harmonic excitation and mounted
on the side panel of the hybrid applicator is shown in Fig. Figure 7.1. The planar
phased array is square with 16.275 cm edge length wide and it comprises of 217 x 217
circular elements with radius A / 4. The center-center spacing between elements is A/ 2.
The compact structure of the array prevents the unwanted grating lobes which can
cause hot spots in normal tissue. The acoustic pressure fields are computed on the
grids sampled at a rate of A/ 2. The distance from the center of the phased array to
the center plane (a: - 0) of the applicator is about 11 cm, the distance to the my plane
with (z :2 0) is about 8.2 cm and there is no offset along y axis. The face of the US

array is parallel to the yz plane and mounted inside of the applicator.
5.2.2 RF phased array and applicator

The design of the RF / US hybrid applicator is based on the RF applicator in [37]. The
size of the top panel is about 33.6cm x 22.5cm. The bottom plane is 18.80m x 22.5cm.
Both the top panel and the bottom panel are parallel to the my plane and the depth
of the of the is about 1700172. In the two hexagonal side panels, which are both
parallel to the yz plane and symmetric to the xz plane, the top edge is 33.6 cm long,
the bottom edge is 18.8 cm long, higher edges are 3.3 cm, lower edges are 15.3 cm,
the two lowest obtuse angles are 150°, the two top angles are right angles, and the
other two obtuse angles are 120°. Except the top plane, other planes are covered

by plastic (Lexan) with 0.50m thickness. Four U-shaped antenna are mounted on

73

 

Y(cm) 0 0 X (cm)

Figure 5.1. The hybrid applicator with 4 U-shaped antennas mounted on the four
sides of the plastic tank, a planar ultrasound phased array mounted on a side panel.

four sides of the applicator shown in Fig. Figure 7.1. The arms of three antenna
are parallel to the z axis. The arms of the fourth antenna on the hexagonal panel
are parallel to the y axis. The planar ultrasound phased array is mounted on one
hexagonal panel. During the treatment, the applicator is filled with deionized water
with a constant temperature 39°C. During the treatment, the patient lies down prone
to the applicator and the body axis is parallel to the y axis and there is 1-2 cm air

gap between the chest wall and the water surface.

5.2.3 Patient model

The 3D patient model is extracted from the MR images of the patients. The 3D
model has two steps: 1) Extract the 2D contours for each organ or bone from every
slice image via ct_con software. 2) Redraw these contours in HFSS and connect them

into solids.

The program CT_con can manually extract tissue contours from medical images.

74

 

Figure 5.2. One slice of MRI images of the patient with contours. Those MR images
were obtained when a patient lay prone to the hyperthermia applicator mounted in a
medical couch. The contours are extracted manually with the Matlab based program
’CT_con’.

An example is given in Fig. Figure 5.2, which shows one MRI slice image with
contours for breast and body. This MR image was obtained from a patient laying
prone to the hyperthermia applicator (without US phased array) mounted in a medical
couch. After all the contours were extracted, a 3D model was reconstructed in HFSS.
This reconstruction is shown in Fig. Figure 7.4. The chest was included in this model,

since the chest influences the E field distribution.

This 3D model includes tumor, breast, external layer of chest and inner chest.
Different parts of the model are assigned different material listed in Table. Table
5.1. The tumor solid is assigned the tumor material. The breast is assigned the
fat material. The external layer of chest is assigned the fat material, shown in Fig.

Figure 5.2 in bright color. The inner chest is assigned the muscle material.

75

Table 5.1. Material properties for the breast model used in paper. The values of E,—
and a are obtained from [1| and the values of cp, 5:, and W are obtained from [2].

 

 

 

 

 

 

 

 

[ Fat rMuscle I Tumor 1
Relative Permittivity 6, 20.4 75 65
Electrical Conductivity a (S/m) 0.12 0.75 0.78
Specific Heat 0,, (J/kg/°C) 2387 3639 3639
Thermal Conductivity r; (W/m/°C) 0.22 0.56 0.56
Blood Perfusion W(kg/m3/s) 1.1 3.6 1.8

 

 

 

 

 

5.3 Methodology

5.3.1 Pressure field calculation and mode scanning techniques

Using a linear model, the total acoustic pressure field is the superposition of the
pressure field of each element. Since the size and shape of the elements in the phased
array are the same, the pressure field generated by elements are identical to each other.
The pressure field generated by a single element source is calculated by combining
the fast near-field method (FNM) [39] and the angular spectrum approach(ASA)[40,
16, 41]. Because of the equivalent sizes of the sampling grids and the array elements,
the total field can be calculated by shifting and adding the field generated by the
center element in a broader dimension. Due to the large number of elements and grid
points, ASA is incorporated to accelerate the pressure field computation, since ASA
utilizes FF T’s to propagate acoustic fields. The pressure field at the first transverse
plane is calculated by FNM and linearly propagated. Computational planes with a
uniform size are slightly larger than the aperture (Zero padding eliminates circular
convolution artifacts for short propagation distances, then the same effect is achieved
by angular restriction for longer propagation distances). Both the array and the
planes are normal to and centered at the z axis. The three dimensional field are
calculated within an hour (Intel P4 2.4GHz CPU and 1GB memory) by using the

combined simulation approach.

76

 

,4» -- 4-

\

’9
/

 

   

 

 

]
/

I / \

l / ‘

] r = 15 cm
/ e
/ .

l x :

' .

‘05 Phase 's—ch—eme—fdffofirfocfisfi—rfidegrfid (b) 3 groups mo e scan fovcusing in a circle

 

 

 

with two cross phase symmetric planes with radius 1.5 cm

 

Figure 5.3. Phase a scheme for four focus mode scan and focusing strategy for planar
ultrasound phased array

In order to reduce the intervening tissue heating, the mode scanning technique [42]
cancels the fields in symmetric planes to reduce unwanted heat accumulated between
the array and the tumor region. In the mode scanning technique, the US array is
subdivided into the 4 regions shown in Fig. Figure 7.3a. Region I synthesizes a single
focus at point ($,y,z). The driven phase distribution for the region II is symmetric
to region I about 3; axis with 180° phase difference and focus at (-$,y,z), region VI is
symmetric to region I about 2: axis with 180° phase difference and focus at (:r,-y,z),
and region III is symmetric to region I about center point and focus at (-2:,-y,z). The
array with four regions generates 4 focal spots and cancels the pressure field in the
xz plane and the yz plane. Multi focus groups, shown in Fig. Figure 7.3b, can afford
better heating pattern than single focus group. Three groups (with 12 focal points)

are used in the present simulation.

After the 3D pressure field is computed, the steady state temperature field is

77

simulated by solving BHTE equation using the SAR pressure as an input. The SAR
generate by the pressure field is calculated by SARP = i |P|2, where c is the velocity

of the acoustic wave, p is the density and a is the acoustic attenuation coeflicient.
5.3.2 B field computation

The E fields radiated by RF antenna are computed in the water tank and breast
model by the commercial software HFSS (Version 8.0, Ansoft Corp, Pittsburgh, PA).
The detail about setting up model and ABC boundary in HFSS is described in [37].
The ultrasound phase array is included in the E field calculation as the copper plate
with the same size as the array. The E field distribution is computed for each separate

antenna. The amplitude of the total E field can be computed from Eq. 5.1.

N

3(7)) = Z A, exp(r,o,~) 0 EA?)

i=1

(5.1)

 

In this equation, A,- and 99,- are the amplitude and the phase of each antenna re-
spectively, and N is the number of the antennas. Once the 3D E-field is computed, the
SAR is calculated from the amplitude of the total E field, the electrical conductivity,

- ~ ('7’ —’ 2 -—> .
and the densrty of the tissue according to SARE = 5% [E] , where 0( 7' ) is the

conductivity, p(?)is the density of the material, 7’ is observation point.

5.3.3 SAR and Temperature Optimization

In this hybrid approach, the SAR generated by electric field and the temperature
distribution produced by the total SAR (the sum of the SAR of E field and the SAR
of pressure field) need to be Optimized. The SAR generated by acoustic pressure
field is optimized by the mode scanning technique. To optimize the SAR by RF
source, one optimization method reported in [19] for Optimizing SAR is easy to use.
The matrices '7’,n’,'y,n are calculated from the E fields separately from 4 antenna
and the computational detail is shown in [19]. All these four matrices are 4 x 4

matrix and not related with the excitation amplitudes and phases of the 4 antennas

78

(A1, A2, A3, A4, 4,91, 992, 993, 4,04). The total SAR in the region of interest is a function of
the four matrices and the antenna excitations. A Matlab function fmz'ncon is used to
minimize the ratio of the average SAR in normal tissue to the average SAR in tumor
tissue.

There are several optimization methods]43, 44, 45, 18] to optimize the temperature
distribution. There is no existing solution to optimize the SAR or temperature for
RF/ US hybrid source simultaneously. In this simulation, the weight of the total RF
energy and the total US energy is optimized to minimized the temperature objective
function. The temperature optimization function used here is in the same vein as
those proposed in 1iterature[46, 47, 48, 49, 50]. The objective is to achieve a specified
tumor and normal tissue distribution: all points in tumor tissue at 43°C, and all

points in normal tissue at or below 42°C.

5.4 Results

If the amplitude of power input of each antenna is set to unity and the phase is set to
zero, there is an E field focus on the central axis of the water tank about 3 cm below
the water surface where the patient breast is placed. Because of the patient motion
and breathing, the positioning errors and that the tumor position is not exactly at the
center region of the breast; therefore it is necessary to steer the amplitudes and phases
for all the antennas to move the E field focus into the tumor region. ( Applicator
focus and practical background)

In the FE simulation, the E field is calculated in the whole 3D region inside the
ABC boundary. The SAR and temperature are evaluated in a rectangular region
with a: from -8.0 cm to 8.0 cm, y from -8.0 cm to 8.0 cm and z from -10.0 cm to 0 cm
on the 0.2 cm interval spacing. This region (16.0 cmx 16.0 cmx 10.0 cm) contains the
breast, the tumor, a portion of skin and deionized water. The pressure field is only

computed in this region specified above. Five faces of the rectangular region are in

79

 

 

Figure 5.4. hybrid applicator including 4 RF antenna and US phased array

the water boundary and the top face of this region is located inside the body. During
calculating the temperature, the five water boundaries are set at a fixed temperature
(39°C) and the sixth boundary is set as the body temperature (37°C). ( Introduce

the thermal region and temperature boundary)

After the B field is calculated for each antenna, the SAR ratio ( the average SAR
inside tumor to the average SAR in the normal tissue) is maximized by the SAR
optimization method. The power input magnitude for four channels are 1, 1, 1, 1 and
the phase are 70.40, -63.8°, 113.5°, 0°. The SAR ratio (E field only) is about 0.45.

( Give the detail of RF SAR optimization results)

In order to reduce intervening tissue heating, ultrasound focal spots are placed at
the distal half of the tumor. Here we assume the tumor is a sphere with 5 cm diameter.

According to the model, the coordinate of the tumor center is around (ray, 2) =

80

[2, 1, —3]cm and the center point of the US phased array is (:13, y, z) 2 [—11,0, —8.2]cm.
The focal spots should be placed around the line which connect the center point of the
array and the tumor center and in the half of tumor which is far from the US array.
The center of the focus ring is selected manually (:1), y, z) = [3.5, 1, -4]cm which gives
better heating performance for this hybrid applicator. { Give the position of the array

and the detail of ultrasound mode scan focusing)

The results in the x f 0 plane and y — 1 cm plane are shown here and one
more plane close to the tumor boundary in x direction is also shown here. Figures
Figure 5.5, Figure 5.6, Figure 5.7 show the temperature distribution on the y :
0.5 cm plane. Figs. Figure 5.5 show the RF-only results and the B field is sealed
with maximum temperature 44°C with the SAR generated by E—field optimized with
the Weisma SAR method[19]. Fig. Figure 5.5 shows the temperature distribution
calculated from the RF-only SAR in this plane. In contrast with the RF-only result
in the same plane (y = 10 mm), the US-only results are shown in the fig. Figure
5.6. The temperature distribution calculated from the US—only SAR with the same
boundary is shown in Fig. Figure 5.6. In these results, the pressure field is sealed
with maximum temperature 44°C. The results on the same plane (y = 10 mm) with
hybrid source are shown in Fig. Figure 5.7. Figures Figure 5.8, Figure 5.9, Figure
5.10 show the temperature distribution on the yz plane (x —.—. 0). (Introduce the view

temperature results of RF only, US only and hybrid in two typical planes)

Figures Figure 5.11, Figure 5.12, Figure 5.13 show the temperature distribution
inside of the breast and tumor. The mesh grid indicates the tumor contour and the
solid surface shows the isothermal surface. Fig. Figure 5.11 shows tumor model and
the 3D View of the isothermal surface with temperature 42°C with RF-only source
which has the same power input and scale as Figs. Figure 5.5 and Figure 5.8. Fig.
Figure 5.12 shows the 3D view of the isothermal surface with temperature 42°C

with US-only source which has the same power input and scale as Figs. Figure 5.6

81

1°C]

 

 

80 _- 44
-43
4o
‘42
g ~41
15,
X
—40

    

-8
—‘l)00 —75 —50 —25 0
Z(mm)

Figure 5.5. Temperature results in the xz plane (y T 10mm) with RF only

82

 

[°C]

70 -—« 44
43
35 - 42
’g {:41
g 0
X

 

 

    

"1‘30 —60 —40 -2o 0
2 (mm)

Figure 5.6. Temperature results in the xz plane (y : 10mm) with US only

83

1°C]

70 '—-' 44
-43
35 - 42
‘g i 41
g o
X

 

 

    

{go —60 —40 -2o 0
Z(mm)

Figure 5.7. Temperature results in the xz plane (y : 10mm) with hybrid RF/ US

84

[°C]
80

40

 

    

-8

-?OO -75 -50 -25 0
2 (mm

Figure 5.8. Temperature results in the yz plane (x : 0) with RF only

and Figure 5.9, Fig. Figure 5.13 shows the 3D view of the isothermal surface with
temperature 42°C with RF/US hybrid source which has the same power input and
scale as Figs. Figure 5.7 and Figure 5.10. (Introduce the 3D view of RF only, US

only and hybrid temperature results)

5.5 Discussion

The intact breast is placed at the geometric center of the phased array and the
antennas are close and face toward the breast. The US phased array is mounted on
the side wall, not the bottom wall of the tank. Because the acoustic pressure wave
will be reflected by the ribs (acoustic impedance mismatch at bone/ tissue boundary),
which will cause pain to the patients, if the array is mounted at the bottom of the

applicator and faces the patient body. This reflection will not occur if the array is

85

1°C]

70 ~44
-43
35 -42

 

 

    

—7
£0 -60 -40 —20 0
Z(mm)

Figure 5.9. Temperature results in the yz plane (x T 0) with US only

86

 

 

  

70 ‘—'44

r 143

35 142

E '41
E. 0 .

X E 40

_35 39

38

37

-7
£0 -60 -40 —20 0
Z(mm)

Figure 5.10. Temperature results in the yz plane (x — 0) with hybrid RF/US

87

Vn’s's'L'v'

'A

‘m
, .
r:
a
A
'4-
a

L

 

Figure 5.11. 3D view of the temperature distribution with optimized RF source only.
The solid surface represents the 3D region over 42°C temperature heating by the RF
only. The external mesh contour is the tumor. The power input magnitude for four
channels are 1, 1, 1, 1 and the phase are 704", -63.8°, 113.5°, 0°. The SAR ratio
(tumor/ normal) is about 0.45.

88

 

Figure 5.12. 3D view of the temperature distribution with only US source, the solid
contour represents over 42°C temperature contour.The external mesh contour is the
tumor. The US phased array is located at(:r, y, z) = [—11,0, —8.2]cm; The center of
the foci ring is at (:r,y, z) = [3.5, 1, —4]cm.

89

 

Figure 5.13. 3D view of the temperature distribution with RF/US hybrid source,
the solid contour represents over 42°C temperature contour and the external mesh
contour is the tumor. The over 42°C region covers more of the tumor region than
that with ultrasound only and that with RF only. The total SAR of hybrid heating
is the sum of 70.14% of the SAR by B field and 67.82% of the SAR by ultrasound
pressure field.

90

mounted on the side wall of the tank. (Explain why the array is mounted at the side

wall of the tank)

The parameters of the US phased array, such as element size, gap size and array
size, are carefully chosen. Small array sizes have small window to deliver ultrasound
power and the power density on the path to the target region will be high, which can
cause intervening heating. Large element center to center spacing could cause severe
grating lobes and the element size has stronger effect than the gap size. Small element
center to center spacing has better performance, but it gives a large number of the

element in the ultrasound phased array. (Explain why picking these array parameters)

The plastic tank and the deionized water are helpful to the focusing of the E field.
Because the E field radiated by the antenna is reflected by the water/ plastic boundary
and the plastic/ air boundary, so the B field inside of the tank will be much stronger
than that without these boundaries. So the absorbing boundary condition can be
move close to the FE model with low computational error. The field radiated by one
antenna also is reflected by the metal surface of other antennae. Because of those
boundaries (air/ water, water/ plastic) and reflections, standing waves exist inside of
the applicator. Actually those standing waves have a dominant effect on the E field
distribution. The overlapping E fields of those standing wave of the RF phased array
generate the E field focus. One important advantage of this 4-antenna applicator is
that the E field can focus close to the center of the water/ air surface where the breast
is located, compared with other applicator design (spherical shell, cylinder, etc). The
shape and the material content of the chest does effect the E field distribution in the
breast and the penetration depth into the chest wall. From the computed E field and
SAR distribution, the power deposition in the chest wall is very low, because the E
field decay very fast in the breast. (Explain the design of the applicator and give the
factors which efiect the E' field distribution)

The E field and SAR inside breast and tumor, which are not shown here, change

91

gradually and decay in the direction to the chest. These temperature in Figs. Figure
5.5, Figure 5.8 and Figure 5.11 indicate that B field by RF antenna array can heat the
the proximal region of the tumor very well, the temperature peak is located inside of
the tumor and close to the tumor/ fat boundary. Most of the over 42° region is inside
the tumor, but the deep region (close to the chest wall) of tumor is not accessible by

RF power. ( Talk about the heating performance by RF)

Single-spot-focus scanning (multi-focus) approach can produce an optimal time-
averaged absorbed power distribution for tumor heating. But the average intervening
heating is also increased as the focus moves around and the power deposition at
the focal points are still high which cause sharp temperature peaks there. Mode
scanning cancels the pressure field in the phase symmetric planes, thus reducing the
intervening heating and achieving low power deposition on the focal points (1 / 4 of
single spot focus power). Mode scanning techniques with 12 foci in this simulation
are employed to heat tumors whose dimension are much larger than one single focus.
In the temperature distribution in the Fig. Figure 5.6, the power deposition right on
those foci are actually lower than the peak power deposition region which is located
about 2 cm behind the foci ring (close to the US array) and is also clearly indicated in
Fig. Figure 5.7. That is the reason that the foci are placed beyond the tumor region
and inside normal tissue. In Fig. Figure 5.12, the shape of the over 42°C regions
heated by US power are not symmetric: the part in the tumor shrinks more than
those outside the tumor, because of different blood perfusion. Although those foci
outside of the tumor do not heat the tumor directly, they can elevate the temperature
in the deep region of tumor. In this simulation, the upper limit of the temperature is
set as 44°C, which will alleviate the pain caused by thermal treatment. { Talk about

the heating performance by US)

Although some of the over 42°C region heating by US only is outside the tumor,

the over 42°C in the hybrid heating shown in the Fig. Figure 5.13 almost fills the

92

whole region of the tumor. Comparing the temperature distribution by RF / US hybrid
applicator with that of the RF only or US only, the over 42°C temperature distribution
by hybrid method covers not only the forepart of the tumor, but also the deep part
of the tumor. The hybrid method decreases the total RF power and the hot spot
temperature caused by the E field. The hybrid method also decreases the power
deposition at the ultrasound foci and increases the size of the heating region with the

44°C up temperature limit. (Talk about the hybrid heating performance)

5.6 Conclusions

The 3D FE model for the RF/ US hybrid applicator was formulated. Also, the 3D
patient models were extracted from the patient CT/ MR images. The E field inside
and outside of the applicator and inside of the patient model is obtained based on the
FE model and the acoustic pressure field in the thermal model region was computed
with a mode scan technique employed for US phased array heating strategy. The 3D
temperature distribution is calculated in the thermal model based on BHTE. The
SAR generated by E field is optimized and the weights of US power and RF power
are also optimized to minimize the temperature objective function and keep the peak
temperature lower than 44°C. Comparisons between RF/ US hybrid method and RF-
only or US-only show that the hybrid heating strategy can heat the whole region of
the LABC tumor better that US or RF alone.

93

CHAPTER 6

SECTOR-VORTEX SCANNING FOR A LARGE SQUARE US
PHASED ARRAY APERTURE

6.1 Introduction

In previous chapters, the RF / US hybrid method is introduced and mode scan and
imaging focus methods are used to generate pressure focus to heat the tumor. But
There is still some disadvantage with those methods, such as the focus size is too
small and peak is too sharp. In this chapter, the ultrasound focusing scheme for
planar array is mainly talked about. How to achieve large heated region is one of the

major problems for ultrasound hyperthermia.

Cain and Umemura [51, 52] proposed a rotating phase scheme (<15 2 M 6) for a
concave sector-vortex ultrasound phased array applicator. This strategy generates
a controllable ring-shaped pattern in the focal plane. In this approach, the total
pressure field is approximated by an Mth-order Bessel function and the size of the
focus (or the radius of the main lobe) in the focal plane is controlled by the mode
number M. The sector-vortex array contains a small number of trapezoidal elements,
and relative to square planar phased array structures populated with square elements,
the sector-vortex array is diflicult to build and calibrate. Although the focus size in
the focal plane is adjustable with the sector-vortex array, the location of the focus is

fixed and the size of the focus along the center axis is also not adjustable.

A large ultrasound phased array with 500 square elements was recently reported
in [53]. With new PZT fabrication technology and circuit designsl54], the number
of the elements in the ultrasound phased array for therapy could grow very large
(71,000 or even more) and the driving electronics system could become much more

compact. The implementation of large phased arrays for hyperthermia could offer

94

flexible control of the focus with a large scan angle while eliminating the formation
of grating lobes. Furthermore, 2D planar arrays with square elements are readily

available for thermal therapy applications.

A new phase scheme prOposed herein can directly synthesize a controllable focus of
pressure field for 2D planar square ultrasound array with square elements or circular
elements. The shape of the focus generated by these large arrays approximates that
generated by a concave sector-vortex array. In addition, the focus can be steered off
the axis and / or along the axis and the size along the axis can also be controlled. In this
paper, a prototype ultrasound phased array applicator for hyperthermia consisting of
73 x 73 square elements generates a broad focus in the target tumor region. Computer
simulations of the ultrasound pressure fields generated by this array are calculated
with the fast near-field method (FNM), and the resulting temperature distribution is
computed with the bio-heat transfer equation (BHTE). Simulation results show that
the combination of several modes can heat large tumors (with radius of 5 cm) with a

broad, uniform temperature distribution.

6.2 Phasing Scheme

The ultrasound phase array is square planar array with N x N square elements. The
square element has an edge length a and the gap between two element is b. The
elements are arranged on the grid points shown in Fig. Figure 7.2. Each element
with index (i, j) is driven by a sinusoidal signal with the phase for each channel gm,-

given by

a, = 1110,,- — mail, (6.1)

for i r 1, 2, ---, N andj : 1, 2, ---, N, where (i,j) is the index of the element,

¢,,-is the phase of the driven signal, 6,7 is the angle of the element center shown

95

 

 

 

 

 

 

 

A l-2.j+2 (Pl-2,j+2

 

Figure 6.1. Schematics of planar array with rectangular element.

in Fig.Figure 7.2, M is the mode number, k is the wavenumber in water, a is the

coefficient for focusing, (1,3 is the distance from the element center to the focal point

 

(difj = \/(:r,j — at;)2 -+- (y,,~ — y,)2 + 2}), (13,-, y,,-) is the coordinate of the (i,j) element
center, and (:1: f,yf,z f) is the coordinate of the focus center. The first term (mode
term) of Equation.7.1 shows that the phase on each element rotates [M [times per

rotation around the center point of the array [51, 52].

The second term (focusing term) of Equation.7.1 works as a concave-converging
acoustic lens shown in Fig. Figure 6.2. With this phase shift, the new wave front of
the acoustic wave radiated by the array is a spherical shell and the spherical shell wave
front converges to the focus specified by (:r,,y,~, 24,) shown in Fig. Figure 6.2a. With
simply changing the location of the focus, the ultrasound focus can be steered ofl the
center axis or steered along the axis shown in Fig. Figure 6.2a. The coefficient of the
focusing term can control the extension of the focus along the center axis. Adjusting
the a coefficients, the shape of the shifted wave front will be changed and the focus

will extend or shrink along the center axis. When 0 equals one, the initial shifted

96

wave front is spherical. Base on the focusing term, the mode term (the first term of

Eq.7.1) modulates the pressure field distribution radiated by the phased array.

6.3 Methods

6.3.1 Pressure field calculation

The 3D pressure field radiated by one element is precisely calculated with unit mag-
nitude is calculated by the fast near—field method]55]. The difference of the pressure
field between any two elements are the complex amplitude and spatial shift. Once
the pressure field by one element with unit magnitude driven signal is calculated as
a sample pressure field, the pressure field by any other channel is the spatial-shifted
sample pressure field times the complex amplitude. The total acoustic pressure field

is the superposition of those pressure fields.
6.3.2 Patient model and thermal calculation

A cross section of the tumor model is shown in Fig.Figure 6.3. The material properties
used in temperature calculation is listed in Table. Table 6.1. The tumor is a sphere
with 4cm diameter, 6.2 cm deep from the skin and the center point of the tumor is
at [0,0,6.2] cm. The thickness of the skin layer is 0.4 cm, the fat layer is 1.5 cm, the
muscle layer is 1.5 cm and the viscera layer is 7.5 cm.

The temperature distribution is computed from the SAR distribution using the
Bio-Heat Transfer Equation (BHTE). The final temperature can be computed with a
steady-state finite difference method]56]. In order to accelerate the calculation, suc-
cessive over-relaxation is utilized in the simulationl43]. Since the water is maintained
a constant temperature, the thermal computational region is cuboid truncated from
the FE region and this cuboid region contains the whole breast model and some wa-
ter around the breast model. The six boundary faces of the cuboid region are set as
constant temperature for BHTE, the top face of the thermal region, located in the

body, is set as 37°C and the other five faces are set as 39°C.

97

 

'p I l
6886/ I
ll- ”‘9 I»? I
Q II“ I l
. $ 09 I
7’ *e‘ 11
II \\‘~\-\ I 8
l \\‘\ I
1‘s 8/ 7* I? ' 01
l 6 Ԥ I
Us; 029 ~. .2
’ ”be ”to ~.‘ : m
g e/ G I x‘ , g
4 00/. er ~\ I
l 0’.- f ‘s‘ I
1 \\ I
l ‘\ I
In ‘w:
1]] Center of the focus
1.
|I

I

l

I

I

n I
I

' \ I
I“ I
a» .
h |

. l
a l
. l
I

I

l

l

I

I

I

(a) focusing stratagy

 

l
. I
.0 ,/ I
[IR 6689/ I
‘\\‘/I;Sb/fi [ -n
71‘ ’0 1 O
I :t\ ' 8
/ \\ I _
D ’I/f s/e ‘§§ ['0
J, 076 ‘Q\ . 5
. II f0 0’0 ‘§:\ I 3
. , 09/0 609“ g m
cl O" f \§\
_1 0 . to ‘t\ I
‘4 I 0’ ‘§\\\ i
m ‘1. L
r =
1 ““ '
I \\‘\ '
1 ‘¥
Ll] Center of the focus
I:
I
b
1

I
I
l
l
l
l
I
I
I
I
I
|
l
O

(1)) Steering the focus off axis or along the axis

Figure 6.2. Focusing strategy and steering the focus

98

 

VISCERA

 

MUSCLE

 

FAT

 

 

SKIN

 

 

Figure 6.3. A cross section of the thermal model used for temperature calculation.
The material properties is listed in Table. Table 6.1. The tumor is a sphere with 4cm
diameter, 6.2 cm deep from the skin and on the z axis. The thickness of the skin
layer is 0.4cm, the fat layer is 1.5cm, the muscle layer is 1.5cm and the viscera layer

 

 

 

 

 

 

 

 

 

is 7.5cm.
skin fat muscle viscera tumor
Blood Perfusion Wb(kg/m3/s) 0 4.0 4.0 4.0 4.0
Thermal Conductivity k(W/m/°C) 0.21 0.16 0.42 0.55 0.56

 

Table 6.1. Material property for human tissue and the blood specific heat is

c], = 4000J/kg/°C

99

 

Mode 4 Mode 8 Mode 12

    

-60 .
_-3o -
- A
x 30
60 60
20 20 20 so 140 200

140
82 (mm) 2 (mm)

60Depth plane

2 (mm)

-60

-30
0
30

6350 30" "“30 so 6.60 .30 3o 60
X(mm)

Y (m)
Y ‘m’a,

CI

     

6
-60 -30 0 30 60
X(mm)

Focal plane

Figure 6.4. Pressure field in the focal plane and the depth plane for different modes.
The focus centers of these modes are all at [0,0,12] cm (with a = 1). Top row is the

depth plane and the bottom row is the focal plane. From the left to the right, they
are mode 4, mode 8 and mode 12.

100

6.4 Results

The driven frequency for the planar array is 1 MHz in the following analysis. The
acoustic field and temperature are evaluated in 3D domain with a: from -6 cm to 6

cm, y from -6 cm to 6 cm and z from 2 cm to 20 cm. on the 0.075 cm interval spacing.

6.4.1 Pressure field patterns

With the same focal point location at [0,0,12] cm, the pressure field distribution on
the depth plane and focal plane for different modes (M : 4, M : 8 and M = 12)
are shown in Fig. Figure 6.4. Actually the M = 0 mode is just the single spot focus
without phase modulation on the focal plane, which is not shown here. For Mode 4,
8 and 12, the pressure field distribution on the depth plane has a peak around 2 : 12
cm and decreases along +2 direction and —z direction. The pressure extension along
2 axis for mode 8 is about 4 cm. The top view of the depth plane of these modes
looks like a big ’X’, but the pressure field along 2 axis equals zero and there is low
pressure gap around 2 axis. From the Figure 6.4, the focus region of mode 8 is larger
than that of mode 4. On the focal plane (2 -_-.— 12 cm), the pressure field looks like
a rectangular ring with four pressure peaks at the four corners corresponding to the
four corners of the square array. The sizes of the pressure ring and the center hole
(or center low pressure region) increase as the mode number increases. The size of

the focal ring for mode 8 is about 3 x 3 cm.

With moving the focus center, the pressure field can be steered off axis and along
the center axis. Off axis steering and along axis steering can be done at the same time,
shown in FigFigure 6.5, by moving the focus center from [0,0,12] cm to [10,0,13.5] cm.
The reference pattern without steering is shown in the left column of Fig.Figure 6.4.
Adjusting the coefficient a, the focus is elongated when a is less than 1 or shrunken,
when a is larger than 1, shown Fig.Figure 6.6 with mode number 8 and the focus

center at [0,0,12] cm. The pressure field for mode 8 with a = 1 is shown in the center

101

X (mm)

 

 

6
(in) so 140 200
-60
.30
E
g o
>—
30
6350 .30 o 30 60

X (mm)

Figure 6.5. Steering the focus of mode 4 off axis and along axis, the focus center from
[0,0, 12] cm to [10, 0, 13.5] cm (with ()1 = 1).

102

column of Fig.Figure 6.4. The actual focus center moves close to the array when a is

larger than 1, and moves away from the array when a is less than 1.

6.4.2 Temperature distribution

The 4 cm diameter tumor, shown in F ig.Figure 6.3, is heated with different modes:
mode 4, mode 8 and mode 12. The coefficient a is set 1 for simplicity in these
analysis. The focus center of mode 4 is located at [0,0,10] cm, the mode 8 center
is at [0,0,10.5] cm and the focus center of the mode 16 is at [0, 0, 13] cm. In these
three 3D temperature distribution, sphere mesh is the target tumor and four 2D
square mesh plots are skin, fat, muscle, viscera boundary respectively. The black solid
surfaces show the isothermal surface of 42°C. The peak temperature are 44°C in each
calculation. The three dimensional field of one mode and temperature distribution in
12 x 12 x 18cm3 region are calculated within half hour (Intel P4 2.4GHz CPU and

1GB memory).

penetration

6.4.3 Frequency penetration study

For breast cancer treatment, ribs behind the breast could be over heated if ultrasound
array is in front of the body and enough ultrasound energy reach the ribs. Ultrasound
should offer enough heat to the deep seated tumor, but the energy on the rib surface
should be lower than a threshold value. Because different frequency has different
attenuation in tissue, ultrasound has different penetration depth for different ultra-
sound frequency. Generally low frequency ultrasound has high penetration depth,
vice versa. So the ultrasound frequency should be carefully studied for breast cancer

treatment with ultrasound phased array in front of the body.

With the same size array (12cm x 12cm) and some focus location (110mm from
the array), ultrasound at frequency range from lMHz to 1.5MHz were used to heated

the breast model, shown in Fig. Figure 6.8. In Fig. Figure 6.8, the top two mesh

103

  
  

6
(20 50 80 110 140 170 200
Z(mm) or.=1.25

6 s
30 5o so 110 140 170 200
Z(mm) a=0.75

Figure 6.6. a steering for Mode 8. The plot of a = lis shown in the center column
of Fig. Figure 6.4.

104

       

o
.20
Y (mm) 4° 40

(a

-2o°
X (mm)

0 l

-20 _200
Y (mm) '40 4° x (mm)
(C)

Figure 6.7. Temperature distribution for Mode 12, 8 and 4 (with a = 1). The mesh
plots are the tumor model and the black solid surfaces are the isothermal surface of
42°C. The peak temperature are 44°C in each calculation. a) Mode 12 with focus
located at [0,0,12] cm, b) Mode 8 with focus located at [0,0,11] cm, c) Mode 4 with

focus located at [0,0,10.5] cm.

105

contours are two ribs, the large external mesh contour is the breast, the center mesh
contour is the tumor model and the solid contour are the 3D temperature iso-surface
at 42°C. From the results, the 1.0MHz ultrasound has the most penetration depth
and the 42°C contour almost reach the rib, while the 1.5MHz ultrasound can not ofl'er
a good heating inside the tumor without causing serious hot spots in health tissue
because of poor penetration depth. The temperature contour of 1.1MHz result is very
close to that of 1.0MHz and 1.4MHz is also close to that of 1.5MHz. Ultrasound at
1.2MHz and 1.3MHz act somewhere between what 1.1MHz and 1.5MHz do and could

be the optimal frequency for breast cancer treatment.

6.5 Discussion

The two terms in Eq.7.1 with different functions work with each other very well.
For mode 0, the first term of Eq.7.1 equals zero, there is no mode modulation on the
focus and the size of the focus is close to one wavelength (1.5 mm for 1MHz in water).
When the mode number is larger than zero, the pressure field in the depth plane has
two peaks and there is a gap between two peaks, the pressure field in the focal plane
is a rectangular ring with one peak at each corner, shown in Fig. Figure 6.4. Because
any array element and it’s center axis symmetry element have 180 degree input phase
difference, the pressure fields generated by these two elements are canceled out along
the center axis. The total pressure field generated by the array is zero along the
center axis and very low close the center axis. The region inside of the focus with
large mode number (For example: mode 12) can not be heated well, which is shown
in Figs Figure 6.4 and Figure 6.7a.

Because the tumor could be off the center axis of the array, the shape of the tumor
could be irregular or the The peaks of the pressure field and peaks of the temperature
are not exactly in the focal plane and they are usually shifted small distance to the

array, the larger mode number has the larger shift distance. In order heat the tumor at

106

 

(e) 1.4MHz (f) 1.5MHz

Figure 6.8. The heating pattern for different frequency range from 1MHz to 1.5MHz.
The array size is 12cm x 12cm, the gap between array and breast is 3cm and the focus
is 11cm from the array. The top two mesh contours are two ribs, the large external
mesh contour is the breast, the center mesh contour is the tumor model and the solid
contour are the 3D temperature iso-surface at 42°C.

107

the same location, different modes have different focus center, shown in Figs. Figure
6.7. From the Fig.Figure 6.5, this phase scheme can also move the focus around the
target region without the shape changed. F ig.Figure 6.6 shows the phase scheme has
the ability to control the focus size along the center axis. Large mode number usually
has a large extension along the center axis, which will result in hot spot outside the
tumor. Adjusting the a coefficient for large mode number can constrain the power

deposition within the tumor region.

With the mode number increasing, the size of the focus is enlarged, shown in Figs.
Figure 6.4, and the size of the over 42°C temperature region is also increased, shown
in Figs.Figure 6.7. From the temperature distribution, large number mode (Mode
12) has hollow region which is not well heated by this mode and small number mode
(Mode 4) can a small solid region very well. Combing the SAR of different modes and
optimizing the weight of each mode, the over 42°C temperature region could extend

the entire tumor region with 5 cm diameter without intervene heating problem.

If the ultrasound energy comes from the side of the body and the propagation path
does not go through the rib region (for example the applicator in previous chapter),
the rib behind the breast will not be over heated by the ultrasound. While the ribs
could be over heated if ultrasound wave propagates through the rib region. The
frequency should be carefully chosen so that ultrasound Offers enough heat to the
deep seated tumor, but the energy on the rib surface is kept at a low level. In this
chapter, frequency in the range 1.0 " 1.5MHz were simulated for 12cm x 12cm planar
array by sector-vortex phase scheme with focus distance 110cm. Although the Optimal
frequency could be effected by the geometry shape of the array, the distance from the
array to the breast, the phase scheme method and the location of the focus, frequency
outside the range 1.0 " 1.5MHz can not give a good heating without overheating ribs
and health fat tissue. The frequency range 1.271.3MHz is the best choice for this

treatment setting in this chapter.

108

 

6.6 Conclusion

Two phase terms, phase for single spot focusing, sector-vortex phase scheme, are
combined together to generate the large and size controllable focus for the 2D planar
array, since the planar array is much easier to build than the concave shaped array.
The large size focus is a good choice for thermal treatment of the large size tumor
(diameter up to 4 or 5 cm). The focus steering property and different modes of this
new phase scheme are investigated to heat the 4 cm diameter tumor which is about 6
cm underneath the skin. This phase scheme also offers a way to adjust the extension of
focus along the axis direction. The ultrasound pressure fields are calculated with fast
near—field method (FN M) and the temperature distribution is computed by solving a
static bio-heat transfer equation. Different modes have difl'erent heating performance.
Large number mode can heat large region, but the region at the center of the focus is
not well heated and intervene heating problem will appear. Small number mode can
heat a small region with uniform temperature distribution. Mode number from 4 to

12 and the combination of them are recommended for hyperthermia.

109

 

CHAPTER 7

HYBRID RF/US PHASED ARRAY APPLICATOR (2) FOR SMALL
BREAST

7 .1 Introduction

External heating devices appropriate for deep hyperthermia in the intact breast in-
clude ultrasound phased arrays[22] and radio-frequency (RF) electromagnetic phased
arrays[28]. Ultrasound phased arrays are an appropriate local modality for heating
small targets in the breast (up to about 2cm diameter[27]) and has the ability to steer
the position of the focus in order to control the power deposition. In contrast, heat
generated by RF electromagnetic devices is delivered regionally across a much larger
area. RF phased arrays have been developed previously for deep hyperthermia in the
pelvis [28] and in the extremitiesl23], respectively. The hybrid approach will exploit
the regional heat generated by the RF applicator that increases the temperature of
the arterial blood supply, thereby reducing the power requirements for the ultrasound
component. The large heated region created by the RF phased array combined with
small focal Spots generated by the ultrasound phased array provides multiple oppor-
tunities for Optimization of the temperature distribution produced by hybrid RF/ US

phased array devices.

One of the hybrid applicator design for large breast [Liyong_Hybrid_2006] base
on the four antenna applicator[37] with ultrasound phased array mounted on one of
the side panel. The US array is mounted on the side face of the tank to reduce the
heating to the ribs caused by the reflection at the bone / tissue boundary. The US array
is operated at 1MHz and is tilted to place the intended tumor target at its geometric
focus. Mode scan method is used for ultrasound focusing. Temperature distributions

show that that hybrid structure is capable of producing a flat with small fluctuation

110

temperature distribution (43°C) in the tumor region. The intervening heating in
normal tissue is significantly reduced.

In this chapter, a new design of RF / US hybrid applicator is proposed for can-
cer treatment in small size breast. Significantly improved temperature distributions
are observed in simulated hyperthermia treatments of LABC. Power distributions
are simulated for a hybrid applicator consisting of an RF phased array and a planar
square US phased array, and temperature responses are evaluated for hyperthermia
treatments in the intact breast. In this paper, a novel RF/ US hybrid device is pro-
posed for hyperthermia treatment of LABC. This device include a four—antenna RF
phased array and a 2D planar US array with 4,000 circular elements. The simulation
results suggest that this hybrid device deliver more therapeutic heat into tumor than

RF device and US device used along.

7.2 Applicator

7.2.1 RF phased array and applicator

The design of the RF / US hybrid applicator is based on the RF applicator in [37]. The
size of the top panel is about 33.6cm x 22.5cm. The bottom plane is 18.8cm x 22.5cm.
Both the top panel and the bottom panel are parallel to the my plane and the depth
of the of the is about 17.00m. In the two hexagonal side panels, which are both
parallel to the yz plane and symmetric to the xz plane, the top edge is 33.6 cm long,
the bottom edge is 18.8 cm long, higher edges are 3.3 cm, lower edges are 15.3 cm,
the two lowest obtuse angles are 150°, the two top angles are right angles, and the
other two obtuse angles are 120°. Except the top plane, other planes are covered
by plastic (Lexan) with 0.5cm thickness. Four U-shaped antenna are mounted on
four sides Of the applicator shown in Fig. Figure 7.1. The arms of three antenna
are parallel to the z axis. The arms of the fourth antenna on the hexagonal panel

are parallel to the y axis. The planar ultrasound phased array is mounted on one

111

 

 

Figure 7.1. The hybrid applicator with 4 U-shaped antennas mounted on the four
sides of the plastic tank, a planar ultrasound phased array mounted on a side panel.

hexagonal panel. During the treatment, the applicator is filled with deionized water
with a constant temperature 39°C. During the treatment, the patient lies down prone
to the applicator and the body axis is parallel to the y axis and there is 1-2 cm air

gap between the chest wall and the water surface.

7.2.2 Planar ultrasound phased array

The ultrasound phase array is square planar array with N x N square elements. The
square element has an edge length a and the gap between two elements is represented
by b. The elements are arranged on the grid points shown in Fig. Figure 7.2. Each
element with index (i, j) is driven by a sinusoidal signal with the phase for each

channel (pg given by

(mi = M0,,- —‘ akdé (7.1)

112

 

 

 

 

 

 

 

 

 

Figure 7.2. Schematic of a planar array with rectangular elements.

for i + 1, 2, ---, N andj — 1, 2, ---, N, where (i,j) is the index of the element,
¢,jis the phase of the driven signal, 6,,- is the angle of the element center shown
in Fig.Figure 7.2, M is the mode number, k is the wavenumber in water, a is the

coefficient for focusing, difj is the distance from the element center to the focal point

 

(difj = \/(.r,-j — rf)2 + (y,,- — yf)2 + z}), (ripyu) is the coordinate of the (i,j) element
center, and (17,, 3);, zf) is the coordinate of the focal point. The first term (mode term)
of Eq. 7.1 shows that the phase on each element repeats [M [times per rotation around

the center point of the array [51, 52].

The second term in Eq. 7.1 focuses the acoustic energy as shown in Fig. Figure 6.2.
By performing these phase adjustments across the aperture, the wave front emitted
by the array converges to the focus specified by (a, y,, 2,) shown in Fig. Figure 6.2a.
The ultrasound focus can also be steered Off the center axis or steered along the axis
as shown in Fig. Figure 6.2a, where the coefficient of the focusing term controls the
extent of the focus along the center axis. Thus, adjusting the a coefficients changes
the shape of the shifted wave front. When (1 equals one, the initial shifted wave

front is spherical. The combination of the first and second terms in Eq. 7.1 therefore

113

 

focuses and modulates the pressure field distribution radiated by the phased array.

7 .3 Methodology

7.3.1 Pressure field calculation

Using a linear model, the total acoustic pressure field is the superposition of the
pressure field of each element. Since the size and shape of the elements in the phased
array are the same, the pressure field generated by elements are identical to each other.
The pressure field generated by a single element source is calculated by combining
the fast near-field method (FN M) [39] and the angular spectrum approach(ASA)[40,
16, 41]. Because of the equivalent sizes Of the sampling grids and the array elements,
the total field can be calculated by shifting and adding the field generated by the
center element in a broader dimension. Due to the large number of elements and grid
points, ASA is incorporated to accelerate the pressure field computation, since ASA
utilizes FFT’s to propagate acoustic fields. The pressure field at the first transverse
plane is calculated by FNM and linearly propagated. Computational planes with a
uniform size are slightly larger than the aperture (Zero padding eliminates circular
convolution artifacts for short propagation distances, then the same effect is achieved
by angular restriction for longer propagation distances). Both the array and the
planes are normal to and centered at the z axis. The three dimensional field are
calculated within an hour (Intel P4 2.4GHz CPU and 1GB memory) by using the
combined simulation approach.

In order to reduce the intervening tissue heating, the mode scanning technique [42]
cancels the fields in symmetric planes to reduce unwanted heat accumulated between
the array and the tumor region. In the mode scanning technique, the US array is
subdivided into the 4 regions shown in Fig. Figure 7.3a. Region I synthesizes a single
focus at point (z,y,z). The driven phase distribution for the region II is symmetric

to re ion I about y axis with 180° phase difference and focus at -:r, ,z , region VI is
g I!

114

; 375

 

symmetric to region I about 2: axis with 180° phase difference and focus at (z,-y,z),
and region III is symmetric to region I about center point and focus at (-:r,-y,z). The
array with four regions generates 4 focal spots and cancels the pressure field in the
3:2: plane and the yz plane. Multi focus groups, shown in Fig. Figure 7.3b, can afford
better heating pattern than single focus group. Three groups (with 12 focal points)
are used in the present simulation.

After the 3D pressure field is computed, the steady state temperature field is
simulated by solving BHTE equation using the SAR pressure as an input. The SAR
generate by the pressure field is calculated by SARp = i |P|2, where c is the velocity

of the acoustic wave, p is the density and o: is the acoustic attenuation coefficient.
7.3.2 Temperature Optimization

In this hybrid approach, the SAR generated by electric field and the temperature
distribution produced by the total SAR (the sum of the SAR of E field and the SAR
of pressure field) need to be optimized. There are several Optimization methods[43,
44, 45, 18] to Optimize the temperature distribution. There is no existing solution to
Optimize the SAR or temperature for RF/ US hybrid source simultaneously. In this
simulation, the weight of the total RF energy and the total US energy is Optimized
to minimized the temperature objective function. The temperature optimization
function used here is in the same vein as those proposed in literature[46, 47, 48, 49, 50].
The Objective is to achieve a specified tumor and normal tissue distribution: all points

in tumor tissue at 43°C, and all points in normal tissue at or below 42°C.

7- 4 Results

In the FE simulation, the E field is calculated in the whole 3D region inside the ABC
boundary. The SAR and temperature are evaluated in a rectangular region with z
from -8.0 cm to 9.0 cm, y from -6.0 cm to 6.0 cm and z from -6.0 cm to 4 cm on

the 0.1 cm interval spacing. This region (16.0 cmx16.0 cmx10.0 cm) contains the

115

X (mm)

 

so
20 so 140 200
Z(mm)

O

Y (mm)

 

-60 -30 0 30 60
X(mm]

Figure 7.3. Phase a scheme for four focus mode scan and focusing strategy for planar
ultrasound phased array

116

, I.‘

/ l

l f, ’
V .
v I l I"
', pl .
v r _ I
. --
\ ~ .
2
1’ l
.
/

 

Figure 7.4. hybrid applicator including 4 RF antenna and US phased array

breast, the tumor, a portion of skin and deionized water. The pressure field is only
computed in this region specified above. Five faces of the rectangular region are in
the water boundary and the top face of this region is located inside the body. During
Calculating the temperature, the five water boundaries are set at a fixed temperature

(39°C) and the sixth boundary is set as the body temperature (37°C).

After the E field is calculated for each antenna, the total E field is maximized
at the selected point. Because the four antennas are all in one plane (parallel to my
plane), the RF array lacks the steering ability along z axis and the E field focus can

Only be steered in my plane. The total E field outside the tumor region (in normal

117

health tissue) is much stronger than that in the tumor because of the impedance
mismatch between tumor and fat materials. Although the E field maximized point is
picked in the tumor region or outside the tumor region, the temperature peaks always
outside the tumor shown in Figs. Figure 7.5, Figure 7.6 and Figure 7.7. Fig. Figure
7.5 shows the isothermal contour at 42°C in the breast models with E field focus
at [5, 15, 0] mm. Fig. Figure 7.6 shows the same isothermal contour with E field
focus at [-5, 0, 0] mm and Fig. Figure 7.7 shows the same isothermal contour with E
field focus at [-20, 20, 0] mm. The E field maximized points around the tumor/ fat
boundary and outside the tumor give more control on the temperature peak location
than those inside the tumor. Figs. Figure 7.5, Figure 7.6 and Figure 7.7 shows that
different focus locations give different heating patterns, the temperature peaks are

distributed at different locations and outside the tumor.

In Figs. Figure 7.5, Figure 7.6 and Figure 7.7, the temperature rises in the tumor
region are about 374.5 degree. Since the temperature peaks around the tumor do not
overlap with each other, the superposing of the three RF SAR with different weights
([05, 0.3, 0.5]) can give higher temperature rise in tumor and the smaller hot spots
outside the tumor shown in Fig Figure 7.8. In Fig Figure 7.8, a large portion of tumor
is heated over 42°C and the isothermal contour looks like a ring around the tumor
in my plane (the plane contains the feed points Of the four antenna). Some portion

of the tumor such as the center tumor region and deep region of tumor are not well

Penetrated by EM.

With the sector—vortex phasing scheme, ultrasound generated by a large 2D planar
array can heat a relatively large region. Mode number 8 is chosen for this simulation
and the temperature distribution heated by mode 8 is shown in Fig. Figure 7.9. The
42°C region has a 5 cm extension along the z axis and average 2 cm diameter. The

tIllrnor region, which is not well penetrated by RF in Fig. Figure 7.8, is well covered

by US heating.

118

2 (mm)

 

‘. \
‘ 4965., g
‘fi‘ffi'tu °
’1 4 4‘ ”We
JZ‘ZQ’IK‘S
wars-5"“
‘KVAw v
an "‘
“.239?"

L‘
s

K‘s

'\
~v

5

 

50
“5° y (m)

Figure 7.5. Temperature results with RF focus at location 1 (0,10,0)

119

'

....
sv‘Wl’W
‘l .5 «uh
I .-

V

20a
0..

.20]

-4o~

2 (mm)

 

 

1! (mm)
x (mm)

Figure 7.6. Temperature results with with RF focus at location 2: (5,0,0)

120

 

 

1! (mm)

Figure 7.7. Temperature results with with RF focus at location 2: (5,0,0)

121

2 (mm)

 

 

-50

y (mm)
x (mm)

Figure 7.8. Temperature results with the combination of above 3 RF heating with
Weight [0.5, 0.3, 0.5]

122

2 (mm)

 

 

- I ”I
and?!
_ a! "’44

7
L‘\

5° y (m)

x (mm)

Figure 7.9. Temperature results with US mode 8

123

 

 

x (mm)

Figure 7.10. Temperature results with hybrid RF/ US

When the RF and ultrasound are combined together, the temperature optimiza-
tion method is applied to adjust the weights of RF SAR and US SAR to minimize the
temperature objective function. The hybrid SAR, which combine the total RF SAR
and ultrasound SAR with mode 8, gives very good heating pattern in tumor shown
in Fig. Figure 7.10. Almost 90% of the tumor region is heated over 42°C and the hot

spots in the health tissue are dramatically reduced.

Fig Figure 7.11 gives a more detail and insight view of the temperature distribution

in the breast model. From these three figure, there is a large region with temperature

124

 

over 43°C at the center of tumor. In Fig. Figure 7.11b and c, the contour of over
42°C follows the tumor boundary and ultrasound has majority contribution to the
temperature rise at the center region of tumor (also the center of the ultrasound focus

with mode 8).

7.5 Discussion

This hybrid applicator with ultrasound phased array mounted at the bottom is de-
signed for small breasts. Because small breasts do not have enough space for ultra-
sound wave converging in breast from the side position. So the most possible position
for US phased array is mounted at the bottom of the applicator in Fig. Figure 7.1 and
facing the breast. In order to reduce the pain caused by the pressure wave reflection
on the ribs, the Operating frequency is increased from 1MHz to 1.1MHz to increase
the attenuation on the pressure wave and the pressure on the rib bone surface is
reduced.

The intact small breast is placed at the geometric center of the phased array and
the antennas are close and face toward the breast. If the amplitude of power input
of each antenna is set to unity and the phase is set to zero, there is an E field focus
on the central axis of the water tank about 3 cm below the water surface where the
patient breast is placed. The four antennas are placed around the breast like a cage
and two Of them are tilted to move the E field focus along 2/ direction. If there is
no tilt and antenna are all vertical to my plane, the E field focus will be in the plane
containing all four feeding point of the array and this focus position is too low to heat
tumor in small breast.

The array in the applicator design[37] is lined up and the antenna arms are parallel
to each other. SO the E field in the breast region is always linearly polarized in
one direction which causes the hot spots aligned in the same direction. In current

arrangement of antenna, the E field in the breast region is still linearly polarized, but

125

" 1.7:.

i

 

Y (mm)

 

 

 

 

2 (mm)

 

(c) YZ view with x = 0 mm

Figure 7.11. 3 cut view of the temperature distribution with RF / US hybrid source.

126

 

the polarization direction is not fixed and can be control by changing the antenna
input. The total polarized E field can be any direction in the my plane. Figs. Figure
7.5, Figure 7.6 and Figure 7.7 show three polarization directions from one hot spot

pointing to another one.

The parameters of the US phased array, such as element size, gap size and array
size, are carefully chosen. Small array sizes have small window to deliver ultrasound
power and the power density on the path to the target region will be high, which can
cause intervening heating. Large element center to center spacing could cause severe
grating lobes and the element size has stronger efl'ect than the gap size. Small element
center to center spacing has better performance, but it gives a large number of the

element in the ultrasound phased array. (Explain why picking these array parameters)

The E field and SAR inside breast and tumor, which are not shown here, change
gradually and decay in the direction to the chest. These temperature in Figs. Figure
7.5, Figure 7.8 and Figure 7.11 indicate that E field by RF antenna array can heat the
the proximal region of the tumor very well, the temperature peak is located inside of
the tumor and close to the tumor/ fat boundary. Most of the over 42° region is inside
the tumor, but the deep region (close to the chest wall) Of tumor is not accessible by

RF power. ( Talk about the heating performance by RF)

Single—spot-focus scanning (multi-focus) approach can produce an optimal time-
averaged absorbed power distribution for tumor heating. But the average intervening
heating is also increased as the focus moves around and the power deposition at
the focal points are still high which cause sharp temperature peaks there. Mode
scanning cancels the pressure field in the phase symmetric planes, thus reducing the
intervening heating and achieving low power deposition on the focal points (~1/4 of
single spot focus power). Mode scanning techniques with 12 foci in this simulation
are employed to heat tumors whose dimension are much larger than one single focus.

In the temperature distribution in the Fig. Figure 7.6, the power deposition right on

127

i

 

those foci are actually lower than the peak power deposition region which is located
about 2 cm behind the foci ring (close to the US array) and is also clearly indicated in
Fig. Figure 7.7. That is the reason that the foci are placed beyond the tumor region
and inside normal tissue. In Fig. Figure 5.12, the shape of the over 42°C regions
heated by US power are not symmetric: the part in the tumor shrinks more than
those outside the tumor, because of different blood perfusion. Although those foci
outside of the tumor do not heat the tumor directly, they can elevate the temperature
in the deep region of tumor. In this simulation, the upper limit of the temperature is
set as 44°C, which will alleviate the pain caused by thermal treatment. ( Talk about

the heating performance by US)

7.6 Conclusions

A new RF/ US hybrid applicator geometry is evaluated for hyperthermia treatments
of locally advanced breast cancer. By scanning the E field focus and adding the
ultrasound broad focus phasing scheme, the EM and US field together achieve a
temperature distribution that covers the majority of the 3D tumor volume. The 3D
temperature distribution is calculated in the thermal model based on BHTE. This
combination heats a breast tumor surrounded by fat better than either modality
alone. Comparisons between RF/ US hybrid method and RF-Only or US-Only show
that the hybrid heating strategy can heat the whole region Of the LABC tumor better
that US or RF alone.

128

 

 

CHAPTER 8

E FIELD POLARIZATION AND CROSS ANTENNA

In most RF thermal therapy system]?, 45, 28, 2, 37], dipole antenna or dual dipole
antenna are used as the element of the phased array. In the most target heating region,
the E field radiated by the dipole antenna or dual dipole antenna is linear polarized
on the plane which is vertical to the antenna. The direction of E field polarization
is parallel to the antenna arms. In the hyperthermia treatment of breast cancer, this
kind Of E field polarization can cause the hot spots outside of the tumor[37] because
of different reflection coefficient along the tumor/normal tissue boundary. The hot
spots are located along the polarization direction and around to the tumor. One way
to eliminate the linear polarization effect is to replace the dipole antenna with the

antenna which. can radiate the circular polarized EM wave.

The polarization of an electromagnetic wave is defined as the orientation of the
electric field vector. Recall that the electric field vector is perpendicular to both
the direction of travel and the magnetic field vector. The polarization is described
by the geometric figure traced by the electric field vector upon a stationary plane
perpendicular to the direction of propagation, as the wave travels through that plane.
There are several types of antennas which can radiate circular polarization, such as
helical antenna, spiral antenna, two orthogonal conductors, etc. The helical antenna
is not compact enough to build an applicator for breast cancer treatment. The spiral
antenna is complicated to build and to compose an array for breast cancer treatment.
In order to make the applicator compact and This cross antenna is composed of
two dipole antennas perpendicular to each other and excited with 90 degree phase
difference. A M 4 phase delay line is designed for this cross antenna to decrease the

number the feed port from two to one.

129

In this paper, a new antenna, cross antenna with delay line, is designed for regional
hyperthermia. This cross antenna composed Of two dipole antenna and two delay
lines can radiate circular polarized electromagnetic wave with just one input. The
electromagnetic (EM) wave propagation along the cross antenna and the electric field
distribution in time domain is inspected with the Finite Difference in Time Domain
(FDTD) simulation. The 90 degree phase shift between two branches and the circular
polarization of E field in the target region are verified. Two heating spots are Observed
close to the tumor/ normal tissue boundary and they rotate around the tumor. The
average SAR. around and inside the tumor is more homogeneous than that by dipoles.

An applicator with four cross antenna is proposed to treat the breast tumor.

8. 1 Boundary condition

Electromagnetic boundary conditions for two different. material boundary are shown

in Eq. 8.1.

E1: = E2t

€1E1n = €2E2n (8-1)

In these two equation, E means electric field and e is the material permittivity. t

is the transverse direction and n is the normal direction.

A simple drawing in Fig. Figure 8.1 will show how E field polarization effects
the E field distribution around the tumor fat boundary. In this figure, tumor is a
circle surrounded by fat tissue and E field direction points to the top of the figure
indicated by black arrows. E field directions are vertical to the tumor/fat boundary
at location 1 and 2 , while they are parallel to the tumor/ fat boundary at location 3

and 4. According to Eq. 8.1, E field in tumor and E field in fat are equal at location

130

3 Tumor 4

fat

Figure 8.1. E field polarization and boundary condition

3 and 4. However E field in fat is much stronger (etumm/efa, z 8 times at 140MHz)
than that in tumor in location, which will cause hot spot in location 1 and 2 in fat.
The simulation results in Chapter. 3 shows there are two hot spots along the E field

direction and proves this theory.

8.2 Cross antenna and applicator

Two dipole antennas are placed perpendicular to each other with the superposing the
center of these two antennas, shown in Fig.Figure 8.2a. The antenna plane is defined
the plane containing the two dipole antennas. The central axis passes through the
common center of the two dipole and is perpendicular to the antenna plane. When
the feeding of two dipole antennas has 90 degree (or -90 degree) phase difference, the
E-field radiated by the structure is circular polarized along the central axis.

But this two dipole structure needs two signal inputs, which have 90 degree delay,
and matching circuits. To simplify the input, a phase delay line is added between two
dipole. One dipole antenna is connected to the other dipole with the delay line.The
total length of the phase delay line is /\/4 (56.5 mm @ 140MHz in water). In order to

make the antenna compact, the length of each arm of the cross antenna is /\/4. The

131

u]
9

 

design of the cross antenna is shown in Fig.Figure 8.21).

A breast model is designed here to compare the heating performance between the
cross antenna and the common dipole antenna. The cross view of the one-antenna
applicator is shown in Fig.Figure 8.3. The antenna is mounted on the large size
plastic layer (Lexan). It is air below the plastic layer and water over this layer. A
breast model is placed at the target heating region, which is along the antenna central
axis and about 15 cm far away from the antenna. The tumor is a sphere with 5 cm
diameter, the breast is a half sphere with 7.5 cm radius and the fat layer behind the

breast is about 3 cm thick.

One-antenna applicator can not reach the desired heating performance for deep
seated breast tumor duo to the decay and reflection of E field. Phased array of cross
antenna with dedicated enclosure is a better choice. The four cross antenna applicator
is shown in F ig.Figure 8.4, which has the same enclosure size as the four antenna
applicator in[37]. The tank enclosure consists of 6 solid Lexan (GE Polymerland,
North America: www.gepolymerland.com) side panels and a Lexan top piece with a
large opening. RF antennas are mounted on four of the rectangular side panels, and
the two remaining side panels are irregular pentagons with one axis of symmetry. The
four rectangular side panels are 12.8cm by 21.5cm, and the pentagonal side panels are
12.8cm by 12.8cm by 12.8cm by 12.8cm by 33.2cm. In each pentagonal side panel,
the obtuse angle measured at the lowest point on the tank is 112°, the two obtuse
angles measured at adjacent vertices are 153°, and the two remaining acute angles

are 61°. Each Lexan panel is approximately 5mm thick.

Four cross antennas are mounted the inner side of the lexan tank, Each antenna is
rotated 45 degree relate to the edges Of each panel and the arms of the cross antenna
has a 45 degree with z axis. In order to decrease the cross talk among the four
antenna, those four antenna are exact the same (RHCP or LHCP) and face to the

applicator center.

132

 

 

 

.....

    

77" Delay line

 

(b)

Figure 8.2. a) Original two dipole antenna which are perpendicular to each other, b)

Cross antenna with A/4 phase delay line

133

 

 

Fat

l\ Q.___7l-_-—Tumor
/4 ————— Breast

 

 

 

 

\\\-//
Water
/ Antenna
1 fl
Lexan
Air

Figure 8.3. A cross section of one antenna applicator with breast model

 

Figure 8.4. 3D view of the four cross antenna applicator. Each antenna is rotated 45
degree relate to the edges of each panel and mounted on the inner side of the panels.

134

 

8.3 Results

The polarization of the electromagnetic field radiated by the cross antenna needs to
be investigated. The E field distributions varying with time, important to understand
this characteristic, are simulated with the finite difference in time domain (FDTD)
method. The FDTD method is a numerical solution for Maxwell’s curl equations,
which are based upon volumetric sampling of the unknown electric field and magnetic

field within and surrounding the structure of interest over a period of time.

The time domain E field is generated by the commercial software XFDTD (Rem-
com, Inc. State College, PA). XFDTD can handle 3D model of antenna and applicator
and subdivide them into equal-space meshes with 1 mm grid size (about 0005A @
140MHz in water). The whole 3D computational domain for the four antenna appli-
cator is about 380mm x 470mm x 225mm. In these simulations, two source type are
Gaussian and sinusoid waveform used for different cases. Gaussian waveform is used
for viewing the electromagnetic wave propagation in the cross antenna and sinusoid
waveform is used for other cases, such us calculating the E field in the target heat-
ing region. Six side boundaries are assigned with the eight-layer PML boundary to

absorb the outgoing wave.
For the four antenna applicator, the E-field distribution is computed for each

antenna separately, loaded into Matlab *(Mathworks Co, Natick, MA), and then the

results are superposed. The magnitude of the total E-field is computed according to

N

Z Un(:r, y, z)I,,

n=l

|E($,y12)| = . (8-2)

 

 

where 1,, is a complex number representing the amplitude and the phase of the n-

th antenna input, and the vector I steers and focuses the E-field. In Eq. 8.2, Un

 

Readers can contact with the author for the script. used to read the output of XFDTD into
Matlab.

135

 

 

represents the electric field contribution produced by the n—th antenna for a unit
input excitation (i.e., 1,, 2 140°), N = 4 is the number of the antennas in the
phased array applicator, E represents the total electric field, and (:13, y, 2) represents

the Cartesian coordinates Of the simulated E-field.

Electromagnetic wave propagation in the cross antenna, E field polarization in
breast model and E field distribution in the four antenna applicator are investigated

by simulations.

8.3.1 Electromagnetic wave prOpagation in antenna

In order to examine the electromagnetic wave propagation along the cross antenna,
the antenna is mounted in air/plastic/ water three layer model, the antenna is on the
plastic surface in the water side and the thickness of the plastic layer is 5 mm. The
dimension of the model is about 16cm x 16cm x 10cm and the outer boundaries are
set PML with 8 layers. The mesh size is 1 mm in the direction parallel to the antenna
plane and 0.5 cm in the direction vertical to the antenna plane. A Gaussian pulse
input with 32 ps pulse width and total 1000 time steps is used as the signal input
and the time step is about 1.926 ps.

E field distribution in the antenna plane is recorded in time sequence with 10 steps
increase. From the recorded results, a pulse wave front is propagating in the antenna
arms and delay lines. Fig. Figure 8.5 shows 4 typical time steps. In Fig.Figure 8.5,
a) shows the EM wave just arrives the port of the antenna. b) shows EM wave is
propagating at the half way Of the vertical branch and the half way in the delay line.
c) shows EM wave arrives the end of the vertical branch and reaches the horizontal

branch, d) shows the EM wave reaches the end of the horizontal branch.

8.3.2 E field distribution for one antenna

The heating performance in the one antenna applicator with the breast model shown

in Fig. Figure 8.3 is compared between the cross antenna and the dual dipole antenna.

136

 

Figure 8.5. A short Gaussian pulse is an input and the E field distribution on the
antenna plane are recorded in different time steps. a) t T 0, b) t : T/8, c) t + T/4,
d) t ,—, T/2(1/T — 140 MHz).

137

The size Of the applicator, the distance from antenna to the breast model, the material
properties and the distance from applicator to the radiation boundary are keep the
same. The E-field distributions in the center plane (parallel to the antenna plane,
through the center of the tumor) are show in Fig.Figure 8.6. Fig.Figure 8.6a shows
the E field distribution by the dual dipole antenna and Fig.Figure 8.6b shows that by
the cross antenna. In Fig. Figure 8.6a, there are two E field peak region around the
tumor along the x direction and also two E field valley region along the y direction.
These two E field peak regions can cause two hot spots outside the tumor. While
the E field distribution is uniform around the tumor for the cross antenna shown in
Fig. Figure 8.6b. Fig. Figure 8.6c shows the magnitude of E field along the tumor

boundary in the center cut plane.
8.3.3 Single cross antenna applicator at 915MHz with Oil bolus

The basic thought about this single applicator is to design a cheap and easy control
hyperthermia applicator for breast cancer treatment, which also need not expensive
accessories such as shielding room. The frequency 915MHz is approved by FDA for
medicine in USA and many other countries and the shield room is unnecessary for
the applicator working at this frequency. But the 915MHz wavelength is too small in
water, which will cause unwanted standing wave and resonance in the applicator. The
mineral oil is chosen here as the media for EM wave propagation and body surface
cooling. The relative permittivity Of the mineral oil is about 3 and the wavelength in
Oil is 18.9 cm. The applicator and the breast model is shown in Fig Figure 8.7. The
enclosure of the applicator is same at that of the five antenna applicator because the
size and shape of this applicator works very well both in simulation and treatment
and prevents the unwanted resonance. The cross antenna is mounted at the center
of the bottom panel because the breast is symmetric when looking from the bottom.
The breast model is similar to the simple breast model which is similar to that in

Chapter 3 and the four cylinders behind the breast are the four ribs.

138

 

 

N —-—Cross antenna
_ —°—Dual dipole

    

_l
0|

Elmean(E)

P
on

 

A
V

O

 

 

160 150
Locatlon
(c) E magnitude along the tumor/fat boundary

Figure 8.6. Cross antenna model and E field distribution in the breast model
139

 

Figure 8.7. Single cross antenna applicator driven at 915MHz, the applicator is filled
with mineral oil, whose relative permittivity is about 3. The enclosure of the applica-
tor is same at that of the five antenna applicator and the cross antenna is mounted at
the center of the bottom panel. The breast model is the simple breast model which
is similar to that in Chapter 3 and the four cylinders behind the breast are the four
ribs.

140

The delay line of the cross antenna should be redesigned for 915MHz, since the
wavelength is different with 140MHz at water. Here for simulation convenience, two
dipole antenna with two power input are used at the circular polarization source. E
field on 3:2 plane and yz plane are shown in Fig. Figure 8.8. The E field distribution
around the tumor region in 272 plane are similar to the distribution in yz plane,
because the E field in tumor region are circular polarized and the breast model are
axial symmetric.

Fig. Figure 8.9 shows the SAR. distribution in yz and xz plane. The SAR peak is
located at the bottom of the tumor which is close to the antenna. In the breast fat
tissue, the SAR peak has a gradient along 2 axis and the peak is close to the breast
skin which can be cooled down by oil circulator. The SAR is very low in other part
Of the body. The temperature distribution is shown in Fig. Figure 8.10. It shows the
temperature peak region is inside the tumor.

The model with tumor off center axis is also simulated. The tumor is moved
-10mm along 1: axis and the other part of the patient model and the applicator are
kept the same as that in the previous simulation. The antenna are still Operated at
915MHz with the oil bolus. The temperature results are shown in Fig.Figure 8.11 and
the temperature peak region is still located inside the tumor. From the temperature
distribution in xz plane, the location of the temperature peak is also move with the

tumor along x axis about -8 mm.
8.3.4 E field distribution in four antenna applicator

The target heating region in the four antenna applicator in Fig.Figure 8.4 is at the
center of and underneath the top water surface, where the patient breast should be
located. If the amplitude of power input of each antenna is set to unity and the
starting phase is set to zero, there is an E field focus at the target heating region.
The four antenna applicator model is placed at the center Of the 60m x 60m x

70m air region with 8 layer PML. The input to four antenna is sinusoidal wave at

141

 

Figure 8.8. E field distribution in the breast model and water tank. a) shows the E
field on the xz plane with y = 0, b) shows the E field on the yz plane with x = 0;
There are strong E field close the antenna region. They also show E field decays with
increasing the penetration depth.

142

SAR in X2 plane with y = 0cm
60 80

60

40

20

 

49900 -50 o 50 100
X (mm)

SAR in Y2 plane with x = 0cm
60

    

"19900 .50 o 50 100

Figure 8.9. SAR distribution in xz and yz plane

143

T in X2 plane with y = 0cm

\_fl

 

-60 -30 0 30 60
Y(mm)

Figure 8.10. Temperature increase distribution in xz and yz plane. The peak tem-
perature is 44°C inside the tumor and the oil bolus temperature is body temperature
37°C.

144

T in X2 plane with y = 0cm

 

-80 -40 0 40 80
X(mm)

(a)

T in Y2 plane with x = 0cm

    

V/

-60 -30 0 30 60
Y(mm)

(b)

 

Figure 8.11. Temperature increase distribution for the breast model with tumor off
axis shifting -1 cm along 2: axis.

145

140 MHZ and the total length of the input 10 periods. The magnitude of the input
in four channel are the same and there are no phase delay among the four input.
The time average E field distribution in three cut planes are shown in Fig. Figure
8.12. All three views Show there is a E field focus at the center of the applicator and
the focus is about 3 cm below the water surface. In time domain, B field focus runs

around the focus region.

8.4 Discussion

8.4.1 The delay line of the cross antenna

The two delay lines are the key components in the cross antenna. The original design
of the cross antenna are just two dipole antennas cross to each other. The excitations
of these two dipole are separate with 90 degree phase shift. In order to simplify the
excitation, these two dipoles can be combined together with the delay line component.
The length of one delay line is equal to A/4(140 MHz). The geometric shape of the
delay line is U shape, which is mainly determined by the shape of the two cross
dipoles, shown in Fig. Figure 8.2. Another advantage of the shape of the delay line
is that the electric current in the two arms of the U shaped delay line are opposite
and the E field radiated by the U shaped delay line is canceled.

To verify the design of the delay line, it is necessary to investigate the detail of the
electromagnetic (EM) wave propagation in the cross antenna. A short Gaussian pulse
is an input and the E field distribution on the antenna plane is recorded in different
time steps. Fig. Figure 8.5 shows four time steps of the EM wave propagating along
the surface of the cross antenna. In Fig. Figure 8.5a, the electromagnetic starts
propagating in the vertical dipole. In Fig. Figure 8.5b, the wave front reaches the
half way of the vertical dipole and half way of the delay line. In Fig. Figure 8.5c,
the wave front reaches the end of the vertical dipole and starts propagating in the

horizontal dipole. Fig. Figure 8.5d shows the wave front reaches the end of the

146

 

Figure 8.12. E field distribution in the four cross antenna. applicator

147

horizontal dipole. The time gap between the wave front reaching the end of the
vertical dipole and the end of the horizontal dipole is T/ 4 or 90 degree (at 140 MHz).
This simulation shows the delay line component delays the input sinusoidal signal 90
degree (at 140 MHz) into the horizontal dipole.

There is cross talks existing among the two dipole arms and the arms of the delay
line, for example there is weak wave propagating in the horizontal dipole at time T/ 8
in Fig. Figure 8.5 b. Although the length of the delay line is /\/4(140 MHz), the
delay is not exact T/4 caused by the cross talk. So the E field radiated by this cross
antenna is not purely circular polarized. In Fig. Figure 8.6b, the heating by the cross
antenna is still not perfect uniform along the boundary of the tumor.

In the Gaussian pulse simulation, the antenna is mounted at the air/ water bound-
ary. There is a impedance mismatch between 509 input impedance and the radiation
impedance of the antenna. The reflection of the wave front is observed in the vertical
dipole in Fig. Figure 8.5d.

The delay line can also isolate the reflected wave from coming into the amplified
system. Since a reflection from a right hand CP wave returns as a left hand wave
and is reflected back the CP antenna, the reflected wave will generated opposite
direction current at the two arm connection and no reflected power flows back to the
generator. If two dipoles are driven by two separate power channels which have a
90" phase shift, then the reflected wave from the body surface would directly return
through each dipole as reflected power to the generator. So when deciding about
applicator choices, the cross coupling and reflection effects need to be considered for

the driving system.
8.4.2 E field distribution in breast model

In breast cancer treatment, the tumor is surrounded by fat tissue and the breast is
immersed in the deionized water. The permittivity of fat material is much less that of

tumor and water and the conductivity of the fat is also much less that of tumor. This

148

three layers structure with impedance mismatch causes difficulty to the applicator
with dipole type antenna aligned in the same direction to heat the large size tumor.
This type applicator has different E field penetrations in different directions, because
the B field radiated by dipole antenna is linear polarized in the tumor region. The
inhomogeneous penetration of B field by dipole type antenna in the breast model in
Fig. Figure 8.3 are shown in Fig. Figure 8.6a. The dipole antenna is along the 2: axis,
the B field on this cut plane is linear polarized along z axis. Around the tumor, the
E field at the region close to z axis is much stronger than that close to y axis and will
cause two hot spots at the tumor/ fat boundary. When the dipole antenna is replaced
by cross antenna in the one antenna applicator, B field in this cut plane is circular
polarized, the E field direction at any point on this plane is rotated 360 degree in one
period and the magnitude of the E field does not change in the rotation. When the
E field is circular polarized, The E field distribution and the penetration of E field

are uniform around the tumor which is shown in Fig. Figure 8.6b.
8.4.3 Single cross antenna applicator at 915MHz

The E field seems better penetration into tumor from the simulation results, although
the poor penetration depth at 915MHz was reported by the clinical application. Dif-
ferent shape, size and material of the bolus can partially explain the different between
the simulation and clinical experience. Some factors could give the reason that the
915MHz has good penetration into to deep tumor. One of them is that the permittiv-
ity difference between the tumor and fat at 915MHz is smaller than that at 140MHz
or 100MHz and it is same to the conductivity, which lows the reflection coefficient at
tumor/ fat boundary. The other one is that the low oil permittivity helps EM energy
entering the breast. But the low oil permittivity also helps EM energy leaking out of
the bolus.

The rib has no blood flow to help cooling down the temperature during the treat-

ment, so the ribs are very easy to heat up if there is amount of SAR. in rib region. The

149

decay of EM at 915MHz in the behind tumor region is higher than that at 140MHz,
because the conductivities of fat, tumor and muscle at 915MHz are higher than that
at 140MHz. So the EM wave decays to very low level when it reaches the ribs, which
is shown in Fig. Figure 8.8 and Fig. Figure 8.9.

The power efficiency of this applicator is not calculated. Given enough power
input into this applicator, it will heat the tumor ( 3-4cm deep) very well without
burning the breast skin. This applicator also showing a promising result on heating
the shifted tumor. The tumor will be heated without positioning the applicator to
align the tumor on the axis of the antenna and the tumor needs not to be at the

center of the breast, which happens to most of the breast cancer cases.
8.4.4 E field distribution in four antenna applicator

Fig. Figure 8.12 shows that simulations predict the approximate shape and location
of the focus generated by the four antenna applicator. Figs. Figure 8.12b and Figure
3.6Figure 8.12b demonstrate that the focus is in the center of the water tank and
is about 3 cm under the water surface. These figures demonstrate that the focus
generated by this RF phased array is quite broad, which suggests that this device is
appropriate for regional heating.

In time domain, the E field focus runs around the focus region and the B field
vector is circular polarized in the cross antenna applicator, while the position of the
focus is fixed and E field is linearly polarized for the applicator with dipole antenna.
This property of the cross antenna applicator is helpful to remove the hot spots and
the tumor/ fat boundary will be heated more uniformly than that with dipole antenna
applicator.

Coupling between antennas is also evident in these measurements. This coupling
is in part caused by standing waves within the water tank. The standing wave pat-
tern changes as the input phases and amplitudes are varied, which also changes the

coupling between antennas. This coupling also changes depending on the load in the

water tank. In particular, the coupling changes for different water levels and for dif-
ferent phantom materials placed in a plastic cup on the water surface. This coupling
is responsible in part for the differences between the measured and simulated phase
and amplitude settings required for a focus in the center of the tank and for a steered
focus. These effects, which are commonly encountered in RF phased array systems
designed for hyperthermia [?, 34, ?I, are reduced by focusing the breast applicator in
center of the water tank so that the reflected powers are minimized. This defines the

location where preferential focusing is achieved for this RF phased array applicator.

8.5 Conclusion

Linear polarization heating is found in the dipole type antenna applicator, which will
cause two hot spots outside the tumor along the polarization direction. A new type
antenna, cross antenna, is designed for regional hyperthermia to avoid this kind of
polarization hot spots. The electromagnetic (EM) wave propagation along the cross
antenna and the electric field distribution in time domain is inspected with the Finite
Difference in Time Domain (FDTD) simulation. The 90 degree phase shift between
two branches and the circular polarization of E field in the target region are verified.
The average SAR around and inside the tumor is more homogeneous than that by
dipoles. A new single cross antenna applicator at 915MHz is proposed to heating
breast cancer with oil bolus. B field, SAR and temperature are simulated for this
applicator with simple breast model. Another four cross antenna phased array is also
proposed to treat breast cancer at 140MHz with water bolus. The results of E field
distribution of this array applicator shows there is a E field focus at the center of
the applicator and under the water/ air surface, where is the tumor location in the

treatment.

151

CHAPTER 9

SUMMARY AND FUTURE RESEARCH

Several novel applicators, RF phased array and RF/ US hybrid applicator are simu-
lated with E field, pressure field, SAR and temperature calculation and/ or measured
with E field probe and MRI scanner. The four antenna applicator and the five an-
tenna applicator had good performance in clinical treatment. The hybrid applicators
show the best heating performance which can be obtained through physical modali-
ties. The electric field polarization effect on hyperthermia was found in the simulation
study of the RF applicator. The applicator should avoid aligning the antenna in the
same direction, which will cause linear polarization with fixed direction. The circular
polarization was first introduced into the design of the RF applicator and the cross
antenna with delay line was designed to radiate the CP wave. The one cross antenna

applicator offers good heating performance in a small breast.

The RF (or hybrid) applicator and MRI integrated system is a promising clin-
ical application or equipment for deep seated tumor hyperthermia treatment. The
MRI system can be used for imaging the target area, tumor region recognition and

temperature guidance for treatment.

Relationship between the wavelength and applicator size

The size or dimension of the applicator should be determined at first when designing
a new applicator. For non-invasive breast applicator, current majority designs are
plastic enclosure with antenna mounted on the inner surface and the enclosure is filled
with some media, such as water, mineral oil and etc. According to simulation results
in this thesis, the dimension of the applicator in any direction should be in the range
1A 2/\(this is experience value, not a mathematic solution). If the size/wavelength

ratio is much larger than 2, some parasitical standing wave would exist somewhere

152

in the applicator, which can cause skin burn. If the size/wavelength ratio is smaller

than 1, it is not easy for B field focusing and steering in the phased array applicator.
Optimization method for hybrid SAR

There are numbers of method for RF or US hyperthermia individually. But there
is no method currentlt available for RF/ US hybrid application. In this thesis, the
weights of RF and US power are optimized, while the ultrasound SAR distribution

and RF SAR distribution are optimized individually.
Close-loop control of applicator by temperature feed back

The integration of hyperthermia system and MR system still needs time to fit the
clinical requirement. The electromagnetic interference between RF applicator and
MR system is not clearly understood. The software controlling the combined system
is not sophisticated enough for clinical application.

The power inputs for all RF antennas and / or US channels currently are optimized
with the geometric shape and material properties of the patient model. But in clinical
treatment, hot spots in healthy tissue are inevitable and some part of the tumor is
not well heated. The method using temperature as optimization parameter to control

the applicator is still not well developed.
Blood flow by MR Angiography (MRA)

The blood flow strongly effects the heat diffusion in the treatment region. In the sim-
ulation, blood perfusion is a constant and homogeneous value in one tissue. Actually
the blood flow is not constant and homogeneous for one tissue and the blood flow is
high in vessel and in the high temperature region. MRA can image the distribution of
the blood vessel and the velocity of the blood flow. With the image of the blood vessel
in the treatment region, the treatment plan can give much more precise result. The
online feed back of the blood flow change offers more parameters for further heating

optimization.

153

BIBLIOGRAPHY

154

BIBLIOGRAPHY

[1] W. T. Joines, Y. Zhang, C. Li, and R. L. Jirtle, “The measured electrical prop-

[2|

[3|

[4|

[6|

[7]

[8|

[9|

[ml

[11]

erties of normal and malignant human tissues from 50 to 900 mhz,” Med. Phys,
vol. 21, no. 4, pp. 547—550, Jan 1994.

H. Kroeze, J. B. Van de Kamer, A. A. C. De Leeaw, and J. J. W. Lagendijk,
“Regional hyperthermia applicator design using FDTD modelling,” Physics in
Medicine and Biology, v01. 46, no. 7, pp. 1919-1935, 2001.

J. van der Zee, “Heating the patient: A promising approach?,” Annals of Oncol-
ogy, vol. 13, pp. 1173—1184, 2002.

B. Hildebrandt, P. Wust, O. Ahlers, and et al., “The cellular and molecular basis

of hyperthermia,” Critical Reviews in Oncology/Hematology, v01. 43, pp. 33-56,
2002.

P. Wust, B. Hildebrandt, G. Sreenivasa, and et al., “Hyperthermia in combined
treatment of cancer,” The Lancet Oncology, vol. 3, pp. 487—497, 2002.

J. L. Volakis, A. Chatterjee, and L. C.Kempel, Finite Element Method for Elec-
tromagnetics, p. 161. New York: IEEE press, 1998.

J. Jin, The Finite Element Method in Electromagnetics, p. 15. New York: John
Wiley and Sons, INC, 2002.

D. M. Sullivan, D. T. Borup, and O. P. Gandhi, “Use of the finite-difference time-
domain method in calculating EM absorption in human tissues,” IEEE Trans.
Biomed. Eng., vol. 34, no. 2, pp. 148—157, 1987.

K. D. Paulsen, X. Jia, and J. M. J. Sullivan, “Finite element computations of spe-
cific absorption rates in anatomically conforming full-body models for hyperther-
mia treatment analysis,” IEEE Trans. Biomed. Eng., vol. 40, no. 9, pp. 933—945,
1993.

S. T. Clegg, S. K. Das, E. Fuller, S. Anderson, J. Blivin, J. R. Oleson, and
T. V. Samulski, “Hyperthermia treatment planning and temperature distribution
reconstruction: A case study,” Int. J. Hyperthenn., vol. 12, no. 1, pp. 65—76, 1996.

J. C. Kumaradas and M. D. Sherar, “Edge-element based finite element analysis
of microwave hyperthermia treatments for superficial tumours on the chest wall,”
Int. J. Hypertherm, vol. 19, no. 4, pp. 414—430(17), Jul-Aug 2003.

155

I12]

[13]

[Ml

[Nil

[17]

I18l

WI

[20]

l21l

E. G. Williams and M. J. D., “Numerical evaluation of the rayleigh integral for
planar radiators using the fft,” J. Acoust. Soc. Am, vol. 72, no. 6, pp. 2020—2030,
1982.

P. T. Christopher and K. J. Parker, “New approaches to the linear propagation
of acoustic fields,” J. Acoust. Soc. Am, vol. 90, no. 1, pp. 507—521, 1991.

P. Wu and T. Stepinski, “Extension of the angular spectrum approach to curved
radiators,” J. Acoust. Soc. Am, vol. 105, no. 5, pp. 2618—2627, 1999.

P. Wu, R. Kazys, and T. Stepinski, “Analysis of the numerically implemented
angular spectrum approach based on the evaluation of two-dimensional acoustic
fields 1. errors due to the discrete fourier transform and discretization,” J. Acoust.
Soc. Am., vol. 99, no. 3, pp. 1339—1348, 1996.

P. Wu, R. Kazys, and T. Stepinski, “Optimal selection of parameters for the
angular spectrum approach to numerically evaluate acoustic fields,” J. Acoust.
Soc. Am, vol. 101, no. 1, pp. 125—134, 1997.

P. Wu, R. Kazys, and T. Stepinski, “Analysis of the numerically implemented
angular spectrum approach based on the evaluation of two-dimensional acoustic
fields 2. characteristics as a function of angular range,” J. Acoust. Soc. Am,
vol. 99, no. 3, pp. 1349—1359, 1996.

S. K. Das, S. T. Clegg, and T. V. Samulski, “Electromagnetic thermal therapy
power optimization for multiple source applicators,” Int. J. Hypertherm., vol. 15,
no. 4, pp. 291—308, 1999.

J. Wiersma, V. M. RAM, and V. D. J DP, “A flexible optimization tool for hyper-
thermia treatments with rf phased array systems,” Int. J. Hypertherm, vol. 18,
no. 2, pp. 73—85, Mar 2002.

G. C. Van Rhoon, D. J. Van der Heuvel, A. Ameziane, P. J. M. Rietveld,
K. Volenec, and J. Van der Zee, “Characterization of the sar-distribution of the
sigma-60 applicator for regional hyperthermia using a schottky diode sheet,” Int.
J. Hypertherm., vol. 19, no. 6, pp. 642—654, Nov-Dec 2003.

S. K. Das, S. T. Clegg, and T. V. Samulski, “Computational techniques for fast
hyperthermia temperature optimization,” Med. Phys, vol. 26, no. 2, pp. 319—328,
1999.

156

l22|

|23l

[MI

[25]

[26]

[27]

WI

[29]

WI

[31]

K. Hynynen, O. Pomeroy, D. N. Smith, P. E. Huber, N. J. McDannold, J. Ket-
tenbach, J. Baum, S. Singer, and F. A. Jolesz, “Mr imaging-guided focused ul-
trasound surgery of fibroadenomas in the breast: A feasibility study,” Radiology,
vol. 219, no. 1, pp. 176-185, 2001.

Y. Zhang, W. T. Joines, R. L. Jirtle, and T. V. Samulski, “Theoretical and mea-
sured electric field distributions within an annular phased array: Consideration
of source antennas,” IEEE Trans. Biomed. Eng, vol. 40, no. 8, pp. 780—787,
1993.

Y. Fujita, H. Kato, and T. Ishida, “An rf concentrating method using inductive
aperture-type applicators,” IEEE Trans. Biomed. Eng, vol. 40, no. 1, pp. 110—
113, JAN 1993.

H. Kato and T. Ishida, “Present and future-status of noninvasive selective deep
heating using rf in hyperthermia,” Medical and Biological Engineering and Com-
puting, vol. 31, no. S, pp. 2—11, Jul 1993.

C. Gromoll, U. Lamprecht, T. Hehr, M. Buchgeister, and M. Bamberg, “An on-
line phase measurement system for quality assurance of the bsd 2000. part i:

technical description of the measurement system,” Int. J. Hypertherm, vol. 16,
no. 4, pp. 365—373(9), July 2000.

M. Malinen, T. Huttunen, K. Hynynen, and J. P. Kaipio, “Simulation study
for thermal dose optimization in ultrasound surgery of the breast,” Med. Phys,
vol. 31, no. 5, pp. 1296-1307, 2004.

P. Wust, J. Nadobny, R. Felix, P. Deuflhard, A. Louis, and W. John, “Strategies
for optimized application of annular-phased-array systems in clinical hyperther-
mia,” Int. J. Hypertherm., vol. 7, no. 1, pp. 157—173, 1991.

P. F. Turner, “Mini-annular phased-array for limb hyperthermia,” IEEE Nans-
actions on Microwave Theory and Techniques, vol. MTT-34, no. 5, pp. 508-513,
1986.

A. J. Fenn, G. L. Wolf, and R. M. Fogle, “An adaptive microwave phased array
for targeted heating of deep tumours in intact breast: animal study results,” Int.
J. Hypertherm., vol. 15, no. 1, pp. 45-61, 1999.

P.F.Turner, “Regional hyperthermia with an annular phased array,” IEEE Trans.
Biomed. Eng, vol. 31, no. 1, pp. 106-14, Jan 1984.

157

[32] K. D. Paulsen, J. W. Strohbehn, S. C. Hill, D. R. Lynch, and F. E. Kennedy,

[33]

I34]

I35]

I36]

WI

[38]

[39]

[40]

MN

“Theoretical temperature profiles for concentric coil induction heating devices in
a two-dimensional, axi-symmetric, inhomogeneous patient model,” Int. J. Radiat.
Oncol. Biol. Phys, vol. 10, pp. 1095—1107, 1984.

K. D. Paulsen, J. W. Strohbehn, and D. R. Lynch, “Theoretical electric field
distributions produced by three types of regional hyperthermia devices in a three-

dimensional homogeneous model of man,” IEEE Trans. Biomed. Eng, vol. 35,
no.1,1988.

S. N. Hornsleth, L. Frydendal, O. Mella, O. Dahl, and P. Raskmark, “Quality
assurance for radiofrequency regional hyperthermia,” Int. J. Hypertherm, vol. 13,
no. 2, pp. 169-185, Mar-Apr 1997.

D. M. Sullivan, “Mathematical methods for treatment planning in deep regional
hyperthermia,” IEEE Trans. Microwave Theory Tech, vol. 39, no. 5, pp. 864—
872, 1992.

D. L. Carter, J. R. MacFall, S. T. Clegg, X. Wan, D. M. Prescott, H. C. Charles,
and T. V. Samulski, “Magnetic resonance thermometry during hyperthermia for
human high-grade sarcoma,” Int. J. Radiat. Oncol. Biol. Phys, vol. 40, no. 4,
pp. 815—822, 1998.

L. Wu, R. J. McGough, O. A. Arabe, and T. V. Samulski, “An rf phased ar-
ray applicator designed for hyperthermia breast cancer treatments,” Phys. Med.
Biol, vol. 51, no. 2006, pp. 1—20, Jan 2006.

L. Wu and R. J. McGough, “The effect of e—field polarization on hyperthermia
breast cancer treatments,” the 2006 Annual Meeting of the Society for Thermal
Medicine, Apr 2006.

R. J. McGough, J. F. Kelly, and T. V. Samulski, “An efficient grid sectoring
method for calculations of the nearfield pressure generated by a circular piston,”
J. Acoust. Soc. Am, vol. 115, no. 5, pp. 1942-1954, 2004.

P. R. Stepanishen and K. C. Benjamin, “Forward and backward projection of
acoustic fields using FFT methods,” J. Acoust. Soc. Am, vol. 71, no. 4, pp. 803—
812, 1982.

G. T. Clement and K. Hynynen, “Field characterization of therapeutic ultrasound
phased arrays through forward and backward planar projection,” J. Acoust. Soc.
Am, vol. 108, no. 1, pp. 441—446, 2000.

158

 

I42]

I43]

I44]

I45l

I46]

I47]

I48]

I49]

I50]

I51]

[52]

R. J. McGough, H. Wang, E. S. Ebbini, and C. A. Cain, “Mode scanning: heating
pattern synthesis with ultrasound phased arrays,” Int. J. Hypertherm, vol. 10,
no. 3, pp. 433—42, May-Jun 1994.

E. S. Ebbini, “Deep localized hyperthermia with ultrasound phased arrays using
the pseudoinverse pattern synthesis method,” Thesis for PHD degree in Electrical
Engineering in the Graduate College of UIUC, 1990.

R. J. McGough, E. S. Ebbini, and C. A. Cain, “Direct computation of ultra-
sound phased-array driving signals from a specified temperature distribution for
hyperthermia,” IEEE Trans. Biomed. Eng, vol. 39, no. 8, pp. 825—835, 1992.

K. D. Paulsen, J. W. Strohbehn, and D. R. Lynch, “Theoretical temperature
distributions produced by an annular phased array-type system in CT-based
patient models,” Radiation Research, vol. 100, pp. 536—552, 1984.

K. S. Nikita, N. G. Maratos, and N. K. Uzunoglu, “Optimal steady-state tem-

perature distribution for a phased array hyperthermia systemn,” IEEE Trans.
Biomed. Eng, vol. 40, p. 1299"C1306, 1993.

D. S. Sullivan, R. Ben-Yosef, and D. S. Kapp, “Stanford 3d hyperthermia treat-
ment planning system. technical review and clinical summary,” Int. J. Hyper-
therm., vol. 9, p. 627"C643, 1993.

S. K. Das, S. T. Clegg, and T. V. Samulski, “Computational techniques for fast
hyperthermia temperature optimization,” Med. Phys, vol. 26, no. 2, pp. 319—328,
Feb 1999.

F. Bardati, A. Borrani, A. Gerardino, and G. Lovisolo, “Sar optimization in a
phased-array radiofrequency hyperthermia system,” IEEE Trans. Biomed. Eng,
vol. 42, no. 12, pp. 1201—1207, DEC 1995.

P. Wust, M. Seebass, J. Nadobny, P. Deuflhard, G. Monich, and R. Felix, “Simu-
lation studies promote technological development of radiofrequency phased array
hyperthermia,” Int. J. Hypertherm, vol. 12, no. 4, pp. 477—494, 1996.

C. Cain and S. Umemura, “Concentric-ring and sector vortex phased array ap—
plicators for ultrasound hyperthermia,” IEEE Trans. Microwave Theory Tech,
vol. 34, no. 5, pp. 542—551, 1986.

S. Umemura and C. Cain, “The sector-vortex phased array: acoustic field syn-
thesis for hyperthermia,” IEEE Trans. Ultrason. Ferroelect. Freq. Contr., vol. 36,
no. 2, pp. 249-257, 1989.

159

[53] K. Hynynen, G. T. Clement, N. McDannold, N. Vykhodtseva, R. King, P. J.

[5

[5

[5

4]

6]

White, S. Vitek, and F. A. Jolesz, “500-element ultrasound phased array system
for noninvasive focal surgery of the brain: A preliminary rabbit study with ex
vivo human skulls,” Magn. Reson. Med, vol. 52, no. 1, pp. 100—107, Jun 2004.

D. Chorman and R. J. McGough, “Development of prototype phased array ultra—
sound system for hyperthermia and targeted drug delivery,” Proceedings of the
acoustics week in Cannada 2005, vol. 33, no. 3, pp. 88-90, Sep 2005.

R. J. McGough, T. V. Samulski, and J. F. Kelly, “An efficient grid sectoring
method for calculations of the nearfield pressure generated by a circular piston,”
Journal of Acoustic Society of America, vol. 115, no. 5, pp. 1942—54, May 2004.

J. Wang and O. Fujiwara, “thd computation of temperature rise in the human
head for portable telephones,” IEEE Trans Microwave Theory Tech., vol. 47,
no. 8, pp. 1528—34, Aug 1999.

160

   

ullgagging