Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
High frequency ultrasound array for ultrasound-guided breast biopsy
(USC Thesis Other)
High frequency ultrasound array for ultrasound-guided breast biopsy
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
1
UNIVERSITY OF SOUTHERN CALIFORNIA
VITERBI SCHOOL OF ENGINEERING
DEPARTMENT OF BIOMEDICAL ENGINEERING
HIGH FREQUENCY ULTRASOUND ARRAY FOR ULTRASOUND-GUIDED BREAST
BIOPSY
Thomas Cummins
PhD Dissertation
May 2016
Committee Members:
K. Kirk Shung, Ph.D. (Chair)
Jesse Yen, Ph.D.
Sue E. Martin, M.D, Ph.D.
2
Dedication
To my parents
Pam and Chris Cummins
3
Acknowledgments
I would like to thank the National Institutes of Health (NIH) for its support in funding the
work presented in this dissertation. I am fortunate to have met and worked with so many talented,
dedicated and caring people during my research at USC. I would especially like to thank Dr. Kirk
Shung for the opportunity to work under his guidance in the UTRC lab. Your encouragement to
explore new ideas throughout my time in the lab allowed me to learn a tremendous amount and
I’ll always be grateful for that. Your thoughtfulness and wisdom is something I’ll carry with me
as I move forward with my career.
I would like to express my gratitude to Dr. Mary Yamashita, Dr. Linda Larsen and Dr. Sue
Martin for their constant guidance and support of my work. It was a pleasure to work with such
intelligent and caring people and I know I am lucky to have collaborated with you all. USC and
all your patients are very fortunate to have you. I am very thankful for Dr. Yen’s support of my
work and his guidance during the preparation of this dissertation. Thank you to all the UTRC lab
members, especially Nestor, Changhan, Ruimin, Jay, Payam, Bong Jin, Takahiro, Teng and
Changyang for all your help. I could not have asked for a better group of colleagues to work with.
Finally, I’d like to thank my parents, Pam and Chris, my brother Ryan, and my sisters Natalie
and Olivia for their unconditional love and support in pursuing this degree.
4
Table of Contents
Introduction to Ultrasound Imaging and High Frequency Applications .......... 16
1.1 Background of Ultrasound Imaging .............................................................................. 16
1.2 Ultrasound Transducers ................................................................................................ 19
1.3 High Frequency Ultrasound Transducers ...................................................................... 25
Clinical Need Identification .................................................................................... 32
2.1 Ultrasound Guided Breast Biopsy ................................................................................. 32
2.2 Defining the Clinical Need ........................................................................................... 35
Imaging Probe Design ............................................................................................. 39
3.1 Array Transducer Design .............................................................................................. 39
Array Simulation Modeling.................................................................................... 47
4.1 Array Simulation Approach .......................................................................................... 48
4.2 Array Simulation Methods ............................................................................................ 48
4.2.1 PiezoCAD .............................................................................................................................. 48
4.2.2 Field II .................................................................................................................................... 52
4.2.3 PZFlex .................................................................................................................................... 55
4.2.4 Final Array Design ................................................................................................................. 59
Fabrication and Packaging ..................................................................................... 60
5.1 Electrical Interconnect Transmission Line .................................................................... 64
5.2 Array Assembly Fabrication ......................................................................................... 70
5.2.1 Conductive Microsphere Electrical Interconnect Solution ..................................................... 79
Array Testing and Performance ............................................................................ 83
6.1 Array Testing Experimental Design.............................................................................. 83
6.2 Array Imaging Experimental Design ............................................................................ 85
6.2.1 Synthetic Aperture Imaging System Software ....................................................................... 89
6.2.2 GUI User Inputs and Controls................................................................................................ 90
6.3 Array Testing Results.................................................................................................... 93
5
Clinical Study Design and Results ....................................................................... 101
7.1 Clinical Study Experimental Design ........................................................................... 101
7.2 Image Processing Methods ......................................................................................... 107
7.2.1 Nakagami Filtering Methods................................................................................................ 108
7.2.2 Backscatter Analysis Methods ............................................................................................. 109
7.2.3 Quantitative Ultrasound Analysis ........................................................................................ 109
7.2.4 Quantitative Backscatter Analysis Methods......................................................................... 110
7.2.5 Statistical Analysis of Backscatter Coefficients ................................................................... 113
7.3 Clinical Study Results ................................................................................................. 115
7.3.1 Nakagami Filtering Imaging Results.................................................................................... 119
7.3.2 Backscatter Analysis Results ............................................................................................... 123
7.4 Statistical Analysis Results ......................................................................................... 135
Summary and Future Work ................................................................................. 142
8.1 Array Development Summary .................................................................................... 142
8.2 Clinical Study Summary ............................................................................................. 143
8.3 Future Array Development ......................................................................................... 143
8.3.1 Electrical Interconnect Challenges and Proposed Solutions ................................................ 144
8.3.2 New Electrical Interconnect Design – Cavity Backing Design ............................................ 144
8.3.3 High Density Interposer Vias and Pad Connections ............................................................ 148
8.4 Clinical Study Future Directions ................................................................................. 149
Bibliography .......................................................................................................... 150
6
List of Tables
Table 1: Elastic, dielectric and piezoelectric crystal properties of [001] poled
PMN-32%PT (Type-B) single crystal ................................................................................................ 23
Table 2 : Material properties of <001> poled PMN-PT single crystal ....................................................... 23
Table 3: Matching layer material properties ............................................................................................... 49
Table 4: Backing layer material properties ................................................................................................. 50
Table 5: Array simulation parameters ......................................................................................................... 51
Table 6: PiezoCAD array element simulation results ................................................................................. 51
Table 7: Field II Simulation Transmit Parameters used for Transmit Beam Profile Simulation. ............... 53
Table 8: Simulation results compiled from PiezoCAD and Field II simulations. ....................................... 54
Table 9: Final array design acoustic stack and element geometry .............................................................. 59
Table 10: APEX glass transmission line substrate material properties. ...................................................... 66
Table 11: Transmission Line Fabrication Steps .......................................................................................... 68
Table 12: Technical Specifications for MUX. ............................................................................................ 69
Table 13: Acoustic Stack Properties ........................................................................................................... 77
Table 14: Synthetic Aperture Scan Settings ............................................................................................... 91
Table 15: Average Measured Properties of 64-Element Array ................................................................... 94
Table 16: Electrical Interconnect Design Parameters ............................................................................... 145
7
List of Figures
Figure 1-1: Single element mechanical sector scanner and linear array scanner used
for hand-held, percutaneous ultrasound imaging probes. ................................................................... 17
Figure 1-2: Piezoelectric electromechanical coupling coefficient depends on material
geometry. Three common geometries are: (a) tall elements commonly used in
1-3 composites, (b) tall, long posts commonly used in 2-2 composites and
linear arrays and (c) a circular disc used in single-element transducers. ............................................ 20
Figure 1-3: Linear array element orientation geometry corresponding with k33'. ..................................... 21
Figure 1-4: Analog receive beamformer diagram ....................................................................................... 26
Figure 1-5: Digital receive beamformer diagram........................................................................................ 26
Figure 1-6: Ultrasound array technology progression with dice-and-fill, laser
micromachining and DRIE micromachining fabrication methods. .................................................... 30
Figure 3-1: Rendering of core biopsy needle (a) oriented in the in-plane technique
within the imaging plane of the external guidance linear ultrasound array
and (b) zoomed view of array with its imaging plane oriented parallel to
that of the guidance probe. .................................................................................................................. 40
Figure 3-2: Rendering of imaging array integrated within large core breast biopsy
needle with (a) sampling aperture open and (b) sampling aperture closed. ........................................ 41
Figure 3-3: (a) 1-3 AND (b) 2-2 composite structure. ................................................................................ 44
8
Figure 4-1: PiezoCAD simulation result plots for 2-way pulse/echo transmit for single array element. ... 51
Figure 4-2: Field II beam profile 2D plot as well as axial and lateral beam profile plots. .......................... 54
Figure 4-3: Grating Lobe angle and magnitude mesh plots. ....................................................................... 55
Figure 4-4: PZFlex simulation result plots for pulse/echo experiment as well as
electrical impedance magnitude and phase plots for individual element. ........................................... 58
Figure 4-5: Exploded view of array acoustic stack including backing, 2-2 composite
and parylene matching layer as well as zoomed view of 3 separate layers and
the final assembled array unit. ............................................................................................................ 59
Figure 5-1: Transmission line (a) outline and (b) zoomed figure of transmission
opening for array aperture. .................................................................................................................. 63
Figure 5-2: Image of (a) single microstrip transmission line laser cut from wafer with
copper traces shown on top layer. Array bonding pads shown around rectangular
cutout (b) match array element bonding pad pattern. The PCB (c) with individual
transformer circuits for electrical impedance matching for array channels
connects to the transmission line via a high density interposer board. ............................................... 65
Figure 5-3: Transmission line (a) outline and (b) zoomed figure of transmission
opening for array aperture. .................................................................................................................. 66
Figure 5-4: HFFS results for cross-talk measurement of transmission line channels
with a color map showing magnitude of signal in each of 5 adjacent channels.................................. 67
Figure 5-5: Image of (b) single transmission line laser cut from (a) wafer with copper
9
traces shown on top layer. Array bonding pads shown around (c) opening for
array aperture. ..................................................................................................................................... 68
Figure 5-6: NI MUX provides independent switching between transmit and receive channels. ................ 70
Figure 5-7: Array composite DRIE artwork for element, sub-elements and surrounding 1-3
composite kerf etching patterns. Artwork for (a) all 49 elements with 2-2 composite
pattern nested within 1-3 composite pattern to prevent crystal fracture propagation
in the sample. Artwork for (b) single array with 64-element aperture size of
1.0 mm x 1.274 mm. Artwork for (c) 3 distinct individual sub-element
lengths for each array element to prevent electrode strain that could
result in open channel circuits. ............................................................................................................ 73
Figure 5-8: Scanning Electron Microscope image of PMN-PT Material directly after
DRIE Processing with sub-element pitch length of (a) 50 µM, (b) 100 µM and
(c) 200 µM. ......................................................................................................................................... 74
Figure 5-9: Micrograph of array element electrodes with bonding pads. Element
bonding pads are patterned above non-conductive epoxy to avoid exciting
the piezoelectric material outside of the element. ............................................................................... 75
Figure 5-10: Finished composite samples a) Have a common ground electrode
sputtered and b) Individual elements, Sub-elements and 1-3 composite
regions can be seen from the backside of the composites. .................................................................. 76
Figure 5-11: Individual arrays diced from (a) grid of 49 arrays to separate them
10
(b), yielding individual arrays (c) 1.5 mm x 1.5 mm in size. .............................................................. 77
Figure 5-12: Array fabrication steps for array composite beginning with bulk
single crystal PMN-PT material and ending with individual
miniaturized arrays. ............................................................................................................................. 78
Figure 5-13: Image of composite film with visible bubbles forming between
silicon carrier wafer and composite material. ..................................................................................... 79
Figure 5-14: Cross-sectional view of interconnect scheme between array and glass
transmission line. Silver-coated glass spheres bonded by conductive epoxy
were used to connect bonding pads on both the array module and
transmission line. ................................................................................................................................ 82
Figure 6-1: Wire targeting imaging experimental setup. ............................................................................ 85
Figure 6-2: Synthetic aperture imaging system design for high frequency array imaging. ........................ 86
Figure 6-3: Diagram of path length between transmit and receive element for synthetic
aperture imaging. ................................................................................................................................ 88
Figure 6-4: Graphical user interface of synthetic aperture imaging system which
allows users to determine the type of array and electronic scanning method
used during the imaging process. ........................................................................................................ 90
Figure 6-5: Synthetic aperture imaging system software control block diagram. ...................................... 92
Figure 6-6: Measured electrical impedance magnitude (solid line) and phase
angle (dashed line) for a typical array element. .................................................................................. 94
11
Figure 6-7: Pulse/Echo measured results for a typical array element
including the (a) echo and (b) FFT magnitude plot. ........................................................................... 95
Figure 6-8: Center frequency and bandwidth values for each element in array. ........................................ 96
Figure 6-9: Impedance magnitude and phase angle values for each element in array. ............................... 97
Figure 6-10: Measured crosstalk values for 3 nearest neighboring elements in array. ............................... 98
Figure 6-11: Reconstructed synthetic aperture image of a single 20 µm wire target
with no thresholding or apodization. The image was displayed with 25 dB
dynamic range and mapped on a linear gray scale. ............................................................................. 99
Figure 6-12: Axial (a) and lateral (b) line plots for the center of the wire phantom. ................................ 100
Figure 7-1: Outline of clinical study to examine efficacy of high frequency
ultrasound imaging in identifying cancer in breast biopsy core samples. ......................................... 102
Figure 7-2: Ultrasound image data acquisition and backscatter analysis design
for high frequency ultrasound imaging of ex vivo breast core biopsy
tissue specimens using a single element transducer and motorized
scanning imaging system. ................................................................................................................. 104
Figure 7-3: Biopsy specimens were secured in an agar gel block to maintain
their orientation during ultrasound and pathological imaging. ......................................................... 105
Figure 7-4: Ultrasound B-Mode images from 3 consecutive image slices captured
at 100 µm intervals. The circled feature is a highly echogenic specular
reflector with acoustic shadowing directly below, which is consistent
12
with a microcalcification. .................................................................................................................. 116
Figure 7-5: H&E stained digital pathology images of biopsy core tissue
sectioned at intervals of 50 µm. ........................................................................................................ 117
Figure 7-6: Comparison between (a) histological and (b) ultrasound images
of an ex-vivo breast core biopsy tissue sample. ................................................................................ 117
Figure 7-7: Ultrasound image sets of (a) B-mode image slices captured at intervals
of 100 µm were compiled together to form a (b) 3D ultrasound rendered
image of the biopsy specimen using MATLAB. .............................................................................. 118
Figure 7-8: (a) Ultrasound B-Mode Image with two regions (A & B) identified for
parameter estimation using the Nakagami filtering technique along with their
respective (c) gray scale histograms. The m-parameter plot (b) shows that the
left and right halves of this biopsy specimen are constituted of fat on the left
and tumor. ......................................................................................................................................... 120
Figure 7-9: Pathology image with m-parameter and Ω-parameter maps of ultrasound
image as well as the original B-Mode ultrasound image. ................................................................. 121
Figure 7-10: Pathology image (a) with regions highlighted by pathologists to indicate
microcalcifications (red) and necrotic tissue (green). These region outlines were
then matched to the original b-mode image (b). Ω-parameter (c) and m-parameter
(d) maps of ultrasound image plotted over original ultrasound image. These
parameter maps had a threshold applied to them so only regions of the
13
highest parameter magnitude were displayed. .................................................................................. 122
Figure 7-11: Case 1 image set includes high frequency ultrasound a) B-mode
image followed by the backscatter parameter plots for b) IB, c) slope
and d) y-intercept and finally the e) H&E stained digital pathology
image. ................................................................................................................................................ 125
Figure 7-12: Case 2 image set include images for the 2 parts of the biopsy specimen.
The image set for parts 1 and 2, respectively, includes a high frequency
ultrasound B-mode image (a,f) followed by the backscatter parameter
plots for IB (b,g), slope (c,h) and y-intercept (d,i) and finally the H&E
stained digital pathology image (e,j). Microcalcifications, fibrous
tissue and adipose tissue are indicated by solid white arrows,
solid black arrows and cross-hatched arrows, respectively............................................................... 128
Figure 7-13: Case 3 image set includes high frequency ultrasound a) B-mode
image followed by the backscatter parameter plots for b) IB, c) slope
and d) y-intercept and finally the e) H&E stained digital pathology
image. The pathology image shows the biopsy specimen composed
of adenocarcinoma. ........................................................................................................................... 131
Figure 7-14: Case 4 image set includes high frequency ultrasound a) B-mode
image followed by the backscatter parameter plots for b) IB, c) slope
and d) y-intercept and finally the e) H&E stained digital pathology
14
image. The white arrows with black outline indicate the positions
of microcalcifications. ....................................................................................................................... 133
Figure 7-15: Box plots for adipose and adenocarcinoma tissue for IB (dB), slope
(dB/MHz) and y-intercept (dB) show the median value, given by the band
in the middle of the box. The upper and lower quartiles are given by the
upper and lower borders of the box, respectively. The 4 outliers outside
of the box plot whiskers are represented by circles. ......................................................................... 135
Figure 7-16: Box plots for microcalcifications as well as fibrous and adipose
tissue for IB (dB), slope (dB/MHz) and y-intercept (dB) show the
median value, given by the band in the middle of the box. The
upper and lower quartiles are given by the upper and lower
borders of the box, respectively. The 4 outliers outside of
the box plot whiskers are represented by circles. .............................................................................. 137
Figure 7-17: Box plots for IB (dB), slope (dB/MHz) and y-intercept (dB) of
adenocarcinoma tissue show the median value, given by the band in
the middle of the box. The upper and lower quartiles are given by
the upper and lower borders of the box, respectively. The outliers
outside of the box plot whiskers are represented by circles. ............................................................. 138
Figure 7-18: Box plots for microcalcifications and surrounding fibrous tissue
for IB (dB), slope (dB/MHz) and y-intercept (dB) show the median
15
value, given by the band in the middle of the box. The upper and
lower quartiles are given by the upper and lower borders of the
box, respectively. The 4 outliers outside of the box plot
whiskers are represented by circles. .................................................................................................. 138
Figure 8-1: Cavity backing electrical interconnect design showing the (a) assembled
unit composed of the flex circuit transmission line, high density interposer and
array composite as well as the (b) exploded view of this assembly. Dimensions
shown are for a comparable 60 – 80 MHz, 64 element linear array. ................................................ 145
Figure 8-2: Array electrical interconnect solution fabrication and assembly process. ............................. 147
Figure 8-3: Diagram of the high density interposer with (a) detailed view of
conductive via layout and (b) overview of interposer. ...................................................................... 147
16
Introduction to Ultrasound Imaging and High Frequency Applications
1.1 Background of Ultrasound Imaging
Medical ultrasound imaging is one of the most prevalent clinical imaging modalities in use
because of its relative low cost and ability to provide real-time images of structure and function of
the human body without the use of ionizing radiation. “Ultrasound” refers to acoustic wave
oscillating at a frequency above the audible range, which is approximately 20 x 10
3
cycles per
second (20 kHz), in humans (Shung 2015). Ultrasound imaging is based on the same principle as
echolocation, which is the process of mapping object location by emitting and receiving acoustic
energy. First observed in nature with dolphins and bats, echolocation found use in military non-
destructive testing before being used for medical applications such as diagnostic imaging (Blitz
and Simpson 1996; D'Amico and Pittenger 2009). The first diagnostic ultrasound images were
produced with mechanical scanning single element transducer systems by John Wild and John
Reid in the early 1950’s. Wild and Reid were the first to demonstrate how acoustic echo signals
could differentiate malignant from benign tissue in the breast and established two paths of medical
ultrasound: diagnostic imaging and tissue characterization (Seimens 2014; Szabo 2004; Wild
1950).
Medical ultrasound was developed based on pulse/echo imaging, which involves the
transmission of acoustic pulses from an ultrasound transducer into the tissue and receiving acoustic
echoes. Received acoustic energy originates from both reflected acoustic pulses off specular
reflector targets and scattered echoes off small structures within the tissue. The reflected acoustic
energy is converted to electrical signals by transducers. Scanning a transducer and mapping the
echo signal amplitude to pixel brightness in an imaging screen creates a 2-D brightness-mode (B-
17
mode) image where the depth is oriented in the vertical direction and horizontal position is
segmented by each acquisition line (A-line).
Figure 1-1: Single element mechanical sector scanner and linear array scanner used for hand-held,
percutaneous ultrasound imaging probes.
Ultrasound provides several advantages over other medical imaging modalities including the
most commonly used: x-ray, x-ray computed tomography, magnetic resonance and optical
imaging. In x-ray and x-ray computed tomography, ionizing x-ray electromagnetic radiation is
projected through the body and received by either specialized film or solid state detectors.
Differences in x-ray energy attenuation provides contrast in both these modalities although in
traditional x-ray imaging a projection image is formed while in x-ray computed tomography
imaging a tomogram is formed around a single axis of rotation (Hasegawa 1990). Magnetic
resonance imaging utilizes the phenomenon of magnetic resonance of water molecules in the body
to generate a computed tomographic image of the body (Herman 2009). Ultrasound is attractive
as a clinical imaging modality since it is non-ionizing, relatively less expensive and portable within
a healthcare facility. Ultrasound imaging has found major use in guiding interventional procedures
18
such as tissue biopsy for cancer diagnosis. Tissue visualization is an ever-present challenge and in
an effort to improve image resolution, research groups and companies have continuously sought
to raise imaging frequencies and also miniaturize ultrasound imaging single-element transducers
and array form factors in order to get the imaging aperture to the location required to provide some
diagnostic value or in some cases directing therapy such as for ultrasound-guided arterial stent
placement (Deaner, Cubukcu, Rees 1992).
The progression of image quality is generally driven by improving resolution, contrast and
frame rate. Better resolution enables visualization of smaller structures, increased contrast enables
better tissue type differentiation and faster frame rate allows images to be displayed in real time
(i.e. above 20 frames per second). During pulse/echo imaging acoustic energy is reflected at the
boundary between two materials that have different acoustic impedances and off of individual
objects when the wavelength of acoustic energy is much smaller than the object. Acoustic energy
can also be redistributed through the process of scattering, which occurs when the wavelength is
comparable to or greater than the object. Since the speed of sound in tissue is relatively constant,
measuring the time a pulse takes to travel into the body and back toward the transducer and
determining the distance of its origin to map out the differing layers of acoustic impedance.
Collecting and processing the echo signals from a transducer oriented towards the body in a fixed
position is an A-line, and collecting a series of A-lines at fixed distance intervals forms a B-mode
image. This image is interpreted as a depth image of tissue, which is the basis of diagnostic
ultrasound imaging.
19
1.2 Ultrasound Transducers
The most efficient mode of ultrasound energy transduction used in medical applications
utilizes the piezoelectric principle and was demonstrated by Curie and Curie in 1880 (Curie and
Curie 1880). The piezoelectric effect can be observed in a material where applied mechanical stress
results in a change in the electrical field and vice versa. The case in which an applied stress results
in an induced electric charge separation is the direct piezoelectric effect while an applied electric
charge induces a mechanical strain is the reverse piezoelectric effect (Cady 1964). An ultrasonic
transducer utilizes the piezoelectric effect by housing a piece of piezoelectric material in a fixed
support structure with electrodes attached to opposite surfaces of the material. A transducer emits
an acoustic pulse when a high voltage potential is applied across the material, causing it to strain.
The resulting strain motion creates a pressure wave and this wave is transmitted into the imaging
medium, which is tissue in the case of diagnostic ultrasound imaging. Similarly, when an
ultrasound echo travels back to the transducer, the acoustic wave applies pressure to the
piezoelectric surface, causing a strain in the material. This strain induces a charge in the electrodes
of the transducer and the voltage across them is recorded. This voltage is the signal recorded by
the imaging system and through a process of amplification, filtering and envelope detection of the
signal amplitude, an image can be formed.
Piezoelectric materials typically have a crystalline structure and the elastic, dielectric and
piezoelectric coefficients are used to describe their behavior. Since these crystals are symmetric, a
simplified notation is used to describe the materials. A piezoelectric plate with its surface cut
parallel to the x-plane is called an x-cut. The x-, y- and z- directions of the crystal are referred to
as 1, 2 and 3, respectively. The piezoelectric strain constant, 𝑑 33
refers to the strain produced in 3-
direction of the material when an electric field is also applied in the 3-direction. The piezoelectric
20
properties depend on the boundary conditions of the material; this value changes as the shape of
the transducer varies. Since the main purpose of ultrasonic transducers is transduction of electrical
energy into mechanical energy and vice versa, the efficiency of energy transduction is a critical
factor in the performance of these materials. This efficiency is measured by the electromechanical
coupling coefficient, k, and has a maximum value of 1.0, which would indicate 100% of the energy
is converted. The value of k for materials depends on the transducer geometry and can be
categorized into 3 types:𝑘 33
𝑘 33
′
and 𝑘 𝑡 . The coefficients 𝑘 33
and 𝑘 33
′
denote the electromechanical
coupling efficiency for tall elements with square and long rectangular cross-sections, respectively
that are much smaller than the thickness. The coefficient 𝑘 𝑡 denotes the efficiency for thin circular
disc elements.
Figure 1-2: Piezoelectric electromechanical coupling coefficient depends on material geometry. Three
common geometries are: (a) tall elements commonly used in 1-3 composites, (b) tall, long posts commonly used
in 2-2 composites and linear arrays and (c) a circular disc used in single-element transducers.
For our application the 𝑘 33
′
coupling coefficient for the geometry shown in Figure 1-2 and the
piezoelectric, elastic stiffness and dielectric constants were provided by the single-crystal
piezoelectric material manufacturer.
21
Figure 1-3: Linear array element orientation geometry corresponding with 𝒌 𝟑𝟑
′
.
The 2-2 composite pillar geometry is applicable for this array design and its electromechanical
coupling coefficient is described by:
𝒌 𝟑𝟑
′
=
√
(𝒆 𝟑𝟑
′
)
𝟐 𝑪 𝟑𝟑
𝑬 ′
𝜺 𝟑𝟑
𝑺 ′
(𝟏 +
(𝒆 𝟑𝟑
′
)
𝟐 𝑪 𝟑𝟑
𝑬 ′
𝜺 𝟑𝟑
𝑺 ′
)
(1-1)
where the piezoelectric stress constant, 𝑒 33
′
, is defined by:
𝒆 𝟑𝟑
′
=𝒆 𝟑𝟑
−
𝒆 𝟑𝟏
𝑪 𝟏𝟑
𝑬 𝑪 𝟏𝟏
𝑬
(1-2)
𝑒 33
′
=32.3 (1-3)
The elastic stiffness under constant electric field, 𝐶 33
𝐸 ′
, is defined by:
22
𝐶 33
𝐸 ′
=𝐶 33
𝐸 (1−
(𝐶 13
𝐸 )
2
𝐶 11
𝐸 𝐶 33
𝐸 )
(1-4)
𝐶 33
𝐸 ′
=2.39
(1-5)
And the clamped dielectric constant, 𝜀 33
𝑆 ′
, is defined by:
𝜀 33
𝑆 ′
=𝜀 33
𝑆 +
𝑒 31
2
𝐶 11
𝐸
(1-6)
𝜀 33
𝑆 ′
=722.8
(1-7)
Giving an electromechanical coupling coefficient of:
𝑘 33
′
=.614 (1-8)
The piezoelectric properties for the PMN-PT single crystal material were provided by the material
manufacturer and are listed in Table 1 and Table 2 below (Channel Technologies Group (formerly
HC Materials, Bolingbrook, IL) (CTG, Santa Barbara, CA).
23
Table 1: Elastic, dielectric and piezoelectric crystal properties of [001] poled PMN-
32%PT (Type-B) single crystal
Table 2 : Material properties of <001> poled PMN-PT single crystal
Properties Value for PMN-PT (PT = 32%)
𝑘 33
93%
𝐸 𝑐 (V/mm) 250
tan δ (1KH, 20°C) <0.008
Depolarization Temperature (°C) >85
Thermal Expansion Coefficient 10
-6
/° (20~70°C) >10
The material properties for <001> poled PMN-PT single crystal including the
electromechanical coupling coefficient, 𝑘 33
, the loss tangent, tan δ, which is the ratio of electrical
energy loss in a material versus in a vacuum, the depolarization temperature, which is the
temperature at which the piezoelectric properties degrade and the thermal expansion coefficient,
which relates the material expansion to changes in temperature are shown in Table 2. The most
24
widely used piezoelectric materials in ultrasound transducer applications was originally lead
zirconate titanate, Pb(Zr, Ti)O3 or PZT. The most common PZT piezoelectric materials used in
medical ultrasound imaging, PZT-5H, has a d33 = 583 x 10
-12
C/N and kt = 0.55. Single crystal
PMN-PT is preferable to PZT for miniaturized high frequency arrays because of its larger
electromechanical coupling coefficient and dielectric constant.
Ultrasound imaging resolution is another critical transducer characteristic and has both axial
and lateral components. Axial resolution (RA) and lateral resolution (RL) is the smallest distance
two objects can be differentiated in the depth and horizontal directions, respectively and is
described by:
𝑅 𝐴 =
𝑐 2
𝜏 −6𝑑𝐵
(1-9)
𝑅 𝐿 =𝑓 #
𝑐 𝑓 =
𝑍 𝑓 2𝑟 𝜆
(1-10)
where c is the speed of sound (m/s) in the imaging medium, 𝜏 −6𝑑𝐵
is the -6 dB transmitted pulse
length, the f-number (f#) is described by 𝑓 #
=
𝐹𝑜𝑐𝑎𝑙 𝐿𝑒𝑛𝑔𝑡 ℎ
𝐴𝑝𝑒𝑟𝑡𝑢𝑟𝑒 𝑆𝑖𝑧𝑒 , f is the frequency of the transmitted
pulse, Zf is the focal distance, r is the transducer aperture radius (for circular transducers) and λ is
the wavelength of the transmitted pulse. Both axial and lateral resolutions are therefore dependent
on transducer frequency. Another fundamental relationship of ultrasound imaging is that
attenuation of acoustic energy, which occurs primarily through the process of absorption of energy
into the tissue, increases with frequency and thus there is a fundamental tradeoff between
resolution and depth of penetration with ultrasound imaging. This limitation becomes critical when
designing imaging devices that operate at high frequencies and will be a key determinant in linear
array design for this project.
25
1.3 High Frequency Ultrasound Transducers
Subsequent to the initial demonstrations of single-element ultrasound scanning, the first real
time linear array scanners were built. These imaging systems could electronically scanned through
multiple elements but initially only provided time-motion (M-Mode) images that necessitated
pattern recognition to provided clinically useful information such as data for heart valve motion.
These arrays were further developed to provide 2-D B-Mode images and eventually
commercialized with the introduction of the Organon Teknika linear scanner in 1973 (Kamp and
Cramer 2008). Linear array design involves determining the optimal geometry of the individual
elements to achieve maximum sensitivity and bandwidth as well as an optimal beam profile. Linear
array elements are long and thin, with the height greater than the width to ensure they vibrate at
their half wavelength thickness resonance (Szabo 2004).
The development of transducer arrays necessitated the use of beamforming electronics,
composed of transmit and receive beamformers. These beamformers enable the system to
electronically scan, steer and focus the ultrasound beam by controlling the delay of transmitted
and received ultrasound pulses from individual elements in the array and summing these signals.
An array beamformer can dynamically focus the array up until the natural focus of the array
aperture and is performed more often on the receive side than on the transmit side since only a
single focal point can be achieved during transmit and multiple focal points can be used to delay
and sum signals on the receive side.
26
Figure 1-4: Analog receive beamformer diagram
Figure 1-5: Digital receive beamformer diagram
Both analog and digital beamformers have been used, with digital beamformers becoming
predominantly used in commercially available clinical imaging systems. Analog beamformers use
a delay-sum-detection-sampling scheme versus digital beamformers which first digitize echo
signals and then perform beamforming through sampling-delay-sum-detection scheme.
Beamforming techniques are based on the beamforming equation (Thomenius 1996):
27
𝑒 (𝑡 )=∑𝐴 𝑟𝑖
∑𝐴 𝑡𝑗
𝑉 [𝑡 −∆𝑡 𝑟𝑖
−∆𝑡 𝑡𝑗
+
2𝑟 (𝑡 )
𝑐 ]
𝑁 𝑗 =1
𝑁 𝑖 =1
(1-11)
Where e(t) is the summed echo waveform at the summing amplifier, V(t) is the transmitted
echo waveform, N is the number of array elements, r(t) is the focal distance at a particular time,
𝐴 𝑟𝑖
and 𝐴 𝑡𝑗
are the weighting functions for reception at channel i and transmission at channel j,
𝛥 𝑡 𝑡𝑗
and 𝛥𝑡
𝑟𝑖
are the time delays applied during transmission and reception to elements j and i,
respectively (Shung 2015).
The combination of advancements in array fabrication, beamforming and data acquisition
pushed array systems to quickly overtake mechanical scanners in popularity as their ability to
eliminate vibration improved image quality and mechanical robustness for commercial products
made them the more attractive solution for a majority of clinical applications. Single element
transducers remain highly useful, however, in applications requiring increasingly high frequency
operation for applications requiring resolution higher than that which was available from array
scanners such as in ophthalmic, small animal and intravascular imaging (Pavlin et al. 1991);
(Foster et al. 2000);(Mallery et al. 1987). The first demonstration of intravascular ultrasound
imaging was for coronary arterial imaging (Yock, Johnson, Linker 1988). This was an example of
a new type of imaging probe being developed specifically to provide much higher resolution
imaging of anatomy that was previously impossible to achieve because of the limitation of imaging
depth of standard hand-held transcutaneous imaging probes. Similarly, endoscopic imaging probes
were first used in the gastrointestinal tract (Tio et al. 1989). Both single-element radial scanning
transducers and curvilinear array probes with integrated fine needle aspiration capability were
demonstrated (Giovannini et al. 1995). These endoscopic probes also utilized miniaturized high
frequency imaging probes to produce high resolution images of anatomy that were not achievable
28
with standard imaging probes. Transrectal probes provided were later developed for detecting and
staging prostate cancer and later for guided prostate cancer biopsy and is another example of
clinical application driving the development of miniaturized, high-frequency ultrasound imaging
probes when hand-held, percutaneous imaging probes cannot provide sufficient imaging resolution
(Smith 1996).
Transducer array development for high frequency, miniaturized imaging probes has met some
critical challenges because of the limitation of both the current array fabrication techniques and
the electrical interconnect solutions. As imaging frequency increases, producing fully kerfed arrays
with element pitches less than 1.5 λ becomes impractical using traditional array fabrication
techniques. This is because these traditional methods have minimum achievable kerf widths and
element post aspect ratios that are greater than those required for these high frequency arrays. To
address the challenge of increasing array center frequencies that require ever smaller kerf widths,
various piezo composite fabrications methods have been developed for high frequency arrays with
center frequencies of 50 MHz and above. As the pursuit of ever smaller array kerf sizes caused the
traditional dice-and-fill array composite fabrication methods to become impractical, more
advanced fabrication processes including laser cutting and chemical etching techniques were
developed to produce kerfs as narrow as 4 µm techniques (Lukacs, Sayer, Foster 1998), (Cannata
and Shung 2003), (Cannata et al. 2006), (Savakus, Klicker, Newnham 1981) (Foster et al. 2009;
Liu et al. 2013). Interdigital pair-bonding, interdigital phase-bonding and stacking methods also
sought to decrease element pitch while still using existing dicing saw technology capabilities (Liu,
Harasiewicz, Foster 2001), (Yin et al. 2004), (Ritter et al. 2001), (Hackenberger et al. 2002). As
for developing new electrical interconnect solutions, capacitive- and piezoelectric-micromachined
ultrasound transducers (CMUT, PMUT) technologies have demonstrated the capability to produce
29
transducers and their electrical interconnects using batch fabrication techniques based on those
previously developed for the semiconductor industry (Oralkan et al. 2003), (Khuri-Yakub and
Oralkan 2011). However, most of these arrays are limited to clinical imaging frequencies (2-15
MHz) and have not been validated for miniaturized high frequency array applications. Kerfless
high frequency array development has been investigated as a fabrication method that avoids the
problem of creating kerfs in the piezoelectric material and instead relies on simply patterning
electrodes directly on the piezoelectric material to create individual array elements at the expense
of increased crosstalk between elements (Zhu et al. 2013), (Chen et al. 2014). Both thick and thin
film deposition techniques have also been investigated for high frequency array development
(Zhang et al. 2011), (Ito et al. 1995). More recently, deep reactive ion etching techniques were
proven for high frequency array development (Zhou et al. 2010), (Liu et al. 2012). To visualize
how array fabrication has developed in recent years, a chart was created to show that the
progression of array development with increasing array center frequencies is enabled by the
development of new fabrication methods. In this chart, arrays developed using the dice-and-fill,
laser etching and DRIE micromachining fabrication methods were selected as the most common
methods used in manufacturing processes. This array technological progress is shown in Figure
1-6 where the center frequency of the array and the publication date describing the device are on
the y- and x-axis, respectively. The box labeled ‘60 MHz Cummins, Shung et al.’ is a journal paper
currently under review for publication. This figure is annotated to show the method of fabrication
used to produce each array. This chart is useful in visualizing how DRIE fabrication techniques
will become critically important to continue high frequency array development.
30
Figure 1-6: Ultrasound array technology progression with dice-and-fill, laser micromachining and DRIE
micromachining fabrication methods.
Sources: (Ritter et al. 2002), (Cannata et al. 2006), (Démoré, Brown, Lockwood 2006), (Visualsonics), (Bezanson, Adamson, Brown
2014)
Additionally, the challenge of making dozens of individual electrical connections to the array
elements becomes pronounced with higher frequency arrays. In fact, all the arrays shown in Figure
1-6 utilized individual wire connections for each element either with standard wire bonding
techniques, flex circuit bonding or manually soldering coaxial cables. Hand soldering coaxial
cables was the initial standardized approach but this soon progressed to direct PCB/Flex circuit
bonding and wire bonding techniques as the element size and pitch decreased. As imaging
frequency rose for smaller and smaller arrays, more compact wire bonding methods were
developed (Bezanson, Adamson, Brown 2014). In this method, a flex circuit is cut and the internal
metal traces exposed in the cross section of the flex substrate serve as wire bonding pads. Wires
can be bumped from these pads to pads on the element electrode bonding pads located on the array
31
composite. This forms a compact electrical interconnect for forward looking arrays. Array
composite fabrication and miniature electrical interconnection has emerged as the two primary
challenges with high frequency, miniaturized ultrasound array development. New electrical
interconnect solutions must be developed to make miniaturized, high frequency ultrasound array
fabrication feasible so these devices can be used to address clinical applications that require
compact, high resolution ultrasound arrays. We will present a clinical need that requires a
miniaturized, high-frequency array probe solution as well as fabrication and novel electrical
interconnect techniques developed to build and test the array.
32
Clinical Need Identification
2.1 Ultrasound Guided Breast Biopsy
Clinical ultrasound imaging has been validated as an effective tool in guiding percutaneous
needle procedures such as central venous catheterization and other needle biopsy procedures
including transbronchial and pancreatic fine needle aspiration (Karakitsos et al. 2006; Herth et al.
2006; Williams et al. 1999). While many ultrasound guided interventional procedures exist, our
group identified percutaneous needle breast biopsy as an ideal application for which to develop a
new imaging tool to improve lesion visualization and biopsy targeting.
Image-guided needle biopsy is the gold standard for cancer diagnosis. These biopsies are
generally performed under ultrasound guidance for masses and stereotactic guidance for
microcalcifications (Parker et al. 1994; O'Flynn, Wilson, Michell 2010). Ultrasound guided biopsy
is the preferred method of choice since it is more comfortable for patients and does not require
ionizing radiation. Stereotactic guided biopsy, however, is chosen over ultrasound guided biopsy
for microcalcifications since current standard ultrasound systems cannot reliably visualize
microcalcifications, the presence of which can suggest early breast carcinoma such as Ductal
Carcinoma In-situ (DCIS).
Currently only external imaging probes are used to guide breast biopsy but these are limited
in the resolution they can provide because of the imaging depth they must maintain throughout the
procedure. There are three types of needles used during these ultrasound-guided minimally
invasive breast biopsy procedures. The first type is for fine needle aspiration (FNA), where a small
gauge needle is used to withdraw fluid and cells from a breast mass. FNA needles have been in
routine clinical practice since 1987 (Fornage, Faroux, Simatos 1987). The second type is a core
33
biopsy needle, which is commonly a coaxial design where the center needle component has a
sampling notch where tissue rests in and an outer component is a hollow metal tube with a sharp
cutting edge that severs a cylindrical core specimen from the surrounding breast tissue. These
needles, introduced in clinical practice in 1993, produce large tissue specimens, making
pathological sectioning and analysis easier and improving diagnostic yield of breast biopsy (Parker
et al. 1993). The third type of needles, VABB technique was first demonstrated in 1996, and
utilizes negative pressure generated by a vacuum pump to pull more tissue into the sampling notch
than could be with a core biopsy needle (Burbank, Parker, Fogarty 1996; Pfarl et al. 2002).
Mammotome brought the first vacuum assisted biopsy needle (VABB) to market with the first
clinical validation of study in 2002.
Before a tissue biopsy is performed a patient will have a mammogram taken either because it
is indicated or for screening. Digital mammography is the standard of care for screening and
diagnostic imaging for the detection of breast cancer. Several multi-center randomized clinical
trials demonstrated that screening mammograms reduce breast cancer mortality for woman over
the age of 50. The standard of care for breast cancer diagnosis is histopathology confirmation
following tissue biopsy, acquired through either fine needle aspiration, large core needle biopsy,
vacuum assisted needle biopsy or open surgical biopsy (Kerlikowske et al. 1995). Ultrasound
guided core biopsy diagnostic accuracy is comparable to open surgical biopsy with a diagnostic
sensitivity of 97.7% and a summary negative likelihood ratio of 0.03%. Ultrasound-guided vacuum
assisted core needle biopsy has a similar diagnostic of 96.5% and a summary negative likelihood
ratio of 0.036 (Bruening, Schoelles, Treadwell 2011). Despite the high diagnostic accuracy of core
needle biopsy and vacuum assisted needle biopsy, the fact that over 1.6 million women undergo
breast biopsy annually in the United States, thousands of patients receive false negative diagnoses,
34
allowing cancer to progress and increasing mortality (Silverstein 2009; Youk et al. 2007). The
challenge of missed diagnosis lies not with the failure of pathological analysis but with tissue
sampling at the time of biopsy. Thus, there is a clinical need to improve tissue sampling during
ultrasound-guided procedures. Microcalcification visualized with x-ray mammography is the most
reliable mammographic feature for predicting breast cancer diagnosis. The problem is that some
lesions and the microcalcifications in them cannot be visualized with standard ultrasound imaging
techniques (Bassett 1992).
Standard clinical ultrasound imaging relies on transmitting and receiving ultrasound energy.
The redistribution of this energy from a transmitted wave occurs through either reflection, when
the wavelength is much smaller than the object it encounters, or scattering, when the wavelength
is greater than or comparable to the dimension of the object. This is particularly relevant in breast
imaging since sampling of microcalcifications gives the highest chance of making a diagnosis
since the presence of microcalcifications is highly correlated with cancer. However, the
wavelengths of current clinical imaging systems (5-15 MHz) with respective wavelengths (300
µm – 100 µm) limits the ability of receiving strong echo signals from these microcalcifications
because scattering is the dominant mode of energy redistribution since the size of
microcalcifications are largely <100 µm (Soo, Baker, Rosen 2003). Since the fundamental
limitation of wavelength cannot be changed for these clinical imaging systems, attempts to
improve image processing of ultrasound echo data to highlight these microcalcifications has been
implemented, such as in the Toshiba Aplipure system (Toshiba America Medical Systems 2008).
The Aplipure system is a compound imaging to improve contrast of breast images which help the
user visualize microcalcifications within the tissue (Jespersen, Wilhjelm, Sillesen 1998).
35
Other ultrasound imaging techniques that seek to improve contrast between tissue types for
better tissue type characterization included elastography, vibro-acoustic imaging and
photoacoustic imaging. Elastographic imaging has been used to differentiate tissue types in the
breast included differentiating between benign and malignant microcalcifications. In a study by
Cho et al., elastographic imaging was performed on patients who had confirmed microcalcification
clusters from mammography and a tissue elasticity score was created for the tissue containing the
microcalcifications. These scores were correlated with pathology by taking a core biopsy sample
of the same tissue (Cho, Moon, Park 2009). Vibro-acoustic imaging utilizes low frequency
ultrasound waves to excite tissue with an oscillating radiation force, producing acoustic emission
from objects within the tissue, and recording these emissions with a hydrophone. By raster
scanning the excitation transducer across the tissue surface, an image can be formed. This
technique was able to image microcalcifications in excised breast tissue as small as 110 µm (Alizad
et al. 2004; Kim et al. 2014). Photoacoustic (PA) imaging has been employed to image
microcalcifications in breast tissue. Kim et al. demonstrated that using 700nm and 800nm light to
excite breast biopsy core tissue samples and a 7 MHz linear array to receive echoes could correctly
identify breast microcalcifications with a sensitivity and specificity of 90.9% and 80.0%,
respectively (Kim et al. 2014).
2.2 Defining the Clinical Need
To summarize the clinical need, physicians need an imaging tool that can provide high
resolution images of tissue adjacent to the biopsy needle in real-time to enable better tissue
visualization including microcalcifications to improve breast cancer diagnostic accuracy by
36
reducing sampling error. This ultrasound imaging tool will be designed to work in conjunction
with the external guidance ultrasound system. The high resolution images produced by the new
imaging tool will be used to identify microstructures within suspicious lesions in the breast and
direct tissue sampling.
Standard clinical ultrasound imaging relies on transmitting and receiving ultrasound energy.
The redistribution of this energy from a transmitted wave occurs through either reflection, when
the wavelength is smaller than the object it encounters, or scattering, when the wavelength is
greater than or comparable to the dimension of the object. This is particularly relevant in breast
imaging as the ability to identify and sample microcalcifications increases the probability of
sampling potentially malignant tissue. The wavelengths of current clinical imaging systems (5-15
MHz) with respective wavelengths of 300 µm – 100 µm, limits the ability of receiving strong echo
signals from microcalcifications. The size of microcalcifications, usually < 100 µm, then entails
that “scattering” becomes the dominant mode of energy redistribution obscuring their
identification (Soo, Baker, Rosen 2003). Since the fundamental limitation of wavelength cannot
be changed for these clinical imaging systems, attempts to improve image processing of ultrasound
echo data to highlight these microcalcifications has been implemented by systems such as the
Toshiba Aplipure System (Toshiba America Medical Systems 2008). The Aplipure system uses
compound imaging to improve contrast of breast images, which helps the user visualize
microcalcifications within the tissue (Jespersen, Wilhjelm, Sillesen 1998). Providing enhanced
microcalcification visualization enables physicians to better target tissues they hope to sample with
the core biopsy needle. As sampling error is the primary cause for false negative results, improving
resolution of specific structures (i.e. microcalcifications) that may represent DCIS, is a promising
solution that will improve the sensitivity of breast cancer biopsy procedures under ultrasound
37
guidance. In summary, radiologists need an ultrasound system that can provide high resolution
images of tissue and structures such as microcalcifications in real-time to guide the biopsy needle
and improve diagnostic accuracy and reduce sampling error.
We have built and tested a prototype miniaturized high frequency linear array integrated
within a core biopsy needle to provide ultra-high resolution images of breast tissue. Such images
obtained within tissue may allow radiologists to identify features such as microcalcifications not
previously seen during conventional biopsy guidance thus procuring tissue with a higher
diagnostic yield for malignancy and a lower false negative rate. The concept of integrating an
ultrasound imaging array within a breast cancer biopsy needle has been demonstrated by Cochran
et al. Cochran where they have fabricated a 15 MHz linear array that can fit within a 2 mm diameter
biopsy needle. This device is still limited however as imaging at 15 MHz cannot reliably provide
the necessary resolution to differentiate small structures within the breast such as lobules, ducts or
microcalcifications (Bernassau et al. 2009). We have thus embarked on building a much higher
frequency array that can resolve these fine structures and does so from within the slender housing
of a biopsy needle.
We have evaluated external and internal imaging arrays to address this clinical need. Given
the small size of microcalcifications and the fact that lesions within the breast can be several
centimeters below the skin surface, imaging systems require axial and lateral resolutions that
correspond to center frequencies above 50 MHz to resolve these individual structures. An external
imaging probe is not a viable solution here since acoustic energy attenuation in tissue would be
prohibitively large. Ultrasound probes that image the lesion from within the tissue, through an
interventional tool such as a core biopsy needle, becomes a viable solution. Under these conditions,
the requirement for depth of penetration is limited to the tissue sampling distance of the core biopsy
38
needle, which is less than 4 mm for most existing core biopsy needles. The maximum depth of the
sampling notch in these core biopsy needles is limited by the external diameter of the needle.
Needle diameters range from 8 AWG – 20 AWG for ultrasound guided needle biopsy which have
an outer diameter ranging from 4.19 mm – 0.91 mm (Sigma Aldrich 2014). We have built and
evaluated this imaging array and have designed a clinical imaging study to validate that high
frequency ultrasound imaging is useful for identifying tissue features such as microcalcifications
to guide the tissue biopsy.
39
Imaging Probe Design
3.1 Array Transducer Design
The array design requirements as informed by the clinical need include array type, aperture
size, orientation, frequency and packaging size. A side-looking, linear array is needed in order to
match existing imaging plane used for needle guidance. In practice, a radiologist would guide the
biopsy needle with integrated imaging array to the suspicious lesion using standard external
ultrasound probe guidance, and once the needle is in close proximity to the lesion, the integrated
array produces a high resolution picture that the radiologist then uses to target the precise point
where the tissue is sampled. During these procedures, the physician primarily uses what is known
as ‘in-plane’ technique where the long axis of the needle shaft is parallel with the imaging plane
of the ultrasound array probe. This is utilized to enable simultaneous visualization of the lesion
within the breast and the needle shaft throughout the procedure. Lesion visualization is critical
since the radiologist is looking to sample a specific tissue section according to lesion
characteristics such as contrast with surrounding tissue and irregularly shaped borders. Needle
visualization is also critical since the radiologist must confirm that the path of advancement for
the needle through the tissue is directly towards the lesion. The radiologist keeps the needle within
the imaging plane so that it does not cause excessive trauma such as hemorrhaging caused by
piercing a blood.
Since radiologist are trained to use in-plane technique during biopsy procedures, any
additional complementary imaging modalities or devices should also produce images with the
same imaging plane so that interpretation of both images, the existing external guidance imaging
probe and the new imaging device, can be done simultaneously during the procedure. Therefore,
40
in designing a miniaturized high frequency ultrasound array we wish to orient the azimuth axis of
the array in parallel with the needle shaft so that the imaging plane is parallel with that of the
external guidance probe.
Figure 3-1: Rendering of core biopsy needle (a) oriented in the in-plane technique within the imaging
plane of the external guidance linear ultrasound array and (b) zoomed view of array with its imaging plane
oriented parallel to that of the guidance probe.
In order the array to be oriented along the long axis of the needle, the array and electronic
connections must be housed within the biopsy needle so as not to increase the cross-sectional
profile of the needle and cause additional trauma for the patient. Needle lengths range from 6-17
cm in length and have a tissue sampling notch approximately 2 cm in length that is oriented
towards the side of the needle so that it faces the external guidance imaging probe during biopsy
procedures. Most core biopsy needles operate by a coaxial cutting cylinder that functions as a
cutting sheath as it slides passed the sampling notch in the inner needle shaft (CR BARD ; Temno ).
41
Figure 3-2: Rendering of imaging array integrated within large core breast biopsy needle with (a)
sampling aperture open and (b) sampling aperture closed.
The array must have significantly high center frequency in order to achieve the resolution
needed to visualize structures not currently visible with clinical imaging systems. Currently,
clinical ultrasound imaging systems that are used to visualize breast lesions are limited to center
frequencies below 15 MHz and thus have limited resolution and cannot reliably visualize small
microcalcification clusters below 100 µm. The original imaging array center frequency was chosen
to be 80 MHz to achieve axial and lateral resolution necessary to reliably visualize
microcalcifications including those below 100 µm. Array geometry has been determined by the
general rules presented by Ritter (Ritter 2000). An array device was chosen because a single
element 80 MHz transducer is not a feasible solution to the clinical problem of providing high-
resolution ultrasound imaging from within a core biopsy needle since the imaging should be in
real-time. To accomplish this with a single element transducer would require either a high speed
linear oscillator with a drive shaft that extended the length of the biopsy needle or a miniaturized
drive motor located within the needle shaft itself. The challenge with both of these is the limitation
in frame rate caused by physically translating the transducer and also the mechanical vibration
resulting from this motion that would interfere with the precise needle placement required to
sample breast tissue. A rotational single-element solution such as those used for intravascular
42
imaging is not an ideal solution since the imaging plane or rotational elements is orthogonal to the
axis of rotation, which would be the same as the needle shaft in this case. Since using an array
imaging probe avoided the issue of mechanical scanning by electronically scanning through each
individual element, issues of frame rate being limited by the inertia of the imaging element and
the mechanical vibration caused by scanning are avoided.
In designing the array, the choice of building a linear, phased and curved-linear array was
considered. A curved-linear array was first eliminated since the array must fit within a straight
core biopsy needle and the imaging aperture should look directly out the side of the needle and
also maintain a minimal cross-sectional profile. Additionally, a phased array was not chosen since
they require kerf widths of less than 0.5 λ and at 80 MHz we could not identify a fabrication
solution to reliably build arrays with this small of kerf width. Therefore, a linear array was chosen
as the model to design the miniaturized array. Among linear arrays, 1D, 1.5D and 2D arrays can
be designed depending on the application and probe size constraints. Limitations of the electrical
interconnect from the channel traces to the individual array bonding pads prevented 1.5D and 2D
array designs from being built since connecting. Thus an 80 MHz 1D linear array was chosen as
the solution.
At this center frequency we needed to have less than a 1.5 λ pitch to avoid undesirable grating
lobes. Grating lobes are pressure lobes that occur in the imaging field because of constructive
interference from pulses emitted from regular spacing of individual elements in an ultrasound
array. These lobes are undesirable since their echoes return to the ultrasound array elements are
combined into the beamformed signal, causing artifacts in the final image. The angular location of
the grating lobes is a function of the wavelength (λ) and element pitch (g) by equation:
43
𝜙 𝑔 =𝑠𝑖𝑛 −1
(
𝑛 λ
𝑔 )
(3-1)
where is an integer = ±1, ±2, … A linear array center frequency of 80 MHz requires a maximum
allowable pitch of 1.5 λ. Given a sound speed of 1525 m/s, a 1.5 λ pitch is 28.6 µm. Traditional
dice and fill techniques cannot achieve the element pitch and kerf width needed in this application
and laser dicing techniques such as excimer laser micromachining has only been demonstrated
with arrays up to 50 MHz (Napolitano et al. 2006; Foster et al. 2009). High frequency composite
arrays have been demonstrated in recent publications with center frequencies from 20 MHz – 60
MHz. Composite materials refers to piezoelectric posts or rods embedded within lower density
polymer materials. Composites are categorized as either 1-3 or 2-2 composites depending on the
orientations. This notation, developed by Newnham, refers to the number of directions (1, 2 or 3)
the piezoelectric and polymer filler material are connected in, respectively. Therefore a 1-3
composite is composed of piezoelectric pillars or rods oriented in a single direction (Z-axis) while
the polymer filler in which they are embedded are connected in 3 directions (X-, Y- and Z-axis).
A 2-2 composite differs in that both the piezoelectric and polymer filler material are both
connected in 2 directions.
The most common piezoelectric material used in medical imaging applications, lead zirconate
titanate, Pb(Zr, Ti)O3 or PZT has been used in composite materials ranging in volume
concentration of piezoelectric material as a fraction of total volume from 20% - 70%. PZT
composites are characterized by lower acoustic impedance (4 to 25 MRayls) than conventional
PZT bulk material (34 MRayls) which better matches the acoustic impedance of skin, enabling
increased signal energy transmission from the array to the skin and vice versa (Shung 2015; Liu et
al. 2013; Newnham, Skinner, Cross 1978).
44
Figure 3-3: (a) 1-3 AND (b) 2-2 composite structure.
More recently advancement in piezoelectric composite material research including the
development of relaxor-based ferroelectric materials have shown higher dielectric constant and
electromechanical coupling coefficient than conventional PZT. Fine grain PZT materials like lead
magnesium niobate-lead titanate (PZN-PT) and single-crystal ferroelectric materials like lead zinc
niobate-lead titanate are examples materials that have proven to be very effective for high-
frequency applications. These materials exhibit high electromechanical coupling coefficient (k33 ~
0.9) and dielectric coefficient (d33 > 2000 pC/N). Composites also have lower electrical impedance
compared to bulk material, which is important for building high frequency arrays with relatively
small element areas (Choi et al. 1989; Shrout and Fielding Jr 1990).
The advantage of higher sensitivity that composites offer is even more significant for
miniaturized arrays that sacrifice element and total aperture size for more compact dimensions.
Composites have been utilized for low frequency, underwater applications for decades and have
been successfully implemented for medical ultrasound imaging transducers. While composites
offer many advantages over bulk materials, they come with a significant drawback. The periodic
microstructure of these composites results in the formation of resonances called Lamb waves (Baer
45
and Kino 1984). These resonances, when coupled with the fundamental thickness-mode resonance
frequency, produce signals across individual ultrasound transducers and degrade the overall
performance of the device (Certon et al. 2001). Ideally each array element operates independently,
but the presence of these lateral resonances causes groups of elements to operate together, causing
changes in beam pattern, electrical impedance and echo response (Démoré, Brown, Lockwood
2006).
As arrays are designed and built for progressively higher frequencies, new fabrication
techniques are implemented to meet the demand of these designs with smaller and smaller element
geometries. A key driver in developing new fabrication techniques is the generation of kerfs
between elements. For this array we had the choice between designing a 1-3 composite and a 2-2
composite. In a 2-2 composite, element posts are regularly spaced allowing lateral modes to
present themselves and should be avoiding by minimizing the kerf width between elements
according to:
𝐾𝑒𝑟𝑓 𝑊𝑖𝑑𝑡 ℎ≤
𝑉 𝑠 4×𝐹𝑐
(3-2)
𝐶𝑒𝑟𝑎𝑚𝑖𝑐 𝑊𝑖𝑑𝑡 ℎ≤
𝑉 𝑙 4×𝐹𝑐
(3-3)
Where Vs is the shear wave velocity in the polymer, Fc is the device center frequency and 𝑉 𝑙
is the longitudinal wave velocity for the width resonator in the piezoelectric material. 𝑉 𝑙 can be
calculated by:
𝑉 𝑙 =
√
𝐶 11
𝐸 𝜌
(3-4)
46
where 𝐶 11
𝐸 is the elastic stiffness constant and 𝜌 is density (Kino 1987). Using these equations we
can estimate the array design parameters needed for this application. Previous examples of high-
frequency composite arrays built from single crystal PMN-PT material were demonstrated and
used as a model for our array development. Previous examples of micromachining for ultrasound
arrays and single elements were especially useful in this design process. Micromachining
techniques have been utilized for high frequency array design including a 60 MHz PMN-PT/epoxy
1-3 composite array. This composite was built from a PMN-PT single crystal material using deep
reactive ion etching (DRIE) to etch through the material and expose individual pillars. Each square
pillar had a width of 105 µm and kerf width of 5 µm (Liu et al. 2013).
Advances in ultrasound array miniaturization has led to accompanying improvements in
electrical interconnect methods including modification of flex circuits for forward looking phased
arrays. The most challenging aspect of making electrical connections for high frequency
miniaturized array is bridging the gap between the flex circuit/transmission line and the individual
array bonding pad. Brown et al. accomplished this by traditional wire-bonding with a novel flex
circuit via machining approach (Bezanson, Adamson, Brown 2014). To date advances in
ultrasound fabrication have been accomplished by making individual, hand-made ultrasound
arrays to demonstrate feasibility in a bench-top, pre-clinical or clinical setting. However, as arrays
are built with increasing frequency and applications demand miniaturization of these arrays, new
fabrication processes are needed that account for the challenge of handling arrays with sizes on the
order of 1 mm
3
throughout the fabrication and testing process. Batch fabrication techniques
developed in the semiconductor industry, may be a suitable solution to this need for piezoelectric
ultrasound array development. Capacitive- and Piezoelectric- micromachined ultrasound
transducers (CMUT, PMUT) have demonstrated the ability to produce transducers and their
47
electrical interconnects using batch fabrication techniques (Oralkan et al. 2003; Khuri-Yakub and
Oralkan 2011). However, these arrays are limited to clinical imaging frequencies (2 – 15 MHz)
and have not been validated for miniaturized high frequency array fabrication.
Array Simulation Modeling
While progressively higher frequency arrays over 50 MHz have been developed, none has
considered the size constraints that may occur in the design of a clinical needle biopsy device. In
addition to the theoretical design constraints for the array design we had to consider the limitations
of the fabrication techniques available. For example a 6 µm kerf width was the smallest we could
reliable use given the etching depth and aspect ratio of the element pillars for the composite. The
etching angle was 89° for this DRIE process, which is very close to the ideal 90°. An element
height of 22 µm which was determined as optimal using the simulation techniques described in
chapter 5 was also feasible given that the DRIE could etch through 30 µm of bulk PMN-PT
material to allow for several micrometers of thickness lost during lapping of the composite. 1-3
composite fabrication was not preferred for the array elements at the time of fabrication because
of risk of pillars breaking during the fabrication process. The applications for these arrays largely
mirrored standard clinical applications such as transcutaneous, intravascular, intracranial or
endoscopic imaging. The challenge that we are addressing is improving breast cancer needle
biopsy procedures and this requires much more demanding physical size constraints since the
entire linear array and electrical connection must fit within a biopsy needle shaft. Both the array
fabrication and electrical interconnect solutions must be specially developed for this application.
48
4.1 Array Simulation Approach
Once general array design requirements have been established, various modeling techniques
can be used to establish ideal array dimensions. These values, when considered in the context of
available fabrication and testing tools and techniques, determine the final array dimensions. In the
case of the miniaturized high frequency array designed considered here, new fabrication
techniques must be considered and thus knowing which array dimension are achievable is difficult
to predict. In fact, array dimensions were calculated using the modeling techniques concurrently
while new fabrication process flows were established. For array simulation, we first performed
single element simulations using the one dimension analogous electrical circuit model in
PiezoCAD modeling software (Sonic Concepts, Bothell, Washington). Once ideal single element
dimensions are established, a beam simulation software package, Field II, a set of program files
running on MATLAB software (MathWorks, Natick, MA), is used to determine the element width,
length and pitch as well as sub-aperture size required for acceptable beam formation (Jensen and
Svendsen 1992; Jensen 1996). Lastly, finite element modeling (FEM) of the array and imaging
field is performed using PZFlex (Weidlinger Associates, Inc., Mountain View, CA).
4.2 Array Simulation Methods
4.2.1 PiezoCAD
Single element modeling using PiezoCAD is an effective way to quickly determine ideal
element parameters such as acoustic stack dimensions and materials by evaluating simulated
pulse/echo impulse and electrical impedance responses. This software produces predicts the
element’s pulse/echo response, insertion loss and electrical impedance. PiezoCAD uses a chain
49
matrix technique to calculate system characteristics beginning with the electrical port to the front
acoustic port. Typically for single element fabrication, two matching layers are used to better
match between the large acoustic impedance of the transducer material and the significantly lower
impedance of the imaging medium, which is effectively equivalent to water in the case (Cannata
et al. 1999).
However, in our initial designs we concluded that only a single matching layer of parylene
could be deposited on the surface of the transducer since attaching another matching layer, ideally
2-3 µm silver particles and INSULCAST 501 (ITW Polymers Coatings, Montgomeryville, PA),
would be impractical given the dimensions, orientation and position of the array and electrical
interconnect. Since the array aperture size must be very small (~1 mm x 1.9 mm maximum) for a
64 element array, a matching layer with thickness of ¼ λ would mean an extremely fragile
component that would need to be attached by hand to each individual array since the array
fabrication process would prevent attaching a large matching layer and then dicing out individual
arrays. Further details of array fabrication will be provided later to demonstrate this necessity. As
a result a single layer of parylene-C polymer was chosen to serve as the single matching layer,
with its properties given in Table 3 below.
Table 3: Matching layer material properties
Property Value
Density (g/cm
3
) 1.1
Velocitylong (m/s) 2350
Acoustic Impedance (MRayls) 2.59
Source: Sonic Concepts, Bothell, Washington
50
A backing layer was chosen given the need for it to serve as a common electrical ground for
all 64 elements and provide rigidity to the thin array composite material. The material should have
high attenuation but low acoustic impedance relative to the composite to maximize sensitivity. The
conductive epoxy E-Solder 3022 (Von Roll USA, Inc., Schenectady, New York) was chosen as it
met all these requirements with its properties as measured at 30 MHz after being centrifuged at
3000RPM for 10 minutes are shown in Table 4 below (Wang et al. 1999).
Table 4: Backing layer material properties
Property Value
Density (g/cm
3
) 3.2
Velocitylong (m/s) 1850
Attenuationlong (dB/mm) 1.1∗10
2
Acoustic Impedance (MRayls) 5.92
Source: (Wang et al. 1999)
Array element dimensions were optimized through a process of design parameter modification
followed by evaluation of the two-way pulse/echo response as well as electrical impedance
response. The final optimized array element dimensions are displayed in Table 5 and the key
simulation results are displayed in Table 6 and Figure 4-1 displays the pulse/echo impulse
responses for these values.
51
Table 5: Array simulation parameters
Parameters Value
Piezoelectric Element Material PMN-PT single crystal
Element Width 14 µm
Element Length 1 mm
Piezoelectric Element Thickness 22 µm
Backing Layer Material E-Solder 3022 (centrifuged)
Backing Thickness 1 mm
Matching Layer Material Parylene-C
Matching Layer Thickness 6.9 µm
Table 6: PiezoCAD array element simulation results
Property Value
-6 dB Center Frequency 87.99 MHz
-6 dB Bandwidth 32.2 MHz (40.01%)
Insertion Loss -42.62 dB
Figure 4-1: PiezoCAD simulation result plots for 2-way pulse/echo transmit for single array element.
52
4.2.2 Field II
Once the one dimensional PiezoCAD modeling yielded ideal element dimensions for the
single array elements, Field II software was used to simulate the sound field produced by multiple
element focused transmit excitations. This is valuable since the PiezoCAD simulation only
accounts for two-way signal attenuation and not for signal diffraction in the imaging medium,
which is unrealistic in actual clinical imaging. The Field II program system uses spatial impulse
responses to predict the sound field, a concept developed by Tupholme and Stephanishen
(Tupholme 1969; Stepanishen 1971a; Stepanishen 1971b). Jensen demonstrated a method to
calculate pressure fields from arrays with arbitrary shapes, element sizes and pitches, apodizations
and excitation waveforms and it is this method which is used in the Field II program. The program
simulates beam profiles by convolving the spatial impulse response of the array with the excitation
function of the transducer elements. The spatial impulse response is the emitted ultrasound field
at any particular point in the sound field at a given time when the array is excited by a Dirac delta
function. In other words, the spatial impulse response is the impulse response as a function of time
at a location within the sound field relative to the array transducer aperture, which can be arbitrarily
shaped. Therefore, this method can predict the sound field of an arbitrarily shaped array with an
arbitrary excitation waveform, making the Field II program a valuable tool for array design. Array
beam profile optimization can be achieved relatively quickly once on the single element
dimensions are estimated using the one-dimensional PiezoCAD simulation and the overall array
dimensional constraints are known based on the clinical need and electrical interconnect design.
A major consideration in beam profile simulation is grating lobe angular position and
magnitude relative to the main lobe. Distancing grating lobes from the main lobe is achieved by
reducing the pitch and for a linear array, a pitch of less than 1.5 λ is needed. Additionally,
53
maximizing the element width as a fraction of the element pitch will minimize the magnitude of
the grating lobe, and thus reducing kerf width is desirable. The simulation parameters and results
for a beam profile for a sub-aperture of 32 elements in the linear array focused at a point 1.5 mm
from the array surface with an element pitch of 20 µm, element width of 14 µm and kerf width of
6 µm are shown in Table 7 and Figure 4-2, respectively. Table 8 shows the simulated bandwidth
and resolution values for the array.
Table 7: Field II Simulation Transmit Parameters used for Transmit Beam Profile
Simulation.
Parameters Value
Active Aperture 32 elements
Element Width 14 µm
Kerf Width 6 µm
Element Length 1 mm
Element Thickness 22 µm
Matching Layer Thickness 6 µm
Focal Distance 1.5 mm
54
Figure 4-2: Field II beam profile 2D plot as well as axial and lateral beam profile plots.
Table 8: Simulation results compiled from PiezoCAD and Field II simulations.
Parameter Value
-6 dB bandwidth 30%
Lateral Resolution 43 µm
Axial Resolution 45 µm
To determine the optimal kerf and element width geometries to both minimize grating lobe
magnitude and maximize its angle, a matrix of 10x10 variables was computed for both the kerf
and element width and displayed on the mesh plots in Figure 4-3 below. These plots were useful
in determining the ideal geometry while considering fabrication limitations.
55
Figure 4-3: Grating Lobe angle and magnitude mesh plots.
4.2.3 PZFlex
Thus far, one-dimension single element modeling and two-dimensional beam profile
modeling were performed to predict which array dimensions are ideal for the 80 MHz linear array
being designed. However, neither of these simulation software packages accounts for movement
in the time domain of array material and therefore are not entirely accurate predictors or ideal array
design. Finite element modeling (FEM) is an ideal solution for predicting vibrational modes
propagating through composite materials and PZFlex is a time-domain FEM software package
optimized for modeling piezoelectric composite arrays for ultrasound imaging. For a high
56
frequency piezoelectric composite ultrasound array, an FEM tool is extremely useful to predict the
pulse/echo response since it allows undesirable and realistic acoustic wave propagation modes
within the composite that contribute to cross talk. Both electrical and mechanical cross-talk are
simulated with PZFlex and thus it is a useful tool in designing array geometries to mitigate these
effects. We evaluated the two-way pulse/echo response, FFT spectrum and electrical impedance
response for various array geometries, focal lengths and excitation waveforms. Using the array
dimensions optimized using PiezoCAD and Field II as a starting point we finalized the array
dimensions using this technique. The optimized simulation parameters were taken from the Field
II simulations and the two-way pulse/echo response, FFT spectrum and electrical impedance
response are shown in Figure 4-4.
57
58
Figure 4-4: PZFlex simulation result plots for pulse/echo experiment as well as electrical impedance
magnitude and phase plots for individual element.
59
4.2.4 Final Array Design
A rendering of the final array design parameters and rendering is shown in Table 9 and Figure
4-5 below. This design was reached after consideration of theoretical array design constraints,
simulation results and fabrication process limitations.
Table 9: Final array design acoustic stack and element geometry
Figure 4-5: Exploded view of array acoustic stack including backing, 2-2 composite and parylene matching
layer as well as zoomed view of 3 separate layers and the final assembled array unit.
Parameter Value
Active Aperture 32 elements
Element Width 14 µm
Kerf Width 6 µm
Element Height 1 mm
Element Thickness 22 µm
Matching Thickness (Parylene) 6 µm
Backing Thickness (E-Solder 3022) 1 mm
Focal Distance 1.5 mm
60
Fabrication and Packaging
Each element in the 2-2 composite was composed of single crystal PMN-PT piezoelectric
material and was separated from its neighbors by epoxy. The array composite was backed by a
highly attenuative conductive epoxy that served as the ground connection for each element. The
array had a single polymer coating matching layer to provide acoustic impedance matching
between the array composite and the imaging medium. Array elements were connected to a
transmission line mounted directly over the surface of the array and the array aperture was exposed
to the imaging environment through a small rectangular cutout in the transmission line. This
integrated array transmission line component can be mounted within a coaxial biopsy needle where
it would image out through the side of the tissue sampling aperture. The array design, fabrication
steps and testing results are described in the following sections.
One critical challenge of fabricating miniaturized high frequency arrays is creating an
electrical interconnect solution. Fabricating arrays with center frequencies above 50 MHz requires
element widths that fall below the standard metal trace width limits of flex circuit manufacturers.
Thus, directly bonding array element electrodes to matching electrode patterns on flex circuits is
not possible since the flex trace pitch is larger than the element pitch of the array. The obvious
solution for this is to increase the array element electrode bonding pad pitch through a fan out
pattern so that it is large enough to match with the flex trace metal pads. If a fanout pattern is used
on the array composite, the next step is to determine whether a flex circuit can be layered on the
top (matching layer side) or bottom (backing layer side) surface of the array composite. The
maximum allowable thickness of a flex circuit material (i.e. polyimide) that can act as a matching
layer falls below what is feasible for these same flex manufacturers. The smallest standard
substrate thickness achievable is ½ mil (~12.5 µm) and for high frequency arrays, including the
61
array described here, the matching layer thickness must be well below 10 µm. Therefore,
positioning a flex circuit on the matching layer side of the array composite is not an option for
direct bonding pad connection. While positioning the flex on the backing side of the composite
would avoid the issue of being too thick, this is not feasible for a fabrication process point of view
since there would not be a way to lap the array composite to its final thickness, de-bond it from its
carrier wafer and flip it over to pattern the individual element electrodes. The composite would be
far too thin to handle and would be destroyed during this step. The delicacy of the array composite
requires that the electrical interconnect steps be compatible with wafer level fabrication processes.
Since layering a flex circuit component on either the top or bottom surface of the array composite
are not possible, a different solution must be developed.
Another design constraint that this electrical interconnect must conform to is the ability to
match the electrical impedance characteristics of the array elements. Since these array elements
are so small (14 µm x 1 mm area and 24 µm thickness), electrical impedance matching becomes a
major issue since all the imaging electronics used have an industry standard 50 Ω electrical
impedance. For instance, the simulated electrical impedance magnitude at 80 MHz for a 22 µm for
a single element was 350 Ω and if we are to match this impedance to that of the pulser/receiver
and multiplexer electronics (50 Ω), a transformer impedance matching circuit for each individual
channel is needed. The surface mount transformers are large enough to where they cannot be
integrated within the housing of a needle shaft. Therefore, a transmission line routing the channel
traces from the individual array element bonding pads to the transformer circuits mounted on a
PCB at the proximal end of the needle shaft must match the electrical impedance of the array
elements. A microstrip transmission line solution was developed for this array. This transmission
line would need a minimum substrate thickness between the traces and ground plane of at least
62
200 µm. Possible substrate materials for this type of transmission line include flex circuit (i.e.
polyimide), PCB (i.e. fiberglass) and glass materials (i.e. borosilicate). The type of substrate
material was determined by its ability to accommodate this needle array’s necessary metal trace
width and space which allowed it to meet the needle diameter size requirements for this clinical
application.
The minimum transmission line thickness of 200 µm made curling the transmission line
around the long axis of the needle shaft impractical given the fact that flex circuit substrates have
a minimum bend radius determined by 10x the substrate thickness. With a 1 mm element height
and a 2 mm bend radius on each side, the total package size would be 5 mm in diameter if a flex
substrate is curled around the array. Thus, a curled flex circuit solution is impractical for needle
based clinical applications such as soft tissue biopsy. Since curling the flex around the array was
not practical, the transmission line substrate would either have to be mounted above or below the
array composite in a single plane while still avoiding the issue of obstructing the array aperture.
Mounting the transmission line below the array composite meant the substrate would also be
serving as the acoustic backing for the composite which made glass or PCB substrates impractical
since their acoustic impedance values are much larger than those of traditional backing materials
such as the highly attenuative E-Solder 3022 silver epoxy commonly used in single element and
array transducer fabrication.
Mounting the transmission line on the front of the array composite with the substrate
completely covering the surface of the composite was impractical for all substrates since it would
be much larger than the maximum allowable thickness for a material serving as an acoustic
matching layer. However, as we demonstrated with the current array, cutting a hole in the substrate
material to allow the array aperture an unobstructed view of the imaging medium solves this
63
problem. The drawback of this solution was that the cutout reduced the area of the substrate on
which all 64 metal channel traces could be patterned. To compensate for this lack of usable
substrate to route traces over, the minimum transmission line width we were able to achieve was
2.7 mm as shown in Figure 5-1 below, which depicts the final transmission line design using a
glass substrate.
Figure 5-1: Transmission line (a) outline and (b) zoomed figure of transmission opening for array
aperture.
This transmission line solution uses only a single layer metal trace design. Future versions of
the transmission line may be able to reduce the 2.7 mm width of the component if a multilayered
trace design is used. If a dual layer design is used to route the channels then the width could be
reduced to 2.0 mm. Further reduction of the transmission line width is constrained by the width of
the rectangular imaging aperture cutout width. Reducing the width of the transmission line further
than 2.0 mm requires a new type of electrical interconnect solution. One potential solution that
meets all the requirements for high frequency miniaturized array electrical interconnects is
proposed in Chapter 8.
64
5.1 Electrical Interconnect Transmission Line
A new electrical interconnect solution was developed to serve as both a high density
interconnect to the 64 array element bonding pads and an electrical impedance matching circuit to
match the high impedance of the array elements to the standard 50 Ω impedance of the multiplexer
and pulser/receiver electronics. This interconnect solution was comprised of both a compact
microstrip glass transmission line and impedance matching PCB connector board. The microstrip
transmission line had a designed electrical impedance magnitude of 350 Ω to match that of each
individual array element. The impedance matching PCB connected to the microstrip transmission
line through a thin, high-density, high frequency matrix interposer board (HCD, Sunnyvale, CA)
and contained a wideband transformer circuit (Coilcraft, Cary, Illinois) with a 1 : 9 impedance
ratio for each of the 64 array channels. The average measured impedance of the assembled array-
transmission line component was 408.4 Ω giving an average matched impedance of 45.3 Ω at the
output of the PCB where it connected through the multiplexer to the pulser/receivers. The higher
than expected impedance may be attributed to thinner than expected metallization of the
transmission line trace metal thickness or perhaps increased impedance through the connection
between the array bonding pads and transmission line.
This novel electrical interconnect solution was chosen for this high frequency miniaturized
array because of its capability to match the electrical impedance magnitude of the individual array
elements and for its compact design. A major design constraint is the size of the transmission line
since it needs to fit within the biopsy needle cylinder while routing all 64 traces that match the
array element electrical impedance. The 1 µm thick copper channel traces were patterned with a
very small pitch (25 µm) on a single layer across a span of 3 cm to validate this is a viable solution
to be housed within a biopsy needle.
65
Each copper trace is 10 µm wide with a 15 µm space between neighboring traces. The ground
layer on the bottom of glass was one uniform 2 µm thick copper layer. Future versions that employ
multilayer transmission line designs could reduce this width even further for smaller gauge biopsy
needles. A small rectangular cutout was laser cut into the end of the glass transmission line where
the channel traces terminated in 50 µm diameter bonding pads lined up on the edge of the cutout.
These bonding pads would match face to face with the bonding pads on the array composite surface
with the cutout providing the array aperture with an unobstructed view of the imaging
environment.
The glass transmission line, its substrate material properties, and PCB are shown in Figure
5-2.
Figure 5-2: Image of (a) single microstrip transmission line laser cut from wafer with copper traces shown
on top layer. Array bonding pads shown around rectangular cutout (b) match array element bonding pad
pattern. The PCB (c) with individual transformer circuits for electrical impedance matching for array channels
connects to the transmission line via a high density interposer board.
66
The glass transmission line substrate material properties and transmission line component
signal trace pattern are shown in Table 10 and Figure 5-3, respectively.
Table 10: APEX glass transmission line substrate material properties.
Parameter Value
Dielectric Constant 6.575
Coefficient of Thermal Expansion (glass sate) 10 ppm/K
Thermal conductivity (glass state) 1.5 W/mK
Electrical resistivity (glass state) ×10
12
Ω−𝑐𝑚
𝑇 𝑔 (glass transition) 452 °C
Young’s Modulus ~78 GPa
Figure 5-3: Transmission line (a) outline and (b) zoomed figure of transmission opening for array
aperture.
The signal propagation and cross-talk of the transmission line was simulated with the high
frequency structural simulator (HFSS) software (ANSYS, Canonsburg, PA), as shown Figure 5-4.
The simulation parameters of trace width, spacing and patterning as well as dielectric thickness
were considered in the simulation. The trace width was 10 µm with 15 µm spacing. In order to
67
calculate the maximum possible cross-talk between channels we input an 80 Vpp sine wave with
center frequency of 80 MHz into one input port and measured the signal at the output port of the
adjacent channel. We measured a maximum cross-talk of -25 dB. We are re-evaluating
transmission line connection options to improve this.
Figure 5-4: HFFS results for cross-talk measurement of transmission line channels with a color map
showing magnitude of signal in each of 5 adjacent channels.
One solution that could improve cross-talk reduction is a flex-circuit design with significantly
increased trace pitch (70 µm). However, solutions such as this would either require multilayer
transmission line fabrication with the glass substrate or the use of a multi-layered or curled flex
circuit, which is currently being investigated. These glass transmission lines were fabricated using
a proprietary material called APEX Glass in a 4” wafer process developed by 3D Glass Solutions
(3D Glass Solutions, Albuquerque, NM). The wafer fabrication steps are outlined as follows:
68
Table 11: Transmission Line Fabrication Steps
Step Fabrication Step
1 Photo-pattern of transmission line on the wafer with resist for the top side copper
2 Inspect wafer for defects in pattern; rework the photoresist if necessary
3 Evaporate copper on to wafer
4 Resist liftoff
5 Photo-pattern for the bump openings
6 Electroplate 2 µm of Ni/Au onto the wafer for the bonding pad openings
7 Sub-mount the wafer and thin the wafer to 200 µm with automatic lapping machine
8 Un-sub-mount wafer
9 Sputter blanket Cu film on backside of wafer [300ATiW + 1 µm Cu]
10 Final quality control inspection
Once the wafer has passed the final quality control inspection, the wafer is sent to a laser
cutting facility to cut out features in the glass wafer then cut out each of the 5 individual parts as
shown in Figure 5-5 below.
Figure 5-5: Image of (b) single transmission line laser cut from (a) wafer with copper traces shown on
top layer. Array bonding pads shown around (c) opening for array aperture.
The PCB connector board enabled all 64 channels of the array to connect to a multiplexer
used for synthetic aperture imaging experiments. The multiplexer was a modular multiplexer
69
system (NI PXIe 2593) built by National Instruments (NI, Austin, TX) which is a configurable
multiplexer platform that uses computer controlled mechanical switches to channel high voltage
(up to 800 Vpp) signals within its500 MHz bandwidth. Each PXIe 2593 module was configured
into a 2x8 multiplexer provide separate transmit and receive channels for 8 individual array
elements. We combined 8 of these modules and shorted 2 rows of 8 transmit and receives channels
separately, in effect serving as a 2x64 multiplexer unit. The assembled multiplexer system
performance specifications and diagram are shown in Table 12 and Figure 5-6, respectively. This
device enabled us to transmit and receive independently on all 64 elements of the ultrasound array.
This was necessary for the synthetic aperture imaging system we implemented for image
acquisition.
Table 12: Technical Specifications for MUX.
Parameter Value
Bandwidth (-3 dB) 500 MHz
Characteristic Impedance 50 Ω
Maximum Impulse Voltage 800 V
Insertion Loss <0.9 dB
Maximum Scan Rate 100 operations/second
70
Figure 5-6: NI MUX provides independent switching between transmit and receive channels.
5.2 Array Assembly Fabrication
The array composite, with array kerfs that are 6 m wide and 24 µm deep, was fabricated using
a Deep Reactive Ion Etching (DRIE) technique optimized by CTG (formerly HC Materials) for
piezoelectric materials. This etching technique was previously used successfully for an 80 MHz
ultrasound array and we modified it for our purpose (Liu et al. 2013). The array composite
fabrication steps are shown in Figure 5-12 and demonstrate how the process begins with bulk
PMN-PT single crystal material and ends with individually backed arrays. Fabrication of a
miniaturized high frequency array presented several key challenges, namely creating and filling
extremely small kerfs and the ability to handle the array parts throughout the fabrication process,
especially for lapping, given their extremely small size.
To achieve the small kerf sizes a deep reactive ion etching (DRIE) process was used to etch
the bulk material DRIE is an anisotropic etching process that is useful for creating microstructures
71
out of bulk materials with especially high aspect ratios and steep sidewall angles. DRIE is a
micromachining process originally developed for dynamic random access memory (DRAM)
fabrication applications. DRIE uses heated plasma to etch into bulk material using a
photolithographic developed pattern of a hard metal such as nickel to serve as the etching mask.
The high resolution of photolithography combined with the ability of DRIE to etch through
materials thicker than 20 µm makes it an ideal choice to create the small kerfs in high frequency
composite arrays.
A bulk piece of PMN-PT single crystal material made by CTG (formerly HC Materials) was
lapped and polished by hand to ensure coplanarity of the top and bottom surfaces until a thickness
of 1 mm was reached. Next, a nickel mask is patterned on to the bulk material so that the nickel is
covering all of the material except for the regions intended to be kerfs. Since the final array size is
only 1.5 mm x 1.5 mm and the piezoelectric bulk material is 15 cm x 15 cm, a 7x7 matrix of arrays
can be patterned during a single batch, enabling 49 arrays to be fabricated from each bulk material
sample. The reason 49 arrays were patterned on to the material is to account for the potentially low
yield of undamaged and useable arrays at the end of the fabrication process. Additionally, we were
interested in developing this batch fabrication method because it has much more potential as a
reliable fabrication approach than traditional array fabrication processes which allow for only 1
array to be fabricated by hand at one time.
After 49 array patterns have been laid down with the nickel mask, the sample is run through
the DRIE process to etch down 30 µm into the bulk material. Etching past the designed composite
thickness of 22 µm was necessary to allow for several microns of material to be removed in the
subsequent lapping fabrication steps. We chose to investigate 3 different composite designs for the
composite posts. Each design had the same element and kerf width in the azimuth direction but
72
each element was sub-diced in the elevation direction at pitches of 50 µm, 100 µm and 200 µm,
each with a sub-element kerf width of 5 µm. This sub-element design was implemented to mitigate
the risk of element electrodes peeling off the composite pillars due to strain. Under the advice of
our fabrication partner, CTG (formerly HC Materials), we etched sub-elements into each
composite pillar so each element had flexible epoxy inserted at regular intervals that could allow
for expansion or contraction of the piezoelectric materials without all of the strain being transferred
to the thin gold electrode patterned on the element. Figure 5-7 shows the etching pattern artwork
used to create the composites. To improve the array fabrication yield, small 30 µm x 30 µm posts
in a 1-3 composite arrangement were patterned between each array to break up the solid pieces of
piezoelectric material and prevent cracks from propagating from one array to another.
73
Figure 5-7: Array composite DRIE artwork for element, sub-elements and surrounding 1-3 composite
kerf etching patterns. Artwork for (a) all 49 elements with 2-2 composite pattern nested within 1-3 composite
pattern to prevent crystal fracture propagation in the sample. Artwork for (b) single array with 64-element
aperture size of 1.0 mm x 1.274 mm. Artwork for (c) 3 distinct individual sub-element lengths for each array
element to prevent electrode strain that could result in open channel circuits.
74
After DRIE processing, scanning electron micrographs of the composites were capture and
shown in Figure 5-8. These images show the effects of the etching process for the 3 different sub-
element designs. Visual inspection indicated that the 50 µm sub-element pitch design had the best
quality in terms of chipping of individual pillars.
Figure 5-8: Scanning Electron Microscope image of PMN-PT Material directly after DRIE Processing
with sub-element pitch length of (a) 50 µM, (b) 100 µM and (c) 200 µM.
In addition to the array element and sub-element kerfs, the region surrounding the 64 elements
was etched away as this would be the location of the individual element electrode bonding pads
and patterning these pads on top of active piezoelectric material would cause unwanted resonances
that would interfere with the operation of the active elements. Each composite had non-conductive
epoxy (EPOTEK-301, Epoxy Technology, Inc., Billerica, MA) wick in between the composite
75
pillars until it completely fills the kerf space between the elements, sub-elements and 1-3
composite region between arrays and covers the composite pillars. The bulk sample is then lapped
down to expose the top of the composite pillars. Once the array pillars are lapped to a fine matte
finish, Chrome/gold electrodes are sputtered on to the individual elements. Photoresist is spun
down on to the sample and the individual electrode patterns are exposed and developed on the
composite. Then, chrome/gold electrodes were then sputtered on to the composite, photoresist was
then removed revealing individual element electrodes (width = 14 µm) and bonding pads (diameter
= 50 µm) with the final element electrodes shown in Figure 5-9. The final composite had a kerf
width of 6 µm and a net piezoelectric volume fraction of 67%.
Figure 5-9: Micrograph of array element electrodes with bonding pads. Element bonding pads are
patterned above non-conductive epoxy to avoid exciting the piezoelectric material outside of the element.
Once the individual elements have been patterned on to the composite, the bulk material is
flipped over, wax bonded onto a carrier wafer and then lapped down through all of the material
until the backside of the composite pillars are exposed and the designed composite thickness of 22
µm is reached. Although 22 µm was the designed composite final thickness, the batch of
composites that was finally used for the array evaluation testing and imaging was slightly thicker
at 24 µm. The 24 µm thick composite was eventually chosen to connect to the electrical
interconnect because its electrodes and bonding pads were evaluated to be the best based on the
76
quality of the gold deposition uniformity. The electrical interconnection proved to be the most
challenging aspect of the array fabrication and therefore the array with non-ideal composite
thickness was chosen because it had the best chance of successful electrical interconnection. After
the composite is lapped down to the final thickness, a common ground chrome/gold electrode is
sputtered onto the entire composite surface as shown in Figure 5-10.
Figure 5-10: Finished composite samples a) Have a common ground electrode sputtered and b) Individual
elements, Sub-elements and 1-3 composite regions can be seen from the backside of the composites.
The next step is to add a backing to the array composites with a conductive epoxy material.
E-Solder 3022 is cast on the common ground electrode and centrifuged at 3000 RPM for 15
minutes, then post-cured overnight in a dry box at room temperature, followed by a 2-hour post-
cure at 40°C. The backed array composite sample is then removed from the carrier wafer, flipped
over and wax bonded onto a square glass carrier part. Each individual array module is then
separated from the sample with a TCAR864-1 dicing saw (Thermocarbon, Inc., Casselberry, FL)
by cutting the modules in a grid pattern as shown in Figure 5-11.
77
Figure 5-11: Individual arrays diced from (a) grid of 49 arrays to separate them (b), yielding individual
arrays (c) 1.5 mm x 1.5 mm in size.
The acoustic stack layer material and acoustic properties for each array module are listed in
Table 13.
Table 13: Acoustic Stack Properties
Layer Material Thickness Density
(kg/m3)
Vlong2
(m/s)
Vshear2
(m/s)
Acoustic Impedance
(MRayl)
Matching Parylene
a
7 µm 1100 2350 1662 2.6
Piezoelectric PMN-PT
b
24 µm 8100 4608 3258 37.0
Kerf Filler Epo-Tek 301
c
24 µm 1150 2650 1270 3.0
Backing E-Solder 3022
d
1 mm 3200 1850 1308 5.9
a
Parylene-C: ONDA Corporation, Sunnyvale, CA
b
PMN-PT: CTG (formerly HC Materials), Santa Barbara, CA
c
Epo-Tek 301: Epoxy Tech., Billerica, MA (Wang et al. 1999)
d
E-solder: Von Roll USA, New Haven, CT (Wang et al. 1999)
78
The array fabrication process is summarized in Figure 5-12 below.
Figure 5-12: Array fabrication steps for array composite beginning with bulk single crystal PMN-PT
material and ending with individual miniaturized arrays.
79
One significant issue arose during the array fabrication process involving the composite film
bonding to the carrier wafer. During the heating process of the chrome/gold electrode deposition,
small air bubbles between the thin composite and the carrier silicon wafer formed, separating film
from the carrier and warping it irreversibly. When these air bubbles formed underneath the
patterned arrays, as shown in Figure 5-13, they became un-usable. This was the primary reason for
the loss of about 25%-50% of the arrays on each sample of 49 arrays. However, because of the
fact we patterned 49 arrays per sample there were at least several useable arrays from each batch.
Figure 5-13: Image of composite film with visible bubbles forming between silicon carrier wafer and
composite material.
5.2.1 Conductive Microsphere Electrical Interconnect Solution
Due to the small size of each array module (1.5 mm x 1.5 mm x 1 mm) a novel interconnect
solution between all 64 transmission line and array element channels was developed. Various,
standardized interconnect methods including wire bonding and flip chip bonding methods were
unsuitable because of the demanding size, pressure, temperature and handling constraints this
miniaturized array presented. The thin, fragile, epoxy-filled composite prevented any processes
80
that required the use of elevated temperatures or pressure. Elevated temperatures would warp the
composite which would develop gaps between elements and surrounding epoxy or cause breaks in
the thin gold element electrodes due to expansion and contraction of the epoxy and applying
pressure during bonding would break the fragile glass transmission line. Most of these high density
interconnect solutions were developed for the semiconductor industry and even the vendors who
specialize in custom interconnect solutions were unable to meet the requirements we had.
Therefore, we developed a hybrid interconnect solution that combined components from
anisotropic film connectors and conductive epoxies with flip chip and solder bumping bonding
methods. The solution was to use conductive microspheres and conductive epoxy to make
connections between all 64 channels in a fashion similar to standard solder-based flip chip
bonding.
In this new process, each array element bonding pad has a small dome-shaped conductive
epoxy bead bumped onto it. Each of these bumps gives the bonding pad a raised profile so that its
connection to the mirrored bonding pad on the transmission line can be made. The conductive
epoxy was 2 - 3 µm diameter silver particles (Sigma Aldrich, St. Louis, MO) mixed with Insulcast
501 epoxy in a 2:1 ratio by weight. Each epoxy bump was 25 µm - 40 µm in height and was applied
with a 5 µm diameter probe tip mounted on a 3-axis manual positioner stage. Next, one conductive
microsphere with a diameter of 28 µm - 32 µm (Cospheric, Santa Barbara, CA) was placed on
each conductive epoxy bump. These conductive spheres are manufactured in bulk and designed to
be integrated into anisotropic conductive films (ACF) products that allow circuit lines to be
connected in the vertical direction, but are spaced in the lateral direction to be electrically
insulating along the plane of the adhesive. Each conductive microsphere has a solid soda glass
core and 75 nm thick pure silver outer surface coating. We have repurposed these conductive
81
microspheres for use as a new type of vertical electrical interconnect scheme similar to flip chip
bonding which does not require pressure or heat to be applied for permanent plane-to-plane circuit
bonding to occur. The epoxy provides both the mechanical support to the conductive microsphere
to hold it in place while also providing a conductive pathway to the bonding pad since it is loaded
with silver particles.
The conductive epoxy was allowed to cure overnight to secure the conductive microspheres
in place. Next, another bump of conductive epoxy is placed on top of each conductive sphere now
bonded to each bonding pad of the array in the same fashion described above. Finally, a custom 3-
axis manual positioner stage is used to align the bonding pads on both the array and transmission
line and lower the transmission line until its bonding pads are in contact with the conductive
microspheres. Now each bonding pad is connected via conductive microspheres with the
conductive epoxy connecting the bonding pads to the conductive microspheres. This connection
technique enabled the array to be connected to the transmission line at room temperature, without
any pressure being applied to the delicate array module and glass transmission and without
expanding the footprint of the connection, which is critical to maintaining a compact connection
scheme for this array. This connection scheme is shown below in Figure 5-14. The array aperture
is exposed to the imaging environment through a small rectangular cutout in the glass transmission
line. The ground connection between the glass transmission line and the conductive backing of the
array is made with a 1 mm wide flexible strip of silver and is secured using E-Solder 3022
conductive epoxy.
82
Figure 5-14: Cross-sectional view of interconnect scheme between array and glass transmission line.
Silver-coated glass spheres bonded by conductive epoxy were used to connect bonding pads on both the array
module and transmission line.
83
Array Testing and Performance
6.1 Array Testing Experimental Design
In addition to image formation there are several array transducer testing experiments that can
provide insight into its performance. Array element electrical impedance and pulse/echo
performance were measured prior to performing 64-channel synthetic aperture imaging. Electrical
impedance was measured with an Agilent E4991A RF Impedance/Material Analyzer (Agilent
Technologies, Santa Clara, CA) and both magnitude and phase angle were recorded over the
frequency range of the transducer pass-band. Pulse/echo testing is useful to determine the effective
bandwidth of array transducers as well as sensitivity, pulse length and focal depth. This test was
performed in de-ionized water with a polished quartz reflector as the target at a distance of 2.2
mm. The pulser used for pulse/echo testing was a Panametrics 5900PR pulser/receiver
(Panametrics, Inc., Waltham, MA) which emitted a unipolar 100 Vpp pulse. To receive echo signals
a bandpass filter (10-100 MHz) and gain of 26 dB was applied before analog signals were digitized
using a GaGe digitizer (Dynamic Signals, LLC, Lockport, IL) with a 1 GHz sampling rate.
For each pulse/echo time domain signal the -6 dB bandwidth was calculated using a fast
fourier transform (FFT) where the lower and upper bandwidth edges were determined by the
frequencies where the power spectrum was equal to -6 dB relative to the maximum value. The
center frequency was taken as the midway point between the lower and upper limits of the -6 dB
bandwidth. Echo amplitude was recorded for sensitivity and insertion loss comparisons. The -6 dB
signal pulse length was determined by measuring the time between the first and last points where
the signal was -6 dB relative to the maximum echo signal value.
84
The level of acoustical and electrical separation between neighboring array elements was
determined by measuring cross talk between a representative element and its 3 adjacent elements.
To perform this test the array was submerged in a degassed and deionized water bath. A Tektronix
arbitrary waveform generator (Tektronix, Beaverton, OR) generated a signal burst to excite one
element in the array with the applied voltage measured as a reference to the measured signal from
adjacent elements. The signal voltage at the 3 neighboring elements is measured and compared to
the reference voltage at discrete frequencies over the bandwidth of the array.
Insertion loss was measured by exciting a single element with a 60 MHz single-cycle signal,
and receiving the echo off of a polished quartz target located at the focal point of the transducer.
The measured amplitude value was corrected for loss due to attenuation in water and from
reflection off a polished quartz target (Lockwood, Turnbull, Foster 1994).
The array was used to image a single 20 µm tungsten wire target (California Fine Wire
Company, Grover Beach, CA) in a de-ionized water bath. For the imaging test an AVTech AVB2-
C-USCC Monocycle Pulse Generator was used to generate a 70 MHz single cycle pulse with an
amplitude of 160 Vpp. Line-spread functions of the wire echo signal from the center of the wire
were evaluated to determine the axial and lateral resolution of the array. The experimental setup
for the wire target imaging is shown in Figure 6-1 below.
85
Figure 6-1: Wire targeting imaging experimental setup.
6.2 Array Imaging Experimental Design
To perform imaging experiments with this array we built a synthetic aperture imaging system
where a multiplexer enables a single channel pulser and single channel receiver to collect
pulse/echo signals from any transmit-receive element pair in the 64 element array; the system
block diagram is shown in Figure 6-2.
86
Figure 6-2: Synthetic aperture imaging system design for high frequency array imaging.
A LabVIEW (NI, Austin, TX) program running on a PC controls a multiplexer that regulates
which array elements are used for transmit and receive channels during the image data capture
sequence. A single channel transmit/pulser and receiver were independently connected to the
multiplexer. The 2x64 multiplexer is a modular NI PXIe 2593 multiplexer system, a configurable
multiplexer platform that uses computer controlled mechanical switches to channel high voltage
signals within the 500 MHz bandwidth. Each PXIe 2593 module was configured into a 2x8
multiplexer to provide separate transmit and receive channels for 8 individual array elements. We
combined 8 of these modules and connected 2 separate rows of 8 transmit and 8 receive channels,
in effect creating a 2x64 multiplexer unit. This enabled the system to transmit and receive
independently on all 64 elements of the ultrasound array. This is necessary for the synthetic
aperture imaging system we implemented for image data acquisition.
This imaging system has the capability to transmit and receive on independent channels and
for each image capture and echo data was captured from each of the 64 x 64 = 4096
transmit/receive pairs. For each echo signal, each individual element was excited using a high
voltage pulser and echo signals were received with an individual element, resulting in 4096
87
transmit/receive signals. These signals were then processed using a synthetic aperture image
reconstruction technique demonstrated by Trots et al. (Trots, Nowicki, Lewandowski 2009). This
reconstruction method calculates the delay of each transmit/receive pair to obtain the signal at each
pixel in the image field. To calculate the delay between transmit and receive elements in the array
the expression is:
𝜏 𝑚 ,𝑛 =𝜏 𝑚 +𝜏 𝑛 ,
(6-1)
Where 𝜏 𝑚 and the 𝜏 𝑛 is the path delay from the transmit and receive element, respectively, to
the focal point in the imaging field. Where 𝜏 𝑚 and the 𝜏 𝑛 are described by:
𝜏 𝑚 =
1
𝑐 √𝑥 𝑚 2
+𝑟 2
−2𝑥 𝑚 𝑟 sin𝜃 ,
(6-2)
𝜏 𝑛 =
1
𝑐 √𝑥 𝑛 2
+𝑟 2
−2𝑥 𝑛 𝑟 sin𝜃 ,
(6-3)
where the point in the image field is given by (r, 𝜃 ) where r is the distance between the center of
the array and the point and 𝜃 is the angle of line from the center of the array this point. Furthermore,
𝑥 𝑚 and 𝑥 𝑛 are the positions of the m-th and n-th elements, respectively.
𝐴 (𝑡 )= ∑ ∑𝑦 𝑚 ,𝑛 (𝑡 −𝜏 𝑚 ,𝑛 )
𝑁 −1
𝑛 =0
𝑁 −1
𝑚 =0
(6-4)
88
where 𝜏 𝑚 ,𝑛 is the delay for the (m, n) transmit and receive combination as given in equation (6-4)
and 𝑦 𝑚 ,𝑛 (𝑡 ) is the received echo signal (Trots, Nowicki, Lewandowski 2009). A diagram of the
transmit and receive delays calculated in (2) is shown in Figure 6-3 and illustrates how each
independent transmit receive pair for each individual pixel position can be simply calculated.
Figure 6-3: Diagram of path length between transmit and receive element for synthetic aperture
imaging.
A MATLAB script) was written to calculate the summed echo signal for each transmit and
receive pair for every pixel position in the imaging field. Summed echo signals were then
processed using envelope detection and log compressed and displayed in a grayscale B-Mode
image.
89
6.2.1 Synthetic Aperture Imaging System Software
The custom built synthetic aperture imaging system utilizes frontend electronic equipment
including a high voltage pulser, multiplexer, pre-amplifier/receiver and digitizer. A custom built
LabVIEW software program controls the system, enabling the user to test different arrays designs
and image data capture techniques. The system captures individual pulse/echo signals in sequence
to generate imaging data used during the synthetic aperture image reconstruction process.
Individual pulse/echo waveforms are captured during the following process. A pulser generates a
single high voltage pulse and sends it to the transmit port of the multiplexer and simultaneously
generates a trigger signal and sends it to the digitizer so data recording begins. The transmit pulse
is routed to a single array element through the MUX and the echo is received by a single, discrete
array element. The received echo signal is routed through the MUX out the receive port, then
through the pre-amplifier and finally to the digitizer RF signal input port. The digitizer, sends the
digitized data to the PC running the LabVIEW virtual instrument software program.
The LabVIEW program controls all aspects of the synthetic aperture imaging process
including RF data acquisition, MUX channel switching, RF data processing and image
reconstruction. The program has a graphical user interface (GUI), as shown in Figure 6-4. This
GUI allows the user to create a custom imaging system profile for both the array and type of image
scanning. Thus, it will serve as a flexible array testing platform for high frequency phase and linear
arrays in the future.
90
Figure 6-4: Graphical user interface of synthetic aperture imaging system which allows users to
determine the type of array and electronic scanning method used during the imaging process.
6.2.2 GUI User Inputs and Controls
The array characteristics that the user can define are the number of elements in the array and
the element-to-element pitch. These two values are utilized during both the data capture and image
reconstruction process, where they are used to determine the delay path lengths for each of the
pixels in the reconstructed image. The imaging characteristics that the user can define are the start
and ending depth of the image field as well as the scanning list which defines the order in which
the transmit and receive elements are selected with the multiplexer. The GUI includes options to
select from 3 scan settings as outlined in Table 14 below.
91
Table 14: Synthetic Aperture Scan Settings
Scan Setting Description
Fully Sampled A pulse/echo waveform is captured from each transmit-receive
element pair in the array. Each waveform is used for the synthetic
aperture image reconstruction process producing a B-mode image.
Same Tx & Rx A pulse/echo waveform is captured for each array element with a
single element acting as both the transmit and receive transducer.
These echo waveforms are treated as scanlines and arranged
together to form a B-mode image.
Manual (Read Scan List) This option enables the user to define any scan list for the array. The
multiplexer can independently select transmit and receive elements,
making image data acquisition flexible. Users define all the
sequential transmit-receive element pairs in an excel file template.
This file is then uploaded to the LabVIEW program and used to
control the scanning process.
Once the array characteristics and scan settings are defined the user can choose to save the RF
data from each pulse/echo waveform by clicking the “Save RF Data” button. This saves the RF
data of each waveform to a .txt file in a folder defined by the user. This data can then be accessed
offline for image processing such as quantitative backscatter analysis. The user can then select the
option to display the reconstructed B-mode image by clicking the “Display Image” button. Then
the user clicks the “Go Imaging” button to initiative the image data capture sequence and image
reconstruction process. The program ends once all RF data is saved and the B-mode image is
displayed on the GUI. The synthetic aperture imaging software program block diagram is shown
in Figure 6-5 and outlines the major operations performed during its execution.
92
Figure 6-5: Synthetic aperture imaging system software control block diagram.
The program begins by first initializing the digitizer so it has the appropriate sampling rate,
voltage amplitude range and data file length. The multiplexer is then initialized so all switches are
open, meaning no transmit or receive element paths are connected. The program then loads the
appropriate scan list excel file and begins reading the list in order to determine the switching
commands for the multiplexer. With each index (1, 2, 3, … , n) in the scan list the appropriate
switches are closed in the multiplexer to form a discrete transmit and receive path for the high
voltage pulse to travel to the transmit element and the low voltage echo signal to travel from the
receive element. The multiplexer progresses from one transmit-receive pair to the next once the
digitizer completes the data transfer process from its onboard memory to the PC memory.
The echo waveform data for each transmit-receive pair saved is appended to a 2D array that
contains the echo data from all transmit-receive pairs in the scan sequence. This 2D array is the
data source used during the image reconstruction process and is the data exported when the user
selects the “Save RF Data” option before the image data capture process begins. Once all echo
waveforms have been captured the multiplexer disconnects all switches and ends the scanning
93
process. The program then moves on to the image reconstruction phase where it runs a MATLAB
script from within LabVIEW designed to read the 2D array of echo waveform data and output the
final B-mode image in the LabVIEW GUI. The MATLAB script first reads all the RF data from
the 2D array, applies bandpass filtering and envelope detection, then performs the image
reconstruction pixel by pixel according to the synthetic aperture image reconstruction technique
described by Trots et al. (Trots, Nowicki, Lewandowski 2009) and outlined here in section 6.2.
The final image is displayed in a log compressed gray scale image with the image depth and
horizontal position in millimeters displayed on the vertical and horizontal axes, respectively. While
this image reconstruction method is used because it is well suited to the data acquisition process
this system is capable of, a different image reconstruction method could be implemented in
MATALB and called during the execution of the LabVIEW program, providing the user flexibility
in how the system both captures data and reconstructs images from that data.
6.3 Array Testing Results
The results for individual array element testing are given in Table 15 below. The array had no
shorted or open elements, however 7 elements had very low sensitivity with insertion loss values
over -60 dB. Figure 6-6 shows the electrical impedance magnitude and phase angle values for one
selected array element. The measurement was made at the point of the transmission line
interconnect pad where it connected to the PCB, which took into account the impedance matching
microstrip transmission line of each individual element channel. While the trace lengths were
designed in the transmission line and PCB to be as similar as possible and the trace metallization
designed to be identical, the variation in lengths or deviations in the trace metal thickness may
94
account for the variation in impedance magnitude measurements since there was a slight trend
towards lower impedance as the element number increased and also some random variable
between even adjacent elements.
Table 15: Average Measured Properties of 64-Element Array
Property Value
Number of Elements 64
Center frequency 59.1 MHz
Bandwidth (-6 dB) 29.4%
Sensitivity 703 mV
Insertion loss -41.0 dB
Electrical impedance magnitude* 408.4 Ω
Electrical impedance phase angle* -70.0°
Focal point 2.2 mm
*
Electrical impedance magnitude and phase angle measurements were made at 60 MHz.
Figure 6-6: Measured electrical impedance magnitude (solid line) and phase angle (dashed line) for a
typical array element.
-85
-83
-81
-79
-77
-75
-73
-71
350
370
390
410
430
450
470
50 52 54 56 58 60 62 64 66 68 70
Phase (Degree)
Magnitude (Ω)
Frequency (MHz)
95
The average array element center frequency and -6 dB bandwidth was 59.1 MHz and 29.6%,
respectively. Pulse/echo measurement results for a typical single element are shown in Figure 6-7.
Figure 6-7: Pulse/Echo measured results for a typical array element including the (a) echo and (b) FFT
magnitude plot.
The selected element #60 had a peak-to-peak voltage of 704 mV, center frequency of 59.6
MHz, bandwidth of 34.7% and -6 dB pulse length of 48 ns. Element #60 was chosen as a typical,
representative element since its performance characteristics were closest to the average of all 64
elements. There was a significant variation in the sensitivity of the elements with 6 elements
exhibiting sensitivity between 45 mV and 95 mV. However, the element center frequencies were
more closely grouped with the minimum and maximum center frequency of 55.1 MHz and 63.7
MHz, respectively. The larger variation in sensitivity values versus center frequency suggests that
variations in the electrical interconnects and traces account for the difference in performance
between elements rather than differences in array element geometry because of the composite
fabrication process. This array employed new array fabrication and electrical interconnect methods
and it is likely that small defects in the metallization of the element electrodes reduced the effective
element size or caused increased resistance in the signal path, thus decreasing the element
96
sensitivity. Another possible issue is that defects in the array-transmission line interconnect or
transmission line itself that formed during array performance characterization experiments
significantly increased the channel resistance of the elements with low sensitivity.
Overall performance of the array uniformity in terms of center frequency, bandwidth and
electric impedance magnitude and phase angle values is illustrated in Figure 6-8 and Figure 6-9,
respectively.
Figure 6-8: Center frequency and bandwidth values for each element in array.
0
10
20
30
40
50
60
70
0
10
20
30
40
50
60
70
0 8 16 24 32 40 48 56 64
Bandwidth (%)
Center Frequency (MHz)
Element Number
Center Frequency
Bandwidth (%)
97
Figure 6-9: Impedance magnitude and phase angle values for each element in array.
The crosstalk measurements taken were consistent across the effective bandwidth of the array
and the lower cross talk for elements farther away from the reference element is as expected. The
measured crosstalk at the center frequency for the 1
st
, 2
nd
and 3
rd
nearest elements was -23.7 dB, -
28.8 dB, -35.9 dB, respectively. The measured crosstalk values are shown in Figure 6-10. The
relatively high crosstalk level (> -30 dB) in the first 1
st
and 2
nd
adjacent elements was likely due
acoustic or mechanical cross-coupling between elements in the 2-2 composite through the narrow
(6 µm wide) epoxy-filled kerfs. Another possible source of the crosstalk is likely the electrical
cross-coupling between both the element electrodes and the high-impedance traces in the
transmission line. In future versions of this miniaturized high frequency array, cross talk may be
reduced by adding particles to the kerf filling epoxy to increase attenuation or by improving the
electrical isolation of the high impedance traces on the transmission line by increasing the spacing
between traces and using multilayer transmission lines with ground layers separating trace layers.
-90
-80
-70
-60
-50
-40
-30
-20
-10
0
0
50
100
150
200
250
300
350
400
450
500
0 8 16 24 32 40 48 56 64
Phase (Degrees)
Magnitude (Ohms)
Element Number
Magnitude
Phase
98
Figure 6-10: Measured crosstalk values for 3 nearest neighboring elements in array.
The measured insertion loss for a typical array element was -45.5 dB and -41.0 dB before and
after correction for attenuation using the method described by Lockwood et al. (Lockwood,
Turnbull, Foster 1994). This value is higher than other high frequency ultrasound arrays built using
micromachining fabrication methods, however, the element size in this array was 6.9 times smaller
than elements in the 2 comparable arrays previously tested (Zhou et al. 2010), (Liu et al. 2012).
The significantly smaller array size coupled with the single matching layer used in this array
explains why the insertion loss measured was higher than what was observed in these other arrays.
Improving both the electrical interconnect method and transmission line fabrication process as
well as adding a second matching layer in the next generation of this array will help improve
sensitivity and decrease insertion loss.
-45
-40
-35
-30
-25
-20
45 50 55 60 65 70 75
Crosstalk (dB)
Frequency (MHz)
Adjacent Element
2 Elements Away
3 Elements Away
99
An image of the 20 µm wire test phantom is shown in Figure 6-11 using a linear gray scale
and 25 dB dynamic range. Image reconstruction did not include any thresholding or apodization.
The line spread functions for the axial and lateral resolution plots for the center of the wire are
shown in Figure 6-12. The measured full-width half-maximum (FWHM) axial and lateral
resolutions were 33.2 µm and 115.6 um, respectively. Faint artifacts are observed to the left and
directly below the wire position. Using apodization could reduce the sidelobes but this would come
at the expense of increasing the main lobe width (Frazier and Brien 1998). Additionally, it is
possible that the artifacts are due to in part by reflections off the thin sidewall of the rectangular
glass transmission line cutout that frames the array aperture.
Figure 6-11: Reconstructed synthetic aperture image of a single 20 µm wire target with no thresholding
or apodization. The image was displayed with 25 dB dynamic range and mapped on a linear gray scale.
This 1-wire phantom target image with 25 dB dynamic range demonstrates that this
miniaturized high frequency linear array is functional and the fabrication and electrical
interconnect methods used to produce this device are feasible.
100
Figure 6-12: Axial (a) and lateral (b) line plots for the center of the wire phantom.
While the axial resolution is an improvement over previously reported high frequency linear
arrays, the lateral resolution is not and this lower than desired lateral resolution may be improved
by improving the alignment mechanism for imaging experiments. The extremely small array
aperture size (1 mm x 1.2 mm) makes the fine adjustments and alignments of any phantom target
difficult. Even a small deviation in the angle of the wire target from parallel alignment with the
array elements elevation axis would negatively affect the lateral resolution measurement of the
array.
101
Clinical Study Design and Results
7.1 Clinical Study Experimental Design
To validate the utility of a miniaturized high frequency imaging array, we will demonstrate
that high-frequency ultrasound imaging enables differentiation of cancerous tissue from normal
tissue by analyzing tissue obtained during a clinical breast tissue study. Our clinical study’s
endpoints are designed to assess the effectiveness of high-frequency ultrasound imaging by
comparing our newly acquired high-resolution ultrasound images of breast biopsy tissue with
digital pathology images, the gold standard for breast cancer diagnosis. The microstructures within
the breast tissue that are relevant to image-guided breast biopsy and breast cancer diagnosis in
general are microcalcifications and necrotic tissue. The presence of microcalcifications can
suggest early breast carcinoma such as Ductal Carcinoma In-situ (DCIS) and thus serve as a very
useful feature to visualize during the biopsy guidance procedure. Providing enhanced
microcalcification visualization can enable physicians to better target the tissue he or she will
sample with the biopsy needle and since sampling error is the primary cause of missed diagnoses,
or false-negatives, of breast cancer, improving lesion feature identification is a promising solution
to improve the sensitivity of breast cancer biopsy procedures. Additionally, necrotic breast tissue
can occasionally mimic malignancy and thus more detailed imaging of necrotic fat tissue could
help physicians avoid taking a biopsy of this tissue since it can be misleading and lead to false-
positive diagnoses (Soo, Baker, Rosen 2003; Toshiba America Medical Systems 2008; Velez,
Earnest, Staren 2000; Taboada et al. 2009). Figure 7-1 illustrates the clinical study design for
imaging breast biopsy core tissue samples.
102
Figure 7-1: Outline of clinical study to examine efficacy of high frequency ultrasound imaging in
identifying cancer in breast biopsy core samples.
We obtained Institutional Review Board (IRB approval for this study and have collected and
imaged tissue samples from our two institutions, the Norris Cancer Hospital, Keck Hospital of
University of Southern California (USC), and the Los Angeles County-University of Southern
California (LAC-USC) Medical Center. The study design is based on prior efforts to determine
how well radiologists can classify tissue using ultrasound images obtained with clinical ultrasound
machines and their correlation with known diagnoses determined by histopathological assessment
(Tresserra et al. 1999; Chaudhari et al. 2000).
Fresh biopsy specimens were obtained at both Norris Cancer Hospital and LAC+USC Medical
Center, placed in plastic containers with phosphate buffered saline (PBS) solution and transported
103
immediately to the ultrasound imaging lab on University Park Campus of USC. To standardize the
plane of imaging for ultrasound image acquisition and the plane of sectioning for pathological
assessment, biopsy specimens were encased in 3% agar gel (Fisher Scientific, Waltham, MA)
within a petri dish with the top portions exposed for ultrasound imaging. The agar gel served to
keep the tissue specimen in a rigid position so that the pathology sectioning can be performed
along parallel planes to the ultrasound image acquisition.
Ultrasound images were generated using a single element transducer with a press-focused,
2.25 mm diameter aperture and a 3.4 mm focal depth, giving an f-number of 1.5. The transducer
had a center frequency and -6 dB bandwidth of 74 MHz and 27 MHz, respectively. The
transducer’s piezoelectric material was PMN-PT and was backed by E-Solder 3022 conductive
epoxy (Von Roll USA, New Haven, CT). The first matching layer was made of 2 - 3 µm diameter
silver particles (Sigma Aldrich, St. Louis, MO) mixed with Insulcast 501 epoxy (ITW Polymers
Coatings North America, Montgomeryville, PA). The second matching layer was vapor deposited
parylene. The ground connection was made via a chrome/gold electrode plated across the front
surface of the 2-3 µm silver epoxy matching layer to the brass transducer housing. An SMA
connector threaded into the back of the brass housing completing the ground connection. The
signal connection though the center signal port of the SMA connector shorted to the conductive
silver epoxy backing of the transducer. This SMA connector then connected to the pulser/receiver
via a coaxial cable and was secured in a fixture attached to the 3-axis motorized positioner stage.
A Panametrics 5900PR pulser/receiver was used to send high voltage pulses to and receive
the low voltage echo signals from the single element transducer. The receiver applied a 26 dB gain
to the received echo signals. A digitizer with 1 GHz sampling frequency captured the RF echo
signal and transferred the data to a PC. A single pulse/echo signal formed each scanline during the
104
image data capture process and scanlines were captured at 10 µm intervals and the scanline data
set was reconstructed to form a log compressed gray scale B-mode image. Once each B-mode
image was captured, the transducer was translated 100 µm perpendicular to the imaging plane and
the image process was repeated until images were captured across the entire width of the biopsy
specimen which ranged from 2 - 4 mm. Figure 7-2 illustrates the ultrasound image acquisition and
data analysis setup for the ultrasound imaging of breast biopsy core tissue and Figure 7-3 shows
an overhead view of the biopsy specimen and the scanning pattern of the transducer.
Figure 7-2: Ultrasound image data acquisition and backscatter analysis design for high frequency
ultrasound imaging of ex vivo breast core biopsy tissue specimens using a single element transducer and
motorized scanning imaging system.
105
Figure 7-3: Biopsy specimens were secured in an agar gel block to maintain their orientation during
ultrasound and pathological imaging.
After the images were collected, a motorized stage with a razor blade attachment cuts smooth,
vertical walls into the agar gel. The agar gel with encased biopsy specimen was then submerged
in 10% neutral buffered formalin (pH 6.8-7.2 @ 25) and transported to the Keck Hospital
Pathology department where the gel-encased specimens were processed for pathology which
included paraffin embedding, sectioning at 50 µm, and plating the tissue on glass slides for
hematoxylin and eosin (H&E) staining. The pathology slides were then imaged with a Leica
SCN400 digital pathology slide scanner (Leica Microsystems, Inc., Buffalo Grove, IL) and
uploaded to the Leica digital image hub (DIH), an image storage database. The pathology images
were then accessible for review and comparison was made to the ultrasound images.
The primary aim of this study is to confirm the presence of microcalcifications within breast
tissue using high frequency ultrasound imaging. To provide this confirmation each biopsy
specimen underwent histopathological sectioning so that H&E pathology images could serve as a
reference for the ultrasound images. Both high frequency ultrasound and pathology image sets
106
were generated in parallel imaging planes using a specimen handling and imaging technique
described in the following sections. By generating ultrasound and pathology image sets with
parallel image planes, features identified in the ultrasound image slices could be confirmed by
matching them with those in the pathology image slices.
The method used to determine matches between ultrasound and pathology images was
possible because of the orientation of both the ultrasound and pathology image planes as well as
the presence of clear landmark features within the images (microcalcifications) which confirmed
that the ultrasound and pathology image slices were correctly matched. The image sets were
matched in the following way. First, the ultrasound and pathology image sets were arranged side
by side. Next, landmark features, such as microcalcifications were identified on the pathology
images. Since the microcalcifications are 3-dimensional objects, the pathology image slice where
they first become visualized and when they no longer appear mark the extent of the width of the
structure. The position of the structure within the biopsy specimen is noted in relation to its top
and bottom surfaces and left and right ends. Once the width and position within the biopsy
specimen of this microcalcification is known, the same structure is searched for in the ultrasound
image set. Since the image sets are in parallel and the position and approximate with of the
structure within the biopsy specimen is known, the microcalcifications can be identified in the high
frequency ultrasound image. Finally, the ultrasound and pathology image slices that both depict
the microcalcification are arranged next to each other and displayed.
Pathology sections were taken from a plane parallel to that of the ultrasound images at 50 µm
intervals instead of 100 µm intervals in order to provide more pathology sections than ultrasound
image slices. This oversampling of pathology image slices was implemented to account for the
slight shrinkage of the specimen as it undergoes histopathological processing. In this study, it was
107
not necessary to have exact 1:1 matching between ultrasound and pathology image slices because
the features we were looking for, microcalcifications, were sufficiently wide that they spanned
several ultrasound and pathology image slices. So the identification of these microcalcifications
could be made by comparing several, adjacent ultrasound-pathology image pairs that all showed
the microcalcification structure. Thus, oversampling the specimen with ultrasound and pathology
image slices accounted for the fact that the ultrasound-pathology image set correspondence would
not be exact.
7.2 Image Processing Methods
In addition to standard B-mode displays of ultrasound imaging data, we sought to improve
tissue characterization and microstructure identification within the breast biopsy tissue samples by
utilizing two ultrasound image processing methods. The first method we applied was the
Nakagami filtering technique, a method that calculates a parameter from the shape of the
backscatter signal distribution for a given windowed region in the imaging field. This parameter
is calculated for each point in the imaging field and the result is a 2D parameter map that provides
insight into the type of tissues and structures within a tissue sample. The second method
implemented was quantitative backscatter analysis, which calculates 3 distinct coefficients from
the backscatter signal of a particular region in the imaging field. The coefficients can then be
calculated for each point in the imaging field in the same way the Nakagami parameters are,
producing a 2D coefficient map for each of the 3 coefficients. These maps also provide useful
tissue and microstructure characterization information. The methods for implementing both of
these image processing techniques as well as their results are described in the following sections.
108
7.2.1 Nakagami Filtering Methods
The first tissue characterization algorithms implemented to improve provide feature
recognition for microcalcifications with the breast biopsy samples was the Nakagami filter
technique. This tissue characterization technique works by calculating the Nakagami parameter of
tissue, which is generated from the ultrasound backscatter envelope of the B-Mode image. The
probability density function (pdf) of the ultrasonic backscatter envelope R is given by (Tsui et al.
2008):
𝑓 (𝑟 )=
2𝑚 𝑚 𝑟 2𝑚 −1
Γ(𝑚 )Ω
𝑚 𝑒𝑥𝑝 (−
𝑚 Ω
𝑟 2
)𝑈 (𝑟 )
(7-1)
where Γ(⋅) and U(⋅) are the gamma function and the unit step function, respectively. E(⋅) denotes
the statistical mean, Ω is the scaling parameter and the m is the Nakagami parameterwhich
corresponds to the Nakagami distribution. The scaling parameter can be obtained from Ω=𝐸 (𝑅 2
)
and the Nakagami parameter is given by
𝑚 =
[𝐸 (𝑅 2
)]
2
𝐸 [𝑅 2
−𝐸 (𝑅 2
)]
2
(7-2)
The scaling parameter Ω provides structural information of the tissue and the Nakagami parameter
m is useful in characterizing tissue type and is a shape parameter determined by the pdf of the
backscatter envelope (Tsui et al. 2008).
109
7.2.2 Backscatter Analysis Methods
Backscatter analysis is a method of processing the backscattered acoustic energy during
ultrasonic imaging to determine characteristics of biological tissue or structures and has been
studied extensively (Campbell and Waag 1984; Lizzi et al. 1983; O’Donnell and Miller 1981;
Shung and Thieme 1992). Various experimental methods have been developed to measure the
backscatter characteristics of various materials, objects and biological tissue types (Lee et al.
2011). Thomas III et al. demonstrated how the integrated backscatter (IB) coefficient, a valuable
backscatter characteristic, could be quantified in myocardium (Thomas III et al. 1989). Backscatter
analysis has more recently been applied to ultrasonic breast tissue imaging where it serves as a
quantitative analysis technique useful in differentiating tissue types (Nam, Zagzebski, Hall 2013).
One microstructure of interest in breast tissue are microcalcifications and statistical approaches to
resolve and identify these in tissue using ultrasound imaging have been explored (Shankar 2013).
The ultrasound backscatter analysis of breast biopsy specimens presented here serves to
demonstrate the potential of this method in differentiating breast microstructures and tissue types
for the purpose of improving breast biopsy guidance and tissue acquisition accuracy.
7.2.3 Quantitative Ultrasound Analysis
Quantitative estimation of backscatter properties in human tissue is possible by measuring the
ratio between the magnitudes of Radio Frequency (RF) echo signals from a sample volume of
interest in a tissue and a flat reference reflector of known reflectivity. Although the ultrasound
backscatter coefficient could be theoretically defined and precisely measured by using a simple
sample target such as a spherical droplet of well-defined homogenous material, actual tissue is
110
fairly inhomogeneous and the scattering signals that can be acquired through ultrasound imaging
vary greatly over the frequency band of the imaging transducer. IB analysis is one successful
method to characterize tissue for cardiac imaging and other applications. It can be expressed as the
ratio of the power spectrum averaged over the effective bandwidth of the transducer relative to
that from a flat reference reflector (O’Donnell and Miller 1981). However, since integration
eliminates frequency components of backscattering, some tissue characteristics are not detected
using IB values (Shung 2015).
Once the ultrasound and pathology image data sets were acquired, the pathology images were
used to determine where to sample the image data during the backscatter coefficient statistical
analysis. Thus, the gold standard pathology image provided the basis for region selection in the
ultrasound image sets. Ultrasound backscatter coefficients were measured, compared and plotted
to demonstrate the feasibility of identifying microstructures such as microcalcifications as well as
distinguishing tissue types including adenocarcinoma, adipose tissue and fibrous tissue using high
frequency ultrasound imaging. There is a clinical need to improve visualization of breast tissue,
including small microcalcifications, during ultrasound guided biopsy in order to reduce sampling
error and improve diagnostic accuracy. The ultrasound backscatter analysis presented in this paper
may improve tissue visualization and address this clinical need.
7.2.4 Quantitative Backscatter Analysis Methods
The spectral features for tissue characterization are useful for the characterization of healthy
and diseased tissues and has been applied to breast tissue imaging at clinical ultrasound frequencies
of 3 - 7 MHz (d'Astous and Foster 1986). Whereas the spectral features of the backscatter
111
coefficient may display a large variation over the frequency band, the fitting of it to a linear
function using a least squares regression can provide useful information about the tissue. The slope
of the linear fit of the backscatter coefficient as a function of frequency corresponds to the size of
the scatterers whereas its y-intercept corresponds to the volume concentration and relative acoustic
impedance of them (Lizzi et al. 2003). Therefore, acquiring the backscatter spectral features of
tissue can serve to identify certain tissue types from within a specimen of unknown, heterogeneous
tissue composition.
In this study, we computed the spectral features of the tissue backscatter according to the
principle demonstrated by Lizzi et al. (Lizzi et al. 2006). We then applied additional filtering and
post-processing steps to show the distinct tissue types and microstructures more clearly. To
normalize the backscattering feature, the spectrum of the pulse/echo signal when captured off a
highly reflected, polished quartz target is first measured as a reference. This reference pulse/echo
signal data is also used to define the performance characteristics of the transducer including center
frequency, bandwidth and focal point. The IB coefficient is the ratio of the backscattered signal
from a target tissue volume to that from a polished quartz reflector target (Thomas III et al. 1989)
(Shung 2015) and is given by:
𝐼𝐵 =
1
2∆𝑓 ∫
|𝑉 (𝑓 )|
2
|𝑅 (𝑓 )|
2
𝑑𝑓 =
∫
|𝑣 (𝑡 )|
2
𝑑𝑡 𝜏 +∆𝜏 𝜏 −∆𝜏 ∫
|𝑟 (𝑡 )|
2
𝑑𝑡 𝜏 +∆𝜏 𝜏 −∆𝜏 ,
𝑓 𝑐 +∆𝑓 𝑓 𝑐 −∆𝑓
(7-3)
where v(t) is a return echo A-line from the target tissue volume and r(t) is a reference signal from
the polished quartz target and V(f) and R(f) are their Fourier transforms. Then, to calculate the
slope and y-intercept of the backscatter coefficient as a function of frequency, the logarithmic
compressed spectral backscatter coefficient is defined as:
112
𝐵𝑆 (𝑓 )[𝑑𝐵 ]=10log
10
|𝑉 (𝑓 )|
2
|𝑅 (𝑓 )|
2
,
(7-4)
where a linear fit is applied to BS(f) giving a fit equation of a1f+a0, where a1 is the slope in dB/MHz
and a0 is the 0 Hz y-intercept in dB.
RF echo signals acquired during the B-mode imaging process are used for the backscatter
analysis processing. After reading one RF echo scanline from the tissue, the 2 neighboring lines
(separated by 10 µm in case 1,2,4 and 25 µm in case 3), are arithmetically averaged to reduce
random noise in the lateral direction. Because of the smaller pitch of lines than the anticipated
feature size of the tissue and microstructures we intend to image, applying this averaging should
not remove information in the lateral direction.
Next, a band-pass filter is applied to the spatially averaged RF data to cut random noise and
low frequency noise along the axial direction. The low- and high-pass frequencies were selected
based on being outside the -6 dB bandwidth of the transducer. A window is set on the region of
interest in the sample tissue volume and the frequency component of the partial RF data within the
window is calculated. In our investigation, a window size of 34 µm is chosen to correspond to the
axial resolution of the B-mode image. Before calculating the partial RF spectrum by using the
Fourier transform, the Hanning window function with length equal to that of the window is applied
to the RF data to avoid emerging noise due to the discontinuity of the data at the edges of the
window.
To acquire the backscatter coefficient of the tissue in each window, the spectral feature of the
windowed RF data is divided by that of the reference pulse/echo spectrum. The effective -6 dB
frequency range of the pulse/echo signal is frequently defined. In our study, a larger bandwidth
113
than the -6 dB band is employed to generate additional backscatter coefficient points which are
then used to plot the linear fit which is used to calculate both the slope and y-intercept. Integrating
the corrected backscatter spectrum over the defined bandwidth produces the IB coefficient. Using
a least squares method, the corrected backscatter spectrum is logarithmically compressed and a
linear fit function is applied. This linear fit function is defined by its slope and y-intercept.
Applying this technique along the axial direction at intervals of one resolved time step and along
the lateral scanning direction at intervals equal to the scan line pitch of the ultrasound measurement
can create the 2D plot of the local spectral backscatter coefficients. A spatial low pass filter is
applied along the axial direction to remove magnitude spikes in the slope and y-intercept 2D maps.
7.2.5 Statistical Analysis of Backscatter Coefficients
In this study, statistical analysis of the backscatter characteristics of each 2D map was
performed to determine if there were any statistically significant differences between tissue types
in breast biopsy core specimens using high frequency ultrasound imaging at 74 MHz. A sample
set of backscatter coefficient values was gathered for each tissue type by selecting a region of
interest from each B-mode image slice based on its match with the corresponding pathology image.
Each region of interest chosen was defined as a 250 µm x 250 µm square centered around a point
manually selected on the B-mode image. The pathology images served as the reference map when
choosing regions from the B-mode image used to calculate the ultrasound IB, slope and y-intercept
values at those locations.
The size of the region of interest for the microcalcification structures was defined as only 50
µm x 50 µm since the 250 µm x 250 µm region of interest extended beyond the boundary of many
114
small microcalcifications. The region of interest needs to be smaller than the structure or tissue
section to be sampled since the backscatter coefficient value should reflect only the tissue of
interest and not any surrounding tissue. Additionally, since there were not 5 microcalcifications in
each image slice only 1 region of interest was selected for each of these structures. In each image
slice produced from the biopsy tissue specimen, 5 regions from each tissue region identified during
pathological analysis were selected to form the sample sets used to compare backscatter
coefficients between these tissue types. In Case 1, 5 regions of adipose and adenocarcinoma were
selected. In case 2 (both part 1 and part 2), 5 regions of fibrous and adipose tissue and a single
region from each microcalcification were sampled. In case 3, 5 regions of adenocarcinoma tissue
were sampled. In case 4, 5 regions of fibrous tissue and a single region from each
microcalcification were sampled.
Sampling multiple regions of interest in each image slice was necessary because we needed
to account for the speckle pattern of the ultrasound data and because since the variability of echo
intensity even within homogenous tissue directly affects the backscatter characteristic values. By
sampling a large amount of regions of interest of each tissue type we were able to gather a
representative data set that was useful for the comparison of backscatter characteristics between
different tissue types. The value of IB, slope and y-intercept for each region was then calculated
and these values constituted the sample data set for each tissue type. The average and standard
deviation of these sample data sets was then calculated. Finally, sample data sets from different
tissue types within each of the 4 biopsy cases were compared to determine if there was a
statistically significant difference between them.
To compare these sample sets we performed the following tests. First we applied the
Kolmogorov-Smirnov test to the sampled data to verify that the sampled data variation for each
115
group is sufficiently normal before progressing to the next step. Then, to compare two variables
with normal distributions that are in good agreement, we applied the F-Test to judge whether these
two variables have the same variance. If their variances are sufficiently close, we applied the T-
Test in order to show a significant difference between them. Otherwise, if their variances are not
close, the Mann-Whitney Rank Sum Test is applied to show significant differences between them.
A threshold value of P ≤ 0.001 was used for both of these test to determine if the difference between
the groups was statistically significant.
7.3 Clinical Study Results
Ultrasound and pathology image sets were recorded and matched to allow for tissue analysis
and feature identification. Results for standard B-mode images as well as Nakagami and
quantitative backscatter analysis image processed images are shown in the following section.
Figure 7-4 below shows 3 successive ultrasound slices capture at 100 µm intervals.
116
Figure 7-4: Ultrasound B-Mode images from 3 consecutive image slices captured at 100 µm intervals.
The circled feature is a highly echogenic specular reflector with acoustic shadowing directly below, which is
consistent with a microcalcification.
The presence of microcalcifications indicated by the white dashed circle in Figure 7-4 was
confirmed by matching these images with the corresponding pathology images as shown in Figure
7-5. Figure 7-5 below shows 3 successive pathology slices capture at 50 µm intervals.
117
Figure 7-5: H&E stained digital pathology images of biopsy core tissue sectioned at intervals of 50 µm.
Once ultrasound and pathology image sets were captured, image slice matches were made
with one shown in Figure 7-6 below.
Figure 7-6: Comparison between (a) histological and (b) ultrasound images of an ex-vivo breast core
biopsy tissue sample.
118
Dashed circles in (b) highlight microcalcifications visualized with high frequency single element
ultrasound imaging system and the dashed circles and arrows in (a) show the matching features
visualized with digital histology images.
3D ultrasound image reconstruction was investigated as a possible method to improve feature
identification in biopsy specimens. Image reconstruction and rendering was performed using
MATLAB image processing tools and on reconstructed biopsy sample is shown in Figure 7-7
below.
Figure 7-7: Ultrasound image sets of (a) B-mode image slices captured at intervals of 100 µm were
compiled together to form a (b) 3D ultrasound rendered image of the biopsy specimen using MATLAB.
119
To date the 3D rendering has not produced images that improve the microcalcification
visualization but this method may be useful as a biopsy tissue visualization tool.
7.3.1 Nakagami Filtering Imaging Results
Figure 7-8 below shows the ultrasound B-Mode image and the corresponding algorithmic
display using the Nakagami filter technique. Rectangular regions A and B show the m-parameter
display plot of magnitude distribution versus normalized magnitude for the speckle pattern
contained within these regions. After ultrasound imaging the pathological imaging confirmed two
distinct tissue types within the specimen, namely fat, highlighted in region A, and adenocarcinoma
tumor, highlighted in region B.
120
Figure 7-8: (a) Ultrasound B-Mode Image with two regions (A & B) identified for parameter estimation
using the Nakagami filtering technique along with their respective (c) gray scale histograms. The m-parameter
plot (b) shows that the left and right halves of this biopsy specimen are constituted of fat on the left and tumor.
The Nakagami filtering technique was applied to image sets from a different biopsy specimen with
more microcalcifications identified during pathological analysis. Figure 7-9 below demonstrates
how both the M-Parameter and Ω-Parameter maps plotted separately along with the corresponding
pathology image.
121
Figure 7-9: Pathology image with m-parameter and Ω-parameter maps of ultrasound image as well as
the original B-Mode ultrasound image.
The m-parameter maps seemed to show the microcalcifications especially when an amplitude
threshold was applied as shown in Figure 7-10.
122
Figure 7-10: Pathology image (a) with regions highlighted by pathologists to indicate microcalcifications
(red) and necrotic tissue (green). These region outlines were then matched to the original b-mode image (b). Ω-
parameter (c) and m-parameter (d) maps of ultrasound image plotted over original ultrasound image. These
parameter maps had a threshold applied to them so only regions of the highest parameter magnitude were
displayed.
This is likely due to the fact that the microcalcifications have very bright, homogenous speckle
patterns which contrasts greatly from the surrounding soft tissue which has more diffuse and lower
intensity speckle patterns. These preliminary clinical imaging results and tissue characterization
maps show promise as providing useful clinical feature identification in breast tissue during high
frequency ultrasound imaging. Further investigation is underway to substantiate this approach.
123
7.3.2 Backscatter Analysis Results
The imaging and statistical analysis results demonstrate that high frequency ultrasound
backscatter analysis is capable of not only visualizing microcalcifications in breast tissue but also
differentiating tissue types. We found that backscatter analysis could distinguish adipose tissue
from adenocarcinoma and how microcalcifications could be distinguished from adipose tissue as
well as fibrous tissue in the breast. In the following discussion, each backscatter characteristic 2D
map set for cases 1 - 4 is shown with the corresponding ultrasound B-mode image and H&E stained
pathology image. The statistical analysis results for cases 1 – 4 are then presented including a
discussion on how microcalcifications and various tissue types were able to be distinguished from
each other.
For each of the 51 ultrasound-pathology image sets generated for each of the 4 biopsy tissue
specimens procured and imaged in this study, a set of ultrasound backscatter 2D maps was
produced to show how the backscatter analysis with high frequency ultrasound imaging is useful
in characterizing different tissue types and microstructures in breast tissue. The following image
sets are composed of a B-mode ultrasound image followed by 2D maps of the IB (dB), slope
(dB/MHz) and y-intercept (dB). The parameters in these 2D maps are depicted in a color scale
with blue and red representing the low and high values, respectively, on the map. Finally, a digital
pathology image of the original biopsy tissue specimen is shown as a reference to correlate regions
of the B-mode image with backscatter parameter maps of different tissue types.
124
Case 1 Results
The tissue in case 1 was obtained by sampling a core biopsy specimen when the sampling
aperture of a 14-Gauge core biopsy needle spanned the boundary between adipose and
adenocarcinoma tissue in the breast as confirmed by an external linear array ultrasound imaging
probe used during the procedure. Thus, the left half of the tissue specimen is composed of adipose
whereas the right half is composed of adenocarcinoma tissue and this is confirmed by the matching
pathology image. This biopsy was obtained from a patient diagnosed with invasive ductal
carcinoma. A total of 9 ultrasound-pathology image sets were created from this specimen and one
selected image set is displayed in Figure 7-11. The speckle pattern of the adipose and
adenocarcinoma tissue appears slightly different in the B-mode image with the adipose region
having a larger, more dispersed speckle pattern whereas the adenocarcinoma region has a relatively
smaller, more tightly distributed speckle pattern.
125
Figure 7-11: Case 1 image set includes high frequency ultrasound a) B-mode image followed by the
backscatter parameter plots for b) IB, c) slope and d) y-intercept and finally the e) H&E stained digital
pathology image.
126
Viewing the B-mode image alongside the 2D maps of the IB, slope and y-intercept shows a
clear difference between the adipose and adenocarcinoma regions. The ultrasound backscatter
parameters of the adipose and adenocarcinoma regions indicated by the black and white arrows,
respectively, were averaged and the standard deviation calculated. The mean and standard
deviation of the IB for the adipose tissue was -73.7 dB and 2.0 dB, respectively, relative to the
echo from a flat reference reflector. The average and standard deviation of the IB for the
adenocarcinoma tissue was -67.3 dB and 4.0 dB. The average and standard deviation of the slope
for the adipose tissue was -0.46 dB/MHz and 0.06 dB/MHz. The average and standard deviation
of the slope for the adenocarcinoma tissue was -0.07 dB/MHz and 0.05 dB/MHz. The average and
standard deviation of the y-intercept for the adipose tissue was -47.3 dB and 4.6 dB. The average
and standard deviation of the y-intercept for the adenocarcinoma tissue was -63.6 dB and 3.5 dB.
The slope and y-intercept maps show a distinct difference between these two regions; the
adipose region of the slope map shows an average value of -0.46 dB/MHz whereas that of the
adenocarcinoma region is nearly horizontal at -0.07 dB/MHz. Similarly, the adipose region of the
y-intercept map has values around -47.3 dB whereas the adenocarcinoma region is at -63.6 dB.
The IB plot does not appear to show as large a difference between the adipose and adenocarcinoma
regions as the slope and y-intercept maps do. It appears that the slope and y-intercept coefficient
2D maps are more useful in differentiating adipose from adenocarcinoma tissue in the breast than
the IB map. These observed differences were subsequently confirmed quantitatively by statistical
analysis which will be described in the following section.
127
Case 2 Results
The tissue in case 2 was obtained from a patient with metastatic adenocarcinoma consistent
with a breast primary. The biopsy specimen separated into 2 pieces at the time of acquisition from
the patient and so the ultrasound B-mode and backscatter parameter 2D maps are segmented into
part 1 (left side) and part 2 (right side) as shown in Figure 7-12. Small microcalcifications were
identified in the tissue in both the pathology and ultrasound images. The diagnosis of metastatic
adenocarcinoma was made using a separate biopsy specimen acquired during the same biopsy
procedure as this specimen imaged in case 2; however no adenocarcinoma tissue was observed in
the tissue sections we examined from this particular specimen. A total of 14 ultrasound-pathology
image sets were created of this specimen and one selected image set is displayed in Figure 7-12.
Microcalcifications appear as small, bright echoes with clear acoustic shadowing directly beneath
and are indicated by white arrows.
128
Figure 7-12: Case 2 image set include images for the 2 parts of the biopsy specimen. The image set for
parts 1 and 2, respectively, includes a high frequency ultrasound B-mode image (a,f) followed by the
backscatter parameter plots for IB (b,g), slope (c,h) and y-intercept (d,i) and finally the H&E stained digital
pathology image (e,j). Microcalcifications, fibrous tissue and adipose tissue are indicated by solid white arrows,
solid black arrows and cross-hatched arrows, respectively.
In this case, the IB and y-intercept maps were most effective in highlighting the positions of
the microcalcifications within part 1 of the biopsy tissue specimen. Both the IB and y-intercept
maps showed a large difference in values as compared to the rest of the surrounding fibrous and
adipose tissue in the biopsy specimen. Microcalcifications were still visibile in the IB and y-
129
intercept maps in part 2, however, the magnitude was not as large as in part 1. This could be due
to the smaller size of the microcalcifications in part 2 as compared with part 1; size difference can
be observed on the pathology image in (e) and (f).Viewing the B-mode image alongside the 2D
maps of the IB, slope and y-intercept shows how the ultrasound backscatter parameter maps are
useful in highlighting the location of the microcalcifications within the biopsy tissue specimen.
The echoes from the microcalcifications are often reflected from only the top edge of these
microstructures with acoustic shadowing occurring below the structure. As a result they look
smaller on the ultrasound B-mode and backscatter 2D maps than they do on the pathology images
where they appear as dark purple structures due to the H&E staining process.
The ultrasound backscatter values of the microcalcifications indicated by the white arrows as
well as those of the surrounding fibrous tissue region were averaged and the standard deviation
calculated. The average and standard deviation of the IB for the microcalcifications was -50.8 dB
and 8.37 dB. The average and standard deviation of the IB for the fibrous tissue was -56.1 dB and
3.73 dB. The average and standard deviation of the IB for the adipose tissue was -56.7 dB and 2.94
dB. The average and standard deviation of the slope for the microcalcifications was -0.30 dB/MHz
and 0.12 dB/MHz. The average and standard deviation of the slope for the fibrous tissue was -0.17
dB/MHz and 0.07 dB/MHz. The average and standard deviation of the slope for the adipose tissue
was -0.23 dB and 0.04 dB. The average and standard deviation of the y-intercept for the
microcalcifications was -38.3 dB and 10.17 dB. The average and standard deviation of the y-
intercept for the fibrous tissue was -46.1 dB and 5.50 dB. The average and standard deviation of
the y-intercept for the adipose tissue was -43.3 dB and 2.94 dB. The slope and y-intercept maps
were most effective in highlighting the positions of the microcalcifications within the biopsy tissue
specimen. Both the slope and y-intercept maps showed a large difference in values as compared
130
to the rest of the surrounding tissue in the biopsy specimen. This difference was subsequently
confirmed by statistical analysis as will be shown in the following section.
Case 3 Results
The tissue in case 3 was obtained from a patient diagnosed with poorly differentiated
adenocarcinoma consistent with a breast primary. The entire biopsy specimen was comprised of
adenocarcinoma tissue. A total of 5 ultrasound-pathology image sets were created of this specimen
and one selected image set is displayed in Figure 7-13.
131
Figure 7-13: Case 3 image set includes high frequency ultrasound a) B-mode image followed by the
backscatter parameter plots for b) IB, c) slope and d) y-intercept and finally the e) H&E stained digital
pathology image. The pathology image shows the biopsy specimen composed of adenocarcinoma.
The ultrasound backscatter parameters of the adenocarcinoma tissue in this biopsy specimen
were averaged and the standard deviation calculated. The average and standard deviation of the IB
was -60.9 dB and 2.41 dB. The average and standard deviation of the slope was -0.14 dB/MHz
and 0.03 dB/MHz. The average and standard deviation of the y-intercept was -52.8 dB and 3.42
dB. A fairly consistent color is seen throughout the 3 backscatter coefficient 2D maps and the B-
mode image shows a consistent speckle pattern throughout the specimen which is consistent with
the fact that the tissue type is homogeneous.
132
Case 4 Results
The tissue biopsy in case 4 was obtained with an 11-Gauge vacuum assisted stereotactic breast
biopsy needle from a patient that was diagnosed with invasive ductal carcinoma. This diagnosis
was made using a separate biopsy specimen with viable tumor obtained during the same biopsy
procedure; the case 4 biopsy specimen had no viable tumor tissue. A total of 23 ultrasound-
pathology image sets were created of this specimen and one selected image set is displayed in
Figure 7-14. The locations of the 3 most prominent microcalcifications in this specimen are
indicated with arrows in each of the 5 images in the case 4 image set.
133
Figure 7-14: Case 4 image set includes high frequency ultrasound a) B-mode image followed by the
backscatter parameter plots for b) IB, c) slope and d) y-intercept and finally the e) H&E stained digital
pathology image. The white arrows with black outline indicate the positions of microcalcifications.
134
Viewing the B-mode image alongside the 2D maps of the IB, slope and y-intercept shows how
the ultrasound backscatter maps are useful in highlighting the location of the microcalcifications
within the biopsy tissue specimen. These microcalcifications appear as bright echoes with acoustic
shadowing directly beneath as would be expected when imaging these small structures at this
frequency. The ultrasound backscatter values of the microcalcifications indicated by the arrows as
well as those of the surrounding fibrous tissue region were averaged and the standard deviation
calculated. The average and standard deviation of the IB for the microcalcifications was -30.8 dB
and 4.0 dB. The average and standard deviation of the IB for the fibrous tissue was -39.0 dB and
5.0 dB. The average and standard deviation of the slope for the microcalcifications was -0.21
dB/MHz and 0.14 dB/MHz. The average and standard deviation of the slope for the fibrous tissue
was -0.09 dB/MHz and 0.07 dB/MHz. The average and standard deviation of the y-intercept for
the microcalcifications was -18.7 dB and 7.6 dB. The average and standard deviation of the y-
intercept for the fibrous tissue was -33.8 dB and 5.9 dB. The IB and y-intercept maps were most
effective in highlighting the positions of the microcalcifications within the biopsy tissue specimen.
Both the IB and y-intercept maps showed a large difference in coefficient values as compared to
the rest of the surrounding tissue in the biopsy specimen as was observed in Case 2. This difference
was subsequently confirmed by statistical analysis as will be shown in the following section.
135
7.4 Statistical Analysis Results
In addition to the ultrasound backscatter 2D maps shown above, statistical analysis was
performed on these data and box plots were created for each case to show the backscatter parameter
values for each tissue type.
Case 1 Results
Box plots in Figure 7-15 show the distribution of IB, slope and y-intercept values of adipose
and adenocarcinoma tissue for the case 1 biopsy specimen.
Figure 7-15: Box plots for adipose and adenocarcinoma tissue for IB (dB), slope (dB/MHz) and y-intercept
(dB) show the median value, given by the band in the middle of the box. The upper and lower quartiles are
given by the upper and lower borders of the box, respectively. The 4 outliers outside of the box plot whiskers
are represented by circles.
The difference in the IB, slope and y-intercept values between the adipose and
adenocarcinoma tissue group was found to be statistically significant. The statistically significant
difference between each of the adipose and adenocarcinoma tissue’s IB, slope and y-intercept
values confirms the large difference seen in the color of the adipose and adenocarcinoma tissue
regions of the 2D maps for each of the 3 backscatter parameters shown in Figure 7-15. As shown
in the pathology image, the typical cell size of adipose tissue is much larger than that of
136
adenocarcinoma tissue and this large distinction is quantified by the large difference in slope
values between the two tissue types. According to Lizzi et al., the slope is dependent on both the
scatterer’s size and frequency of ultrasound (d'Astous and Foster 1986). The diameter of the cell
can be represented by 2a = 12 µm and ka=1.8 when the center frequency of the ultrasound is 70
MHz, where a is the radius of the scatterer and k is the wave number. The adipose cells in the
tissue we measured are larger than 12 µm and therefore the result that most of the slope values are
evaluated as negative is reasonable.
The adipose and the adenocarcinoma tissue are not hyperechoic and as a result have both a
low mean IB (-73.7 dB and -67.3 dB) and y-intercept (-47.3 dB and -63.6 dB). Notwithstanding
these low IB and y-intercept magnitudes, the difference in these values among the two tissue types
demonstrates the utility of this type of analysis since differentiating adipose from adenocarcinoma
for purposes of an image guided breast biopsy is a critical part of avoiding sampling error and
achieving high diagnostic accuracy of breast cancer.
Case 2 Results
Box plots in Figure 7-16 show the distribution of IB, slope and y-intercept values of
microcalcifications as well as fibrous and adipose tissue for the case 2 biopsy specimen.
137
Figure 7-16: Box plots for microcalcifications as well as fibrous and adipose tissue for IB (dB), slope
(dB/MHz) and y-intercept (dB) show the median value, given by the band in the middle of the box. The upper
and lower quartiles are given by the upper and lower borders of the box, respectively. The 4 outliers outside of
the box plot whiskers are represented by circles.
The difference in the IB, slope and y-intercept values between the microcalcification and
fibrous tissue groups is statistically significant. The difference in those values between the
microcalcification and adipose tissue groups is also statistically significant (P ≤ 0.05). Backscatter
slope and y-intercept values showed a statistically significant difference between adipose and
fibrous tissue groups. However, no significant difference was observed between adipose and
fibrous groups when comparing the IB values. The higher IB values of the microcalcifications as
compared to fibrous and adipose tissue is understandable since the wavelength at 74 MHz in tissue
is comparable to the size of the microstructures and therefore they act as specular reflectors when
imaged. These specular reflectors produce a larger backscatter signal as compared with the
surrounding soft fibrous and adipose tissue.
Case 3 Results
Box plots in Figure 7-17 show the distribution of IB, slope and y-intercept values of
adenocarcinoma tissue for the case 3 biopsy specimen.
138
Figure 7-17: Box plots for IB (dB), slope (dB/MHz) and y-intercept (dB) of adenocarcinoma tissue show
the median value, given by the band in the middle of the box. The upper and lower quartiles are given by the
upper and lower borders of the box, respectively. The outliers outside of the box plot whiskers are represented
by circles.
Since this specimen had only one type of tissue present no comparative analysis was
performed.
Case 4 Results
Box plots in Figure 7-18 show the distribution of IB, slope and y-intercept values of
microcalcifications as well as fibrous tissue for the case 4 biopsy specimen.
Figure 7-18: Box plots for microcalcifications and surrounding fibrous tissue for IB (dB), slope (dB/MHz)
and y-intercept (dB) show the median value, given by the band in the middle of the box. The upper and lower
quartiles are given by the upper and lower borders of the box, respectively. The 4 outliers outside of the box
plot whiskers are represented by circles.
139
The difference between IB, slope and y-intercept values for the microcalcification and fibrous
tissue groups is statistically significant. The microcalcifications again show different backscatter
characteristics with most IB values being higher than that of fibrous tissue because of higher
acoustic impedance and therefore higher echogenicity as compared to fibrous tissue. In this study,
we did not consider the effect of attenuation for each tissue type. Attenuation can affect the
calculation of the ultrasound backscatter parameters, especially in cases where one is imaging deep
into tissue or when the tissue has a large attenuation coefficient. The reason that we did not
consider attenuation for this study is because it was impractical to evaluate the attenuation
coefficients of each part of tissues precisely since the tissue specimen would not have survived
additional testing to determine the attenuation coefficients due to that fact that they were delicate,
difficult to handle and fresh (not fixed in formalin prior to ultrasound imaging), leaving only a
short time period to capture all ultrasound image data. Additionally, these tissues would break
apart easily and therefore prohibit capturing actual attenuation measurements of the tissue in its
original shape. Maintaining precise control of the orientation of the specimen was critical in order
to ensure that the ultrasound and pathology image planes were parallel with each other and enable
the comparison of the two image sets. Although it may improve the accuracy of our results to
compensate for attenuation, all of the results presented here are consistent with the theoretical
results previously published in the literature (d'Astous and Foster 1986). Furthermore, the purpose
of the study was to evaluate if high frequency ultrasound backscatter analysis is capable of
characterizing different tissue types in one biopsy specimen image and not to characterize the
absolute backscatter parameters values of the tissue. Therefore, the analysis we performed for each
biopsy case allowed us to confirm that differences between tissue types can be visualized for an
individual core biopsy specimen. This is clinically useful since a physician must make a
140
determination about where in the breast to sample tissue from in order to make a diagnosis and
better tissue characterization at the time of biopsy can improve sampling accuracy which in turn
improves diagnostic accuracy of breast cancer.
In this study we demonstrate how quantitative analysis of high frequency ultrasound images
of breast tissue may be useful in identifying microcalcifications and also in differentiating tissue
types within breast core biopsy specimens. Ultrasound backscatter parameters (IB, slope and y-
intercept) were calculated from RF data obtained from imaging ex vivo breast biopsy specimens.
Backscatter parameters were calculated for each location in the B-mode image and were used to
create 2D maps of these parameters. These maps were compared with the B-mode and pathology
images of the tissue in order to confirm the nature of the tissue on which backscatter analysis was
performed. In future work more precise characterization of backscatter spectral features may be
achieved by acquiring ultrasound imaging data using a single-element transducer with broader
bandwidth. Alternatively, utilizing a high frequency array may improve data acquisition even
further by enabling data capture at multiple focal depths within the tissue specimen.
As has been previously reported in the literature for clinical ultrasound frequencies (2 – 20
MHz), quantitative breast ultrasound image analysis is useful as a diagnostic tool and this paper
offers new evidence that backscatter analysis of breast tissue at a much higher frequency (74 MHz)
is capable of providing useful information in differentiating adenocarcinomas from adipose tissue
and microcalcifications from both fibrous and adipose tissue. Backscatter analysis at higher
frequencies may differ considerably from that at lower frequencies since the wavelength is
drastically decreased and is now approximately equal to or even smaller than some microstructures
in the breast. One practical application of this work is that this type of quantitative ultrasound
processing could be applied to data captured by the high frequency linear array mounted on a
141
biopsy needle that is being developed for imaging surrounding tissues during breast biopsy. This
sort of automated, quantitative analysis may enable better identification of microstructures and
tissue types during biopsy, further improving the sampling accuracy during breast core biopsy
procedures, thus increasing overall diagnostic accuracy of breast cancer.
142
Summary and Future Work
8.1 Array Development Summary
The design, fabrication and testing of a 60 MHz miniaturized linear array using DRIE
micromachining fabrication techniques has been described. The array is intended to be integrated
into a breast core biopsy needle. The array characterization and image formation testing served as
a successful proof of concept for the new array fabrication and electrical interconnect assembly
processes. The axial resolution was acceptable and steps to improve the lateral resolution may
include increasing the aperture size and improving the bandwidth of the array elements by adding
a second matching layer. Improving the array element uniformity and performance and
implementing a more robust and simple electrical interconnect solution are the focus of the future
work for this miniature high frequency array development. This array fabrication and testing
confirmed that the most challenging aspect of fabricating this type of array is the connection of the
miniature high frequency ultrasound array to the imaging system and future development paths
include using micromachining technologies to make even simpler, smaller and easier to process
interconnect components. Developing robust high density electrical interconnect solutions for this
high frequency array and improving acoustic properties of array elements will be critical for
progressing towards clinical applications for this array. In the future, this type of miniaturized high
frequency array has the potential to improve tissue visualization during breast biopsy procedures,
making tissue sampling more accurate and thus increasing the accuracy of breast cancer diagnosis.
143
8.2 Clinical Study Summary
Additionally, we have preliminary confirmation that microcalcifications in the breast less
than 100 µm in size can be clearly visualized with high frequency ultrasound imaging. 3D
reconstruction and rendering, Nakagami filtering and quantitative backscatter analysis techniques
have been investigated in order to provide improved feature identification for breast cancer biopsy
specimens. The results from the clinical ex vivo study suggest that high frequency ultrasound is
useful in identifying microcalcifications differentiating cancerous from normal tissue in breast
tissue.
8.3 Future Array Development
We are currently improving the electrical interconnect design of the array-to-transmission
line connection to improve the signal propagation to and from the array in order to decrease signal
loss and increase sensitivity. To date, a custom synthetic aperture imaging system controlled by a
LabVIEW/MATLAB hybrid software program has been used to produce images using the 60 MHz
miniaturized linear array. The system successfully imaged a tungsten wire target in a de-ionized
water bath. Future versions of this high frequency miniaturized array built for a wide range of
clinical applications can improve upon the sensitivity, bandwidth and electrical interconnect
design of the current array prototype.
144
8.3.1 Electrical Interconnect Challenges and Proposed Solutions
The challenge of making robust electrical interconnects with all elements in this miniaturized
high frequency array was significant, especially since the transmission line used in the current
prototype was made with a thin glass substrate, making it extremely fragile. Repeated attempts to
connect the two components has resulted in cracking of the transmission line, destroying all 64
channel traces in the process. It was only with the development of the conductive microsphere
interconnect method that we were able to make the connection to all 64 elements without
destroying the array, transmission line, or both. However, drawbacks still remain with this process
since the conductive microspheres had to be placed manually one at a time. The next prototype
built will need to have this process automated with a more advanced computer controlled
conductive epoxy bumping station. These machines are currently used in industry to bump circuits
with conductive epoxies and are well suited to place all the conductive epoxy bumps and
conductive microspheres, all without direct human intervention, which will make the connections
more uniform and reliable.
8.3.2 New Electrical Interconnect Design – Cavity Backing Design
A new electrical interconnect scheme must improve upon the current design according the 4
requirements listed in Table 16 below.
145
Table 16: Electrical Interconnect Design Parameters
Design Parameter Requirement
Electrical Impedance Maintain electrical impedance matching with array elements.
Cross-Section Size Enable more compact array/transmission line assembly; the
current minimum needle diameter is 11 AWG.
Bonding Pad Connection Simplify array-to-transmission line bonding pad connection.
Robust Transmission Line Transmission line substrate must be a robust material that can
withstand the temperature and pressure that will be experienced
during the fabrication and testing processes
The solution that satisfies these 4 design constraints is designated as the cavity backing design.
The solution is so named because the critical component is a small high density interposer with a
center cavity that serves as the bridge between the array composite and the transmission line
component. The overview of this design is shown in Figure 8-1 with the array composite, cavity
backing interposer and flex circuit transmission line are labeled. This solution was developed
based on what we learned about the fabrication and testing of the current array prototype.
Figure 8-1: Cavity backing electrical interconnect design showing the (a) assembled unit composed of the
flex circuit transmission line, high density interposer and array composite as well as the (b) exploded view of
this assembly. Dimensions shown are for a comparable 60 – 80 MHz, 64 element linear array.
146
Since this high frequency array uses a composite that is extremely thin (i.e. 24 um thick for
60 MHz array), the electrical interconnect assembly process must be integrated with the array
composite fabrication process since handling the composite by itself is not feasible. Any process
that involves moving the thin composite with a tool such as tweezers or vacuum chuck would
destroy it. Therefore, all fabrication and electrical interconnect assembly steps must be completed
in a single, integrated wafer level process. It is critical to ensure that there is always a firm, flat
support structure for the composite throughout the fabrication and electrical interconnection
processes. The array composite and electrical interconnect assembly process are detailed in Figure
8-2 below.
147
Figure 8-2: Array electrical interconnect solution fabrication and assembly process.
The high density interposer conductive via/pad layout pattern matches the element electrode
bonding pad layout of the array composite as well as that of the flex circuit transmission line,
allowing for direct connection of these 3 components. The via/pad layout pattern of the high
density interposer is shown in Figure 8-3 below.
Figure 8-3: Diagram of the high density interposer with (a) detailed view of conductive via layout and (b)
overview of interposer.
148
8.3.3 High Density Interposer Vias and Pad Connections
The high density interposer component can be made with industry standard wafer fabrication
processes using either silicon or glass as the substrate material. The constraining requirement is
the conductive via/bonding pad pitch as well as the ability to add a metalized bump to the top and
bottom surface of the interposer bonding pads. The bump height should at least 20 µm in order to
ensure that each bonding pad on the array composite and flex circuit transmission line makes solid
contact with the metalized bump on the high density interposer. This connection will be made
without solder so as to avoid the need to reflow the solder during a flip chip assembly process.
Exposing the array composite to high temperatures should be avoided so as not to cause expansion
and contraction of the delicate array composite. This cavity backing electrical interconnect design
satisfies the 4 critical design criteria (electrical impedance matching, cross-section size, bonding
pad connection, robust transmission line) previously described. It also maintains the beneficial
aspect of being mass producible since it involves a wafer level batch fabrication process. With this
process 49 arrays/batch can be produced and this number could be further increased by expanding
the supporting wafer size. This interconnect design would make significantly smaller array
package sizes possible and open up additional clinical applications that require smaller needle
gauges or increased element counts.
149
8.4 Clinical Study Future Directions
Future directions for the clinical study used to determine the effectiveness of high frequency
ultrasound in characterizing tissue types and microstructures include expanding the focus to other
clinical needs including sentinel lymph node and liver biopsy. These are 2 soft tissue biopsy
applications that may benefit from the type of high resolution tissue visualization to improve
sampling accuracy during biopsy procedures. Clinical studies that image ex vivo tissue samples of
lymph node and liver could be useful in determining if a miniaturized high frequency array may
also be used to improve sampling accuracy in these biopsy procedures.
150
Bibliography
Alizad A, Fatemi M, Wold LE, Greenleaf JF. 2004. Performance of vibro-acoustography
in detecting microcalcifications in excised human breast tissue: A study of 74 tissue samples.
Medical Imaging, IEEE Transactions On 23(3): 307-12.
Baer R and Kino G. 1984. Theory for cross coupling in ultrasonic transducer arrays.
Applied Physics Letters 44(10): 954-6.
Bassett LW. 1992. Mammographic analysis of calcifications. Radiologic Clinics of North
America 30(1): 93-105.
Bernassau A, Hutson D, Demore CE, Flynn D, Amalou F, Parry J, McAneny J, Button
TW, Desmulliez MP, Cochran S. 2009. Progress towards wafer-scale fabrication of
ultrasound arrays for real-time high-resolution biomedical imaging. Sensor Review 29(4):
333-8.
Bezanson A, Adamson R, Brown JA. 2014. Fabrication and performance of a
miniaturized 64-element high-frequency endoscopic phased array. Ultrasonics, Ferroelectrics,
and Frequency Control, IEEE Transactions On 61(1): 33-43.
Blitz J and Simpson G. 1996. Ultrasonic methods of non-destructive testing. Springer
Science & Business Media.
Bruening W, Schoelles K, Treadwell J. 2011. Comparative effectiveness of core-needle
and open surgical biopsy for the diagnosis of breast lesions. Comparative Effectiveness
Review, Prepared by ECRI Institute Evidence-Based Practice Center 19.
Burbank F, Parker SH, Fogarty TJ. 1996. Stereotactic breast biopsy: Improved tissue
harvesting with the mammotome. American Journal of Surgery 62(9): 738-44.
Cady W. 1964. Piezoelectricity. Dover. New York.
Campbell JA and Waag RC. 1984. Measurements of calf liver ultrasonic differential and
total scattering cross sections. Journal of the Acoustical Society of America 75(2): 603-11.
Cannata JM, Williams J, Zhou Q, Ritter T, Shung KK. 2006. Development of a 35-MHz
piezo-composite ultrasound array for medical imaging. Ultrasonics, Ferroelectrics, and
Frequency Control, IEEE Transactions On 53(1): 224-36.
Cannata JM, Zhao J-Z, Ayyappan S., Ritter TA, Chen W. and Shung K. Kirk. 1999.
Fabrication of high frequency (25-75 MHz) single element ultrasonic transducers. 1999 IEEE
Ultrasonics Symposium Proceedings. 1099.
151
Cannata Jonathan M. and Shung K. Kirk. 2003. A comparison of model and experiment
for a high frequency (35 MHz) linear ultrasonic array. Ultrasonics, 2003 IEEE Ultrasonics
Symposium Proceedings. 1658.
Certon D, Felix N, Lacaze E, Teston F, Patat F. 2001. Investigation of cross-coupling in
1-3 piezocomposite arrays. Ultrasonics, Ferroelectrics, and Frequency Control, IEEE
Transactions On 48(1): 85-92.
Chaudhari M, Forsberg F, Voodarla A, Saikali F, Goonewardene S, Needleman L, Finkel
G, Goldberg B. 2000. Breast tumor vascularity identified by contrast enhanced ultrasound and
pathology: Initial results. Ultrasonics 38(1): 105-9.
Chen R, Cabrera-Munoz NE, Lam KH, Hsu H, Zheng F, Zhou Q, Shung KK. 2014. PMN-
PT single-crystal high-frequency kerfless phased array. Ultrasonics, Ferroelectrics, and
Frequency Control, IEEE Transactions On 61(6): 1033-41.
Cho N, Moon WK, Park J. 2009. Real-time US elastography in the differentiation of
suspicious microcalcifications on mammography. European Journal of Radiology 19(7):
1621-8.
Choi S, Shrout RT, Jang S, Bhalla A. 1989. Dielectric and pyroelectric properties in the
pb (Mg1/3Nb2/3) O3-PbTiO3 system. Ferroelectrics 100(1): 29-38.
CR Bard Magnum Biopsy Reusable Core Biopsy System, 2014.
http://www.bardbiopsy.com/products/magnum.php.
Curie P and Curie P. 1880. Development by pressure of polar electricity in hemihedral
crystals with inclined faces. Bulletin of the Chemical Society of France 3: 90.
D'Amico A and Pittenger R. 2009. A Brief History of Active Sonar. Space and Naval
Warfare Systems Center. San Diego.
d'Astous F and Foster F. 1986. Frequency dependence of ultrasound attenuation and
backscatter in breast tissue. Ultrasound in Medicine and Biology 12(10): 795-808.
Deaner AN, Cubukcu AA, Rees MR. 1992. Assessment of coronary stent by intravascular
ultrasound. International Journal of Cardiology 36(1): 124-6.
Démoré CE, Brown J, Lockwood GR. 2006. Investigation of cross talk in kerfless annular
arrays for high-frequency imaging. Ultrasonics, Ferroelectrics, and Frequency Control, IEEE
Transactions On 53(5): 1046-56.
Fornage BD, Faroux MJ, Simatos A. 1987. Breast masses: US-guided fine-needle
aspiration biopsy. Radiology 162(2): 409-14.
152
Foster FS, Pavlin CJ, Harasiewicz KA, Christopher DA, Turnbull DH. 2000. Advances
in ultrasound biomicroscopy. Ultrasound in Medicine and Biology 26(1): 1-27.
Foster FS, Mehi J, Lukacs M, Hirson D, White C, Chaggares C, Needles A. 2009. A new
15–50 MHz array-based micro-ultrasound scanner for preclinical imaging. Ultrasound in
Medicine and Biology 35(10): 1700-8.
Frazier CH and Brien W. 1998. Synthetic aperture techniques with a virtual source
element. Ultrasonics, Ferroelectrics, and Frequency Control, IEEE Transactions On 45(1):
196-207.
Giovannini M, Seitz JF, Monges G, Perrier H, Rabbia I. 1995. Fine-needle aspiration
cytology guided by endoscopic ultrasonography: Results in 141 patients. Endoscopy 27(2):
171-7.
Hackenberger Wesley, Kwon Seongtae, Rehrig Paul, Snook Kevin and Rhee Sorah.
2002. 2-2 PZT-polymer composites for high frequency (> 20 MHz) ultrasound transducers.
2002 IEEE Ultrasonics Symposium Proceedings. 1253.
Hasegawa BH. 1990. The physics of medical x-ray imaging. Medical Physics Publishing
Co.
Dielectric and piezoelectric properties of [001] poled PMN-32%PT (Type-B) single
crystal. 2014. http://www.hcmat.com/Pmn_Properties.html.
Herman GT. 2009. Fundamentals of computerized tomography: Image reconstruction
from projections. Springer Science & Business Media.
Herth FJ, Eberhardt R, Vilmann P, Krasnik M, Ernst A. 2006. Real-time endobronchial
ultrasound guided transbronchial needle aspiration for sampling mediastinal lymph nodes.
Thorax 61(9): 795-8.
Ito Y, Kushida K, Sugawara K, Takeuchi H. 1995. A 100-MHz ultrasonic transducer
array using ZnO thin films. Ultrasonics, Ferroelectrics, and Frequency Control, IEEE
Transactions On 42(2): 316-24.
Jensen JA. 1996. Field: A program for simulating ultrasound systems. 10
th
Nordic-Baltic
Conference on Biomedical Imaging, 4(1): Part 1: 351-353.
Jensen JA and Svendsen NB. 1992. Calculation of pressure fields from arbitrarily shaped,
apodized, and excited ultrasound transducers. Ultrasonics, Ferroelectrics, and Frequency
Control, IEEE Transactions On 39(2): 262-7.
Jespersen SK, Wilhjelm JE, Sillesen H. 1998. Multi-angle compound imaging. Ultrasonic
Imaging 20(2): 81-102.
153
Kamp O and Cramer M. 2008. History of echocardiography in the netherlands: 30 years
of education and clinical applications. Netherlands Heart Journal 16(1): 16-20.
Karakitsos D, Labropoulos N, De Groot E, Patrianakos AP, Kouraklis G, Poularas J,
Samonis G, Tsoutsos DA, Konstadoulakis MM, Karabinis A. 2006. Real-time ultrasound-
guided catheterisation of the internal jugular vein: A prospective comparison with the
landmark technique in critical care patients. Journal of Critical Care 10(6): R162.
Kerlikowske K, Grady D, Rubin SM, Sandrock C, Ernster VL. 1995. Efficacy of
screening mammography: A meta-analysis. JAMA 273(2): 149-54.
Khuri-Yakub BT and Oralkan Ö. 2011. Capacitive micromachined ultrasonic transducers
for medical imaging and therapy. Journal of Micromechanics and Microengineering 21(5): 4-
14.
Kim GR, Kang J, Kwak JY, Chang JH, Kim SI, Youk JH, Moon HJ, Kim MJ, Kim E.
2014. Photoacoustic imaging of breast microcalcifications: A preliminary study with 8-gauge
core-biopsied breast specimens. PLoS one 9(8): e105878.
Kino GS. 1987. Acoustic waves: Devices, imaging, and analog signal processing.
Prentice-Hall. Englewood Cliffs, NJ.
Lee J, Chang JH, Jeong JS, Lee C, Teh S, Lee A, Shung K. 2011. Backscattering
measurement from a single microdroplet. Ultrasonics, Ferroelectrics, and Frequency Control,
IEEE Transactions On 58(4): 874-9.
Liu C, Djuth FT, Zhou Q, Shung KK. 2013. Micromachining techniques in developing
high-frequency piezoelectric composite ultrasonic array transducers. Ultrasonics,
Ferroelectrics, and Frequency Control, IEEE Transactions On 60(12): 2615-25.
Liu C, Zhou Q, Djuth FT, Shung KK. 2012. High-frequency (> 50 MHz) medical
ultrasound linear arrays fabricated from micromachined bulk PZT materials. Ultrasonics,
Ferroelectrics, and Frequency Control, IEEE Transactions On 59(2): 315-8.
Liu R, Harasiewicz KA, Foster FS. 2001. Interdigital pair bonding for high frequency
(20-50 MHz) ultrasonic composite transducers. Ultrasonics, Ferroelectrics, and Frequency
Control, IEEE Transactions On 48(1): 299-306.
Lizzi FL, Feleppa EJ, Alam SK, Deng CX. 2003. Ultrasonic spectrum analysis for tissue
evaluation. Pattern Recognition Letters 24(4): 637-58.
Lizzi FL, Alam SK, Mikaelian S, Lee P, Feleppa EJ. 2006. On the statistics of ultrasonic
spectral parameters. Ultrasound Medicine and Biology 32(11): 1671-85.
154
Lizzi FL, Greenebaum M, Feleppa EJ, Elbaum M, Coleman DJ. 1983. Theoretical
framework for spectrum analysis in ultrasonic tissue characterization. Journal of the
Acoustical Society of America 73(4): 1366-73.
Lockwood GR, Turnbull DH, Foster FS. 1994. Fabrication of high frequency spherically
shaped ceramic transducers. Ultrasonics, Ferroelectrics, and Frequency Control, IEEE
Transactions On 41(2): 231-5.
Lukacs Marc, Sayer Michael and Foster Stuart. 1998. Single-element and linear-array
transducer design for ultrasound biomicroscopy. Medical imaging 1998 International Society
for Optics and Photonics. 272.
Mallery J., Gregory K., Morcos NC, Griffith J. and Henry W. 1987. Evaluation of an
ultrasound balloon dilatation imaging catheter. Circulation 76(4): 371.
Nam K, Zagzebski JA, Hall TJ. 2013. Quantitative assessment of in vivo breast masses
using ultrasound attenuation and backscatter. Ultrasonic Imaging 35(2): 146-61.
Napolitano D, Chou C, McLaughlin G, Ji T, Mo L, DeBusschere D, Steins R. 2006.
Sound speed correction in ultrasound imaging. Ultrasonics 44: 43-6.
Newnham R, Skinner D, Cross L. 1978. Connectivity and piezoelectric-pyroelectric
composites. Material Research Bulletin 13(5): 525-36.
O’Donnell M and Miller J. 1981. Quantitative broadband ultrasonic backscatter: An
approach to nondestructive evaluation in acoustically inhomogeneous materials. Journal of
Applied Physics 52(2): 1056-65.
O'Flynn E, Wilson A, Michell M. 2010. Image-guided breast biopsy: State-of-the-art.
Clinical Radiology 65(4): 259-70.
Oralkan Ö, Cheng C, Johnson J, Karaman M, Lee TH, Khuri-Yakub BT. 2003.
Volumetric ultrasound imaging using 2-D CMUT arrays. Ultrasonics, Ferroelectrics, and
Frequency Control, IEEE Transactions On 50(11): 1581-94.
Parker SH, Burbank F, Jackman RJ, Aucreman CJ, Cardenosa G, Cink TM, Coscia JL,
Jr, Eklund GW, Evans WP, 3rd, Garver PR. 1994. Percutaneous large-core breast biopsy: A
multi-institutional study. Radiology 193(2): 359-64.
Parker SH, Jobe WE, Dennis MA, Stavros AT, Johnson KK, Yakes WF, Truell JE, Price
JG, Kortz AB, Clark DG. 1993. US-guided automated large-core breast biopsy. Radiology
187(2): 507-11.
Pavlin CJ, Harasiewicz K, Sherar MD, Foster FS. 1991. Clinical use of ultrasound
biomicroscopy. Ophthalmology 98(3): 287-95.
155
Pfarl G, Helbich TH, Riedl CC, Wagner T, Gnant M, Rudas M, Liberman L. 2002.
Stereotactic 11-gauge vacuum-assisted breast biopsy: A validation study. American Journal
of Roentgenology 179(6): 1503-7.
Ritter T. 2000. Design, fabrication, and testing of high frequency (>20 MHz) composite
ultrasound imaging arrays.
Ritter T, Shrout TR, Tutwiler R, Shung KK. 2002. A 30-MHz piezo-composite
ultrasound array for medical imaging applications. Ultrasonics, Ferroelectrics, and Frequency
Control, IEEE Transactions On 49(2): 217-30.
Ritter TA, Gerber ER, Shung K, Shrout TR. 2001. Method for manufacture of high
frequency ultrasound transducers. US 6183578 B1.
Savakus H, Klicker K, Newnham R. 1981. PZT-epoxy piezoelectric transducers: A
simplified fabrication procedure. Material Research Bulletin 16(6): 677-80.
Siemens Healthcare - From the X-ray sphere to Molecular MRI - Ultrasound Diagnosis.
2014. https://www.siemens.com/history/en/innovations/healthcare.htm.
Shankar P. 2013. A statistical model for the ultrasonic backscattered echo from tissue
containing microcalcifications. Ultrasonics, Ferroelectrics, and Frequency Control, IEEE
Transactions On 60(5): 932-42.
Shrout Thomas R. and Fielding Jr Joseph. 1990. Relaxor ferroelectric materials. 1990
IEEE Ultrasonics Symposium Proceedings. 711.
Shung KK. 2015. Diagnostic ultrasound: Imaging and blood flow measurements. CRC
Press. Boca Raton.
Shung KK and Thieme GA. 1992. Ultrasonic scattering in biological tissues. CRC Press.
Boca Raton.
Syringe Needle Gauge Chart. 2014. http://www.sigmaaldrich.com/chemistry/stockroom-
reagents/learning-center/technical-library/needle-gauge-chart.html.
Silverstein M. 2009. Where's the outrage? Journal of the American College of Surgeons
208(1): 78-9.
Smith JA. 1996. Transrectal ultrasonography for the detection and staging of carcinoma
of the prostate. Journal of Clinical Ultrasound 24(8): 455-61.
Soo MS, Baker JA, Rosen EL. 2003. Sonographic detection and sonographically guided
biopsy of breast microcalcifications. American Journal of Roentgenology 180(4): 941-8.
156
Stepanishen PR. 1971. The Time‐Dependent force and radiation impedance on a piston
in a rigid infinite planar baffle. Journal of the Acoustical Society of America 49(3B): 841-9.
Stepanishen PR. 1971. Transient radiation from pistons in an infinite planar baffle.
Journal of the Acoustical Society of America 49(5B): 1629-38.
Szabo TL. 2004. Diagnostic ultrasound imaging: Inside out. Academic Press.
Taboada JL, Stephens TW, Krishnamurthy S, Brandt KR, Whitman GJ. 2009. The many
faces of fat necrosis in the breast. American Journal of Roentgenology 192(3): 815-25.
Temno Evolution Biopsy Device. 2014. http://www.carefusion.com/medical-
products/interventional-procedures/biopsy/soft-tissue-biopsy/temno-evolution.aspx.
Thomas III LJ, Barzilai B, Perez JE, Sobel BE, Wickline S, Miller JG. 1989. Quantitative
real-time imaging of myocardium based on ultrasonic integrated backscatter. Ultrasonics,
Ferroelectrics, and Frequency Control, IEEE Transactions On 36(4): 466-70.
Thomenius Kai E. 1996. Evolution of ultrasound beamformers. 1996 IEEE Ultrasonics
Symposium Proceedings. 1615.
Tio T, Cohen P, Coene P, Udding J, den Hartog Jager F, Tytgat G. 1989. Endosonography
and computed tomography of esophageal carcinoma. Gastroenterology 96(6): 1478-86.
Toshiba America Medical Systems. 2008. Microcalcification detection breast ultrasound
- A new perspective - A case study report.
Tresserra F, Feu J, Grases PJ, Navarro B, Alegret X, Férnandez‐Cid A. 1999. Assessment
of breast cancer size: Sonographic and pathologic correlation. Journal of Clinical Ultrasound
27(9): 485-91.
Trots I, Nowicki A, Lewandowski M. 2009. Synthetic transmit aperture in ultrasound
imaging. Archives of Acoustics 34(4): 685-95.
Tsui P, Yeh C, Chang C, Liao Y. 2008. Classification of breast masses by ultrasonic
nakagami imaging: A feasibility study. Physics in Medicine and Biology 53(21): 6027.
Tupholme GE. 1969. Generation of acoustic pulses by baffled plane pistons.
Mathematika 16(02): 209-24.
Velez N, Earnest DE, Staren ED. 2000. Diagnostic and interventional ultrasound for
breast disease. The American Journal of Surgery 180(4): 284-7.
Wang H, Ritter TA, Cao W, Shung KK. 1999. Passive materials for high-frequency
ultrasound transducers. Medical Imaging '99. pp. 35. International Society for Optics and
Photonics.
157
Wild JJ. 1950. The use of ultrasonic pulses for the measurement of biologic tissues and
the detection of tissue density changes. Surgery 27(2): 183-8.
Williams DB, Sahai AV, Aabakken L, Penman ID, van Velse A, Webb J, Wilson M,
Hoffman BJ, Hawes RH. 1999. Endoscopic ultrasound guided fine needle aspiration biopsy:
A large single centre experience. Gut 44(5): 720-6.
Yin Jianhua, Lukacs Marc, Harasiewicz Kasia and Foster Stuart. 2004. Ultra-fine
piezoelectric composites for high frequency ultrasonic transducers. 2004 IEEE Ultrasonics
Symposium Proceedings. 1962.
Yock P, Johnson E, Linker D. 1988. Intravascular ultrasound: Development and clinical
potential. Am J Card Imaging 2(3):185-93.
Youk JH, Kim E, Kim MJ, Lee JY, Oh KK. 2007. Missed breast cancers at US-guided
core needle biopsy: How to reduce them 1. Radiographics 27(1):79-94.
Zhang JY, Ji XM, Zhou J., Huang YP, Xu WJ, Carlier J., Gao JM and Nongaillard B.
2011. A novel method to fabricate full-kerfed highfrequency (> 100 MHz) ultrasonic array
transducers. IEEE 2011 Ultrasonics Symposium. 1739.
Zhou Q, Wu D, Liu C, Zhu B, Djuth F, Shung KK. 2010. Micro-machined high-frequency
(80 MHz) PZT thick film linear arrays. Ultrasonics, Ferroelectrics, and Frequency Control,
IEEE Transactions On 57(10): 2213-20.
Zhu B, Chan NY, Dai J, Shung KK, Takeuchi S, Zhou Q. 2013. New fabrication of high-
frequency (100-MHz) ultrasound PZT film kerfless linear array. Ultrasonics, Ferroelectrics,
and Frequency Control, IEEE Transactions On 60(4): 854-7.
Abstract (if available)
Abstract
Clinical ultrasound imaging has been validated as an effective tool in guiding percutaneous needle procedures including core needle breast biopsy. In fact, image-guided core needle biopsy is the gold standard of breast cancer diagnosis. One current challenge of ultrasound guided breast cancer biopsy is that useful radiographic findings on mammography, including microcalcifications, are difficult to reliably visualize using standard transcutaneous ultrasound imaging systems. The size of these microcalcification structures within the breast suggest that high resolution ultrasound imaging may be able to resolve them. However, an ultrasound imaging probe operating at a significantly higher frequency than standard transcutaneous imaging systems used during ultrasound guided procedures would have a limited imaging depth, thereby necessitating the probe to be miniaturized within a core biopsy needle so that it can be placed immediately adjacent to a lesion within the breast during imaging. ❧ This research has investigated the development of a solution to this clinical problem by fabricating and testing a novel, high frequency, miniaturized ultrasound imaging array as well as performing an ex vivo clinical study to validate this new imaging device may be a viable solution to the clinical problem. The ultrasound array fabrication and assembly process is described and a novel electrical interconnect solution is presented and discussed. Array performance and imaging results as well as imaging results from the clinical study are presented and discussed. Lastly, future work for this project is proposed and discussed including improvements to the miniaturized, high frequency array and electrical interconnect design.
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
High frequency ultrasonic imaging for the development of breast biopsy needle with miniature ultrasonic transducer array
PDF
Miniature phased-array transducer for colorectal tissue characterization during TEM robotic surgery; and, Forward-looking phased-array transducer for intravascular imaging
PDF
Array transducers for high frequency ultrasound imaging
PDF
Synthetic aperture imaging platform based on CMOS high voltage 1 to 64 multiplexer / de-multiplexer for ultrasound guided breast biopsy needle
PDF
Development of high-frequency (~100mhz) PZT thick-film ultrasound transducers and arrays
PDF
High-frequency ultrasound array-based imaging system for biomedical applications
PDF
High frequency ultrasonic phased array system and its applications
PDF
Development of novel 1-3 piezocomposites for low-crosstalk high frequency ultrasound array transducers
PDF
Design and development of ultrasonic array transducers for specialized applications
PDF
Single-cell analysis with high frequency ultrasound
PDF
Novel beamforming with dual-layer array transducers for 3-D ultrasound imaging
PDF
A high frequency array- based photoacoustic microscopy imaging system
PDF
High-frequency ultrasound imaging system with Doppler features for biomedical applications using 30~35 mHz linear arrays and an analog beamformer
PDF
Intravascular imaging on high-frequency ultrasound combined with optical modalities
PDF
Development of front-end circuits for high frequency ultrasound system
PDF
Lesion enhancement for three dimensional rectilinear ultrasound imaging
PDF
Highly integrated 2D ultrasonic arrays and electronics for modular large apertures
PDF
Development of high frequency focused transducers for single beam acoustic tweezers
PDF
Microfluidic cell sorting with a high frequency ultrasound beam
PDF
Development of back-end processing system for high frequency ultrasound b-mode imaging
Asset Metadata
Creator
Cummins, Thomas
(author)
Core Title
High frequency ultrasound array for ultrasound-guided breast biopsy
School
Viterbi School of Engineering
Degree
Doctor of Philosophy
Degree Program
Biomedical Engineering
Publication Date
03/15/2016
Defense Date
10/28/2015
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
breast biopsy needle,DRIE,Guidance,high frequency,imaging,micromachining,miniaturized,OAI-PMH Harvest,transmission line,ultrasound array,ultrasound transducer
Format
application/pdf
(imt)
Language
English
Contributor
Electronically uploaded by the author
(provenance)
Advisor
Shung, K. Kirk (
committee chair
), Martin, Sue E. (
committee member
), Yen, Jesse (
committee member
)
Creator Email
tcummins07@gmail.com
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-c40-221491
Unique identifier
UC11277247
Identifier
etd-CumminsTho-4205.pdf (filename),usctheses-c40-221491 (legacy record id)
Legacy Identifier
etd-CumminsTho-4205.pdf
Dmrecord
221491
Document Type
Dissertation
Format
application/pdf (imt)
Rights
Cummins, Thomas
Type
texts
Source
University of Southern California
(contributing entity),
University of Southern California Dissertations and Theses
(collection)
Access Conditions
The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...
Repository Name
University of Southern California Digital Library
Repository Location
USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA
Tags
breast biopsy needle
DRIE
high frequency
imaging
micromachining
miniaturized
transmission line
ultrasound array
ultrasound transducer