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 imaging system with Doppler features for biomedical applications using 30~35 mHz linear arrays and an analog beamformer
(USC Thesis Other)
High-frequency ultrasound imaging system with Doppler features for biomedical applications using 30~35 mHz linear arrays and an analog beamformer
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
HIGH-FREQUENCY ULTRASOUND IMAGING SYSTEM WITH DOPPLER
FEATURES FOR BIOMEDICAL APPLICATIONS USING 30~35 MHZ LINEAR
ARRAYS AND AN ANALOG BEAMFORMER
by
Xiaochen Xu
A Dissertation Presented to the
FACULTY OF THE GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(BIOMEDICAL ENGINEERING)
May 2007
Copyright 2007 Xiaochen Xu
ii
ACKNOWLEDGEMENTS
I first wish to acknowledge the members of my guidance committee. I wish to
express my special thanks to Dr. K. Kirk Shung, for his supervision and support
during the past five years. I also wish to thank Dr. Jesse Yen and Dr. Jonathan
Cannata for their many valuable suggestions when I was facing difficulties in my
research. Finally, I want to thank Dr. Richard Leahy and Dr. Ellis Meng for their
indispensable advices for this dissertation.
I also want to acknowledge the financial support provided by the National
Institute of Health, Dr. K. Kirk Shung, Dr. Mark S. Humayun, and the Doheny Eye
Institute during the period of finishing this dissertation.
I sincerely thank the collaborators in the NIH Resource on Medical Ultrasonic
Transducer Technologies, the Doheny Eye Institute, and the Institute for Genetic
Medicine at USC. Without their collaborations and assistances, it is impossible for
me to complete this dissertation. They include: Dr. Qifa Zhou, Dr. Lei Sun, Dr.
Ruibin Liu, Mr. Jay Williams, Dr. Jianzhong Zhao, Dr. Bin Huang, Dr. Emanuel
Gottlieb, Dr. Changhong Hu, Dr. Jung-woo Lee, Ms. Rachel Bitton, Mr. Jin-Ho
Chang, Mr. Hyung-Ham Kim, Mr. Dawei Wu, Mr. Chih-Chung Huang, Mr. Joe Han,
Mr. Peter Lee, Mr. Bruce Lai, Mr. Charles Sharp, Ms. Suvimol Sangkatumvong, Dr.
Hossein Ameri, Mr. Charles DeBoer, Dr. Laurence Kedes, and Ms. Yan Bai.
I want to thank Dr. Sharon Myers and Ms. Mary Ann Murphy for polishing this
dissertation.
iii
Finally, I want to thank my family for their unconditional support and love in the
past thirty years, especially my parents, my sister, and my grandparents. Last but not
the least, I want to thank my wife Qi, who has been living and studying alone in
Texas for five years, for her patience, understanding, encouragement, and
unconditional support and love.
iv
TABLE OF CONTENTS
ACKNOWLEDGEMENTS .........................................................................................ii
LIST OF FIGURES ...................................................................................................vii
LIST OF TABLES ......................................................................................................xi
ABSTRACT...............................................................................................................xii
CHAPTER 1 ................................................................................................................1
INTRODUCTION .......................................................................................................1
MEDICAL ULTRASOUND............................................................................................1
ULTRASONIC TRANSDUCER.......................................................................................2
ULTRASOUND IMAGING.............................................................................................4
ULTRASOUND DOPPLER ............................................................................................5
A. Continuous-wave Ultrasound Doppler.........................................................6
B. Pulsed-wave Ultrasound Doppler................................................................8
C. Color Doppler............................................................................................10
HIGH-FREQUENCY ULTRASOUND ............................................................................11
A. High-frequency B-mode imaging ..............................................................11
B. High-frequency Ultrasound Doppler..........................................................12
C. High-frequency Ultrasonic Linear Arrays .................................................13
DISSERTATION STRUCTURE.....................................................................................14
CHAPTER 2 ..............................................................................................................16
HIGH-FREQUENCY HIGH-FRAME RATE ULTRASOUND SYSTEM FOR
SMALL ANIMAL IMAGING WITH LINEAR ARRAYS ......................................16
INTRODUCTION........................................................................................................16
METHODS................................................................................................................17
A. Beamformers in Ultrasound System ..........................................................17
B. Digital Beamformers vs. Analog Beamformer ..........................................20
C. High-frequency Beamformer Design.........................................................22
SYSTEM DESCRIPTION.............................................................................................22
A. Array Characteristics..................................................................................24
B. Microprocessor Timing Control Board......................................................24
C. 64 channel Transceiver Board....................................................................25
D. 16-channel Amplification Board................................................................28
E. Receive Beamformer Board.......................................................................29
F. The Complete System ................................................................................31
v
EXPERIMENTS AND RESULTS...................................................................................33
A. Phantom Tests............................................................................................33
B. In vitro Experiments...................................................................................34
C. In vivo Experiments ...................................................................................36
SUMMARY...............................................................................................................37
CHAPTER 3 ..............................................................................................................38
IMAGE QUALITY IMPROVEMENTS ...................................................................38
INTRODUCTION........................................................................................................38
METHODS................................................................................................................39
A. Efficient Bipolar Pulse Generator ..............................................................39
B. Low-noise Front-end Electronics...............................................................49
C. Grounding Isolation and PCB Layout Considerations...............................53
D. High-resolution A/D..................................................................................54
SUMMARY...............................................................................................................56
CHAPTER 4 ..............................................................................................................57
HIGH-FREQUENCY PLUSED-WAVE DOPPLER SYSTEM WITH PMN-PT
NEEDLE TRANSDUCERS FOR OPHTHALMOLOGY APPLICATIONS...........57
INTRODUCTION........................................................................................................57
SYSTEM DESCRIPTION.............................................................................................62
A. Transducer Fabrication...............................................................................62
B. Micro-flow Phantom..................................................................................68
C. System Design............................................................................................69
D. Flow Estimation.........................................................................................73
EXPERIMENTS AND RESULTS...................................................................................75
A. Minimal Detectable Velocity Measurements.............................................75
B. Micro-flow Phantom Evaluation................................................................77
C. In vivo Rabbit Experiments .......................................................................81
SUMMARY...............................................................................................................87
CHAPTER 5 ..............................................................................................................88
HIGH-FREQUENCY ULTRASOUND PULSED-WAVE DOPPLER SYSTEM
FOR BIOMEDICAL APPLICATIONS WITH 30 MHz LINEAR ARRAY ............88
INTRODUCTION........................................................................................................88
SYSTEM DESCRIPTION.............................................................................................89
A. Array Characteristics..................................................................................92
B. Timing Circuits..........................................................................................92
C. Bipolar Pulsers and Analog Front-ends .....................................................94
D. HF Pulsed-wave Doppler Block ................................................................96
EXPERIMENTS AND RESULTS...................................................................................97
C. Micro-flow Phantom Tests.......................................................................101
D. In vivo Studies .........................................................................................103
vi
E. In vitro B-mode Imaging Studies.............................................................108
SUMMARY.............................................................................................................113
CHAPTER 6 ............................................................................................................114
HIGH-FREQUENCY COLOR DOPPLER AND POWER DOPPLER WITH
ARRAYS .................................................................................................................114
INTRODUCTION......................................................................................................114
METHODS..............................................................................................................115
A. Clutter Signal Rejection ...........................................................................116
B. Autocorrelation Based Flow Velocity Estimation ...................................118
C. Velocity Mapping and Power Doppler ....................................................121
D. System Setup............................................................................................122
EXPERIMENTS AND RESULTS.................................................................................123
A. Single Vessel Detection ...........................................................................124
B. Multiple Vessels Detection ......................................................................127
SUMMARY.............................................................................................................129
CHAPTER 7 ............................................................................................................130
FUTURE WORK AND CONCLUSION ................................................................130
FUTURE WORK......................................................................................................130
A. Better Lateral Resolution and Wider Field of View ................................130
B. Duplex Imaging System with Higher Frame Rate ...................................131
C. 2-D Color Doppler ...................................................................................134
D. In vivo Microcirculation Studies Enhanced by Contrast Agents.............135
E. Digital Beamforming and Doppler Processing ........................................136
CONCLUSION.........................................................................................................137
BIBLIOGRAPHY....................................................................................................139
vii
LIST OF FIGURES
Figure 1.1 Block diagram of a typical ultrasound system............................................1
Figure 1.2 Photos of single element and array transducers..........................................3
Figure 1.3 Diagrams of mechanical and electronic scanning. .....................................4
Figure 1.4 A-mode image line (a) and B-mode image of mouse superficial
abdominal vessels (b). ................................................................................5
Figure 1.5 The setup and block diagram of the continuous-wave Doppler system. ....7
Figure 1.6 The setup of the pulsed-wave Doppler system. ..........................................9
Figure 1.7 Base band signals (a) and extracted Doppler signals (b) in the pulsed-
wave system................................................................................................9
Figure 1.8 A color Doppler image shows internal carotid stenosis............................10
Figure 2.1 Focused acoustic beam by applying appropriate time delays...................18
Figure 2.2 Delay time calculations based on the geometry structures of an array.....19
Figure 2.3 Electronically scanned and focused acoustic beams in a linear array. .....19
Figure 2.4 Block diagrams of an analog beamformer (a) and a digital
Beamformer (b). .......................................................................................21
Figure 2.5 Block diagram of the HF imaging system with arrays. ............................23
Figure 2.6 Interconnections among the transducer, the multiplexers, the
cross-point switch and the analog receive beamformer............................26
Figure 2.7 The unipolar pulse (solid line) and its corresponding spectrum
(dashline). .................................................................................................27
Figure 2.8 Block diagrams of the HF programmable analog delay circuit (a)
and the summation circuit (b) in the analog receive beamformer. ...........30
Figure 2.9 Calculated (lines) and measured (solid triangle) time delays of each
channel in the analog beamformer............................................................31
viii
Figure 2.10 Photos of the unpacked system (a) and packed system (b). ...................32
Figure 2.11 Tissue and wire phantom images............................................................34
Figure 2.12 Rabbit eye images obtained by Transducer 1. ........................................35
Figure 2.13 Rabbit eye images obtained with Transducer 2. .....................................35
Figure 2.14 Mouse heart images. ...............................................................................36
Figure 3.1 Schematic of the timing control circuit for the bipolar pulser..................41
Figure 3.2 Schematic of the bipolar pulser.................................................................41
Figure 3.3 The timing signals and output of the bipolar pulser. ................................41
Figure 3.4 Simulated monocycle bipolar pulse..........................................................42
Figure 3.5 Photo of the prototype bipolar pulser........................................................43
Figure 3.6 Experimental results of the bipolar pulser with a 50 Ω load. ...................45
Figure 3.6 Comparison between the bipolar pulser and the Panametrics 5900PR. ...48
Figure 3.8 Block diagram of measuring the minimal detectable signal.....................49
Figure 3.9 Block diagrams of AD8332 applications.................................................. 51
Figure 3.10 Measured characteristics of AD8332......................................................52
Figure 3.11 Comparison between AD8332 and Miteq AU-1313...............................52
Figure 3.12 Zebra fish images with different dynamic ranges...................................56
Figure 4.1 Eye anatomy..............................................................................................59
Figure 4.2 Overview of microsurgerical instruments inserted through sclera. ..........59
Figure 4.3 Vein locations in CRVO (a) and BRVO (b)..............................................63
Figure 4.4 Diagrams of different probes for CRVO (a) and BRVO (b)
measurements. ..........................................................................................63
Figure 4.5 Photos of the needle transducer (a), and detailed 0° and 45° tips (b).......65
ix
Figure 4.6 Measured needle transducer properties. ..................................................66
Figure 4.7 Measured lateral profile of the needle transducer at 2 mm depth.............67
Figure 4.8 Photo of the micro-flow phantom.............................................................68
Figure 4.9 Block diagram (a) and photo (b) of the HF directional pulsed-wave
Doppler system.........................................................................................70
Figure 4.10 Timing sequences in the HF directional pulsed-wave Doppler system..72
Figure 4.11 Minimal detectable velocity measured from the wire phantom. ............76
Figure 4.12 Detected Doppler signals from the micro-flow phantom. ......................78
Figure 4.13 Velocity variance measurements on the micro-flow phantom................80
Figure 4.14 Measured retinal blood flows with the 0° needle transducer..................82
Figure 4.15 Measured the blood flows in CRV and CRA with the 45° needle
transducer................................................................................................84
Figure 4.16 Blood flow velocity estimation error caused by human positioning. .....86
Figure 5.1 Block diagram (a) and photo (b) of the HF ultrasound Doppler with
array. .........................................................................................................91
Figure 5.2 Timing sequences in the HF Doppler system with 30 MHz array............94
Figure 5.3 Measured 7-cycle 30 MHz bipolar pulse and its spectrum.......................95
Figure 5.4 Block diagram of one channel analog front-end circuit. ..........................96
Figure 5.5 Measured lateral resolutions of the 30 MHz array....................................99
Figure 5.6 Measured minimal detectable velocity with the wire phantom..............101
Figure 5.7 Measured Doppler signals from the micro-flow phantom......................102
Figure 5.8 B-mode image of abdomen superficial vessels. .....................................104
Figure 5.9 Measured Doppler signals from the abdomen superficial vessels..........105
Figure 5.10 In vivo Doppler studies on mice with the 30 MHz array. .....................107
x
Figure 5.11 Wire phantom images obtained by the array Doppler system (a) and
the 35 MHz UBM (b). .......................................................................... 110
Figure 5.12 Pig eye images in vitro.......................................................................... 111
Figure 5.13 Rabbit eye images in vitro. ................................................................... 112
Figure 6.1 Block diagram of a color Doppler imaging system................................ 116
Figure 6.2 Block diagram of the single echo canceller. ........................................... 117
Figure 6.3 Frequency response of the single echo canceller.................................... 117
Figure 6.4 IQ signals used for calculating the angular Doppler shift frequency......120
Figure 6.5 Block diagram of the system with 30 MHz array for color Doppler
studies. ....................................................................................................122
Figure 6.6 M-mode color Doppler image of a mouse vessel (a) and its
corresponding pulsed-wave Doppler spectrogram at the depth of
8 mm.......................................................................................................125
Figure 6.7 Power Doppler image (a) and plot (b) at the time of 0.33s. ...................126
Figure 6.8 M-mode color Doppler image (a) and B-mode image of two vessels....128
Figure 7.1 Block diagram of the proposed new system (a) and its mother-
daughter boards’ structure (b).................................................................132
Figure 7.2 Block diagram of RAMs based control structure (a) and
corresponding timing sequences (b).......................................................134
xi
LIST OF TABLES
TABLE 2.1 Characteristics of the linear arrays used in this study ............................24
TABLE 4.1 Characteristics of PMN-PT single crystal. .............................................63
TABLE 5.1 Measured characteristics of the 30 MHz array. ......................................92
TABLE 5.2 Lateral resolutions of the array with 16-channel beamformer................99
xii
ABSTRACT
Small animal research is gaining more attention since mice and rats are preferred
animal models for gene and drug therapy. High-frequency (HF) ultrasound imaging,
capable of achieving superior spatial resolution at an affordable price, has been
shown to be useful for imaging and visualizing blood flow in small animals for
biological and pharmaceutical research. The utilization of HF arrays in a HF imaging
system can alleviate the limitations of the current systems with single element
transducers, such as the uneven image quality and frame rate.
In this dissertation, a HF ultrasound B-mode imaging system and a HF
ultrasound Doppler system using 30~35 MHz linear arrays and an analog
beamformer were investigated, visualizing both the movements and blood flow of
the mouse heart. First the B-mode imaging system, which included HF linear arrays,
timing control circuits, analog front-end circuits, a 16-channel analog receiving
beamformer and an 8-bit digitizer, was developed. The system with the 48-element
linear arrays was used to image rabbit eyes in vitro and mice heart in vivo in real
time. Furthermore, this system was used to successfully acquire images of mouse
heart movements at 100 frame/s. For further improving the performance of the B-
mode imaging system, efficient bipolar pulsers, low-noise front-end electronics, and
a 14-bit digitizer were studied. Meanwhile, a HF pulsed-wave Doppler system,
which can be interfaced to either a single element transducer or the analog
beamformer with arrays, was developed. The Doppler system was capable of
xiii
detecting blood flow in vessels with diameters less than 200 µm and measuring
velocities from 100 µm/s to 1 m/s. In addition, this Doppler system with intraocular
needle transducers can be used to evaluate the treatment of retinal vein occlusion.
Finally, based on the Doppler system and part of the improved B-mode imaging
system, an array-based HF Doppler system, which can perform pulsed-wave Doppler,
color Doppler, and power Doppler measurements, was implemented and used in
small animal cardiovascular research. The experimental results have demonstrated
that the HF ultrasound system with the arrays and analog beamformer may achieve
better performance and higher frame rate than current systems with single element
transducers.
1
CHAPTER 1
INTRODUCTION
Medical Ultrasound
Ultrasound imaging is one of the most important medical imaging methods due
to its safety, cost effectiveness and real-time capability. Conventional ultrasonic
imaging systems use frequencies from 2 to 15 MHz with a resolution of millimeter
level. They have been widely used in monitoring fetuses, as well as diagnosing
diseases in the internal organs, such as the heart, liver, gallbladder, spleen, pancreas,
kidneys, and bladder (Shung 2005). Current ultrasound systems can visualize both
internal organs with B-mode imaging and blood flow with ultrasound Doppler. A
block diagram of a typical ultrasound system is illustrated in Fig. 1.1.
Figure 1.1 Block diagram of a typical ultrasound system.
2
As shown in Fig. 1.1, the main components in an ultrasound system include a
central control unit, either electronically scanned array (or mechanically scanned
single-element) transducers, high-voltage transmitting circuits with a transmit
beamformer, low-noise receive circuits with a receiving beamformer, Doppler
processing units, and a B-mode image processing unit. The processed B-mode
images and Doppler images can be displayed separately or superimposed onto each
other, depending on the objects to be diagnosed. The principles of B-mode imaging
and Doppler ultrasound will be discussed in this chapter.
Ultrasonic Transducer
The most important component in an ultrasound system is the ultrasonic
transducer, which consists of a piezoelectric element and other supporting structures.
The transducer’s performance primarily determines the performance of a whole
ultrasound system. In Fig. 1.1, the transducer is linked to both the high-voltage
transmit (Tx) amplifiers (AMPs) and the low-noise receive (Rx) amplifiers through
the high-voltage multiplexers and de-multiplexers. The transducer, which is excited
by the high-voltage Tx AMPs, converts electrical energy to acoustic energy due to
the piezoelectric effect. Therefore, the transducer transmits acoustic waves to a
medium, which can be human tissue, liquids or solid structures depending on
applications. Due to the inhomogeneity in a medium, the acoustic waves are
reflected or scattered and received by the transducer again. The transducer converts
3
the returned acoustic waves to corresponding electrical signals, which are amplified
and processed by other units in the system as shown in Fig. 1.1.
Both single element transducers and array transducers are used in ultrasound
systems, depending on cost-performance issues and application areas. Ultrasound
systems with mechanically scanned single element transducers illustrated in Fig. 1.2
(a) and Fig. 1.3(a) are inexpensive with sacrificing the image quality and frame rate.
Better image quality and higher frame rates can be accomplished by utilizing array
transducers and beamformers to scan and focus the acoustic beam electronically as
illustrated in Fig. 1.2(b) and Fig. 1.3(b). Naturally, the cost of ultrasound systems
with arrays increases dramatically, because array transducers usually have more than
64 elements and the corresponding electronic systems have numerous channels. The
design and fabrication of ultrasonic transducers has been reviewed in recent books
and papers (Shung and Zipparo 1996; Shung 2005). Further studies on this topic are
ongoing as described in recent papers (Ritter et al. 2002; Cannata et al. 2003;
Cannata et al. 2006).
(a) (b)
Figure 1.2 Photos of single element and array transducers. 40~60 MHz light-weight
single-element transducers (a) and GE 3-8 MHz array transducer (b).
Connector
Array
4
(a) (b)
Figure 1.3 Diagrams of mechanical and electronic scanning. Mechanically scanned
single-element transducer (a) and electronically scanned linear array (b).
Ultrasound Imaging
After a transducer receives the echoes, appropriate processing units should be
implemented in order to convert these signals to interpretable information for
sonographers or other end-users. In the earliest ultrasound systems, useful clinical
information was obtained by displaying the amplitude of echoes as a function of
time-of-flights of these echoes. This is the A-mode (amplitude-mode) ultrasound
imaging system, which displays 1-D scan lines as shown in Fig. 1.4(a). Thus only
limited information is available to the sonographers. More information can be
obtained from 2-D grayscale or color images.
In order to construct 2-D images, the acoustic beam of a transducer needs to be
scanned over a certain region, and multiple A-mode scan lines are obtained during
the scanning. These scan lines constitute one frame of image on which the echo
amplitudes in the scan lines are mapped to the pixels’ brightness level in a linear or
5
nonlinear manner. Real-time images can be achieved when the acoustic beam of the
transducer scans fast enough. These images are called as B-mode (brightness-mode)
images. The first commercial B-mode ultrasound imaging systems appeared in the
1970s (Shung 1992). Since then, it has been an accepted medical image modality. To
date, more and more novel ultrasound images appear on the latest commercial
systems, such as 3-D and 4-D images (Shung 2005). Nevertheless, 2-D B-mode
image is still the most widely-used imaging features, because it is the basis for
producing 3-D and 4-D images.
(a) (b)
Figure 1.4 A-mode image line (a) and B-mode image of mouse superficial abdominal
vessels (b).
Ultrasound Doppler
In addition, most clinical ultrasound systems include another indispensable
feature: ultrasound Doppler, which is based on the Doppler effect. The Doppler
effect describes that the wavelength of a wave is not a constant due to motion of the
medium. If a wave is emitted from a source moving away from an observer, its
wavelength increases, and vice versa. Therefore as the acoustic wave propagating in
6
human body and being reflected by moving objects which include the tissue, organ
and blood, the wavelengths of the transmitted pulses and received echoes (i.e. center
frequencies) are different. This frequency difference is the Doppler shift frequency
which can be used to calculate the velocities of the moving objects:
c
f
f
v
θ cos 2
0
∆
= (1.1)
where f ∆ is the Doppler shift frequency, f
0
is the center frequency of the transmitted
pulses, c is the sound velocity in a medium, and θ is the angle between the
ultrasound beam and the moving objects.
Ultrasound Doppler
has been used in medical applications since the late 1950s.
Currently ultrasound Doppler is used not only to evaluate blood flows, but also to
study tissue movements (Evans and McDicken 2000). In the past 50 years, several
Doppler techniques were developed to provide different sets of information to their
users. In this chapter, three most popular Doppler techniques in ultrasound, including
continuous-wave (CW) Doppler, pulsed-wave (PW) Doppler and color Doppler, will
be introduced, as well as their advantages and disadvantages.
A. Continuous-wave Ultrasound Doppler
The continuous-wave Doppler is the earliest ultrasound Doppler technique, and it
can be implemented by extracting the exact shifted Doppler frequency in the
received echoes. Its experimental setup is shown in Fig. 1.5.
7
(a) (b)
Figure 1.5 The setup and block diagram of the continuous-wave Doppler system.
Two transducers Tx and Rx are used. Tx is for transmitting CW ultrasound, and
Rx is for receiving echoes from the moving objects. Tx transmits a cosine wave to a
medium, and Rx receives the shifted cosine signal described by:
) cos(
1
ϕ ω + = t Tx
c
(1.2)
] ) cos[(
2
ϕ ω ω + + = t Rx
d c
(1.3)
where ω
c
is the center frequency of a transducer and ω
d
is the Doppler shift
frequency introduced by the moving objects. The Doppler shift frequency ω
d
can be
extracted by demodulation with a mixer as:
) 2 cos( ) cos(
) cos( ] ) cos[(
1 2 1 2
1 2
ϕ ϕ ω ω ϕ ϕ ω
ϕ ω ϕ ω ω
+ + + + − + =
+ × + + =
t t t
t t f
d c d
c d c d
(1.4)
The harmonic component in the above equation can be filtered out by a low-pass
filter. Then the signals containing Doppler shift frequencies are amplified to audio
outputs.
In the continuous-wave Doppler system, any Doppler shift frequency can be
extracted; no matter it is high or low. Thus it can be used to detect extremely high
velocity without any aliasing problem which commonly exists in pulsed-wave
Rx Tx
Transmitter
Receiver LPF+AMP
Mixer Rx
Tx
8
Doppler and color Doppler systems. As a result, the continuous-wave Doppler is still
used for detecting high-speed blood flows in the cardiovascular circulation. In
addition, the continuous-wave Doppler can be built with simple circuits, but it has
poor spatial resolution due to transmitting continuous waves and can only detect the
average flow velocities in a wide sample volume which is the shaded area in Fig.
1.5(a). This sample volume depends on the crossed transducer focal zones. In order
to improve its spatial resolution, the pulsed-wave Doppler was developed in the late
1960s.
B. Pulsed-wave Ultrasound Doppler
The implementation of pulsed-wave Doppler is based on demodulation and
sample-and-hold techniques (Jensen 1996). The experimental setup for a pulsed-
wave Doppler system is shown in Fig. 1.6. Only one transducer is required and the
shaded area illustrates the sample volume which is determined by the spatial
resolutions of the transducer (Jensen 1996). Typically the transducer transmits a 4-
16 cycle tone burst cosine signal at a particular period repetition frequency (PRF)
and receives the reflected signal. The received signals are amplified and
demodulated by a cosine waveform with the transducer center frequency of ω
c
. Since
the received signals are scattered by moving particles in the flow, the demodulated
signal is shifted in time domain differently compared to its neighbor signals as
Figure 1.7(a) shows (Jensen 1996). At a certain depth as the dash line illustrated in
Fig. 1.7(a), the amplitudes of the demodulated signals are different. Those
9
amplitudes are plotted in Fig. 1.7(b). Using a low-pass filter to smooth the discrete
waveform in Fig. 1.7(b), the Doppler shift signal is obtained as well.
Figure 1.6 The setup of the pulsed-wave Doppler system.
(a) (b)
Figure 1.7 Base band signals (a) and extracted Doppler signals (b) in the pulsed-wave
system.
Compared to the continuous-wave Doppler system, the pulsed-wave Doppler
requires more complex circuits and allows detecting flow velocities at a specific
Sample
Volume
Base Band Signals Doppler Signals
Sample & Hold
Position
10
depth with only one transducer which is often shared with the one used in B-mode
imaging. In additional, the Doppler features can be acquired on a B-mode imaging
platform by modifying digital signal processing software. In Chapter 4, the
implementation of a pulsed-wave Doppler system will be discussed in detail.
C. Color Doppler
Visualizing 2-D blood flows in real time is extremely useful in diagnosing
cardiovascular diseases, such as the occlusion of vessels and heart valve
regurgitation. A typical color Doppler image which shows the carotid stenosis is
illustrated in Fig. 1.8.However, both the pulsed-wave Doppler and continuous-wave
Doppler are only capable of detecting flows in a limited sample volume, independent
of a wide sample volume or a narrow sample volume.
Figure 1.8 A color Doppler image shows internal carotid stenosis. (Courtesy of GE)
Originally, multiple-gate pulsed-wave Doppler systems was used to produce the
2-D flow mapping; nevertheless, real-time mapping was impossible for these
systems due to the large number of individual sample volumes. So real-time flow
mapping has to be achieved with other approaches. Since 1970, several approaches
11
have been studied and utilized to map flows in real time. Two most commonly used
algorithms include autocorrelation in phase domain and cross-correlation in time
domain. The phase domain algorithm uses in-phase and quadrature (IQ) demodulated
signals, and the time domain one uses RF signals respectively. Based on the
autocorrelation method, the first commercial color Doppler imaging system was
introduced by Aloka Co. in 1982 (Evans and McDicken 2000; Kasai 1985).
Thereafter, commercial color Doppler imaging systems using the cross-correlation
appeared in the 1980s (Evans and McDicken 2000). At present, color Doppler is still
a very active research area, and new algorithms are being developed to better
visualize flows. In Chapter 6, we will discuss the implementation of color Doppler in
more detail.
High-frequency Ultrasound
A. High-frequency B-mode imaging
Conventional ultrasonic imaging systems as mentioned above use frequencies
from 2 to 15 MHz with a resolution of millimeter level. Better spatial resolutions can
be achieved by increasing operational frequencies, since the lateral (R
L
) and axial (R
A
)
resolutions can be calculated respectively as:
λ
r
Z
f
c
f R
f
L
2
#
= = (1.5)
dB A
c
R
6
2
−
= τ (1.6)
12
where c is the sound velocity in a medium, Z
f
is the focal distance, r is the aperture
size of a transducer, and τ
-6dB
is the –6 dB duration time of an emitted pulse. In
addition, τ
-6dB
is proportional to the cycle number of an emitted pulse. Therefore,
both the lateral and axial resolutions are proportional to the operational frequency of
a transducer. For example, at 50 MHz, the lateral resolution for a transducer with f
#
of 2.5 would be better than 75µm (Lockwood et al. 1996). State-of-the-art high-
frequency ultrasound systems can obtain a spatial resolution better than 15 µm
(Hozumi et al. 2004). However, the depth of penetration would be limited to less
than 1 cm due to the high attenuation in biological tissues.
As a result, high-frequency ultrasound is more suitable in applications requiring
higher resolution but less penetration. Many clinical applications for high-frequency
B-mode imaging were reported by Lockwood et al. (1996). In ophthalmology, high
frequency scanners have been used to assess anterior segment tumors, segment
lesions, and types of glaucoma. In dermatology, typical applications of high-
frequency ultrasound include tumor assessments and dynamic studies of structures in
the skin. High-frequency B-mode imaging has also been used in studies of small
animals like mice and rats (Foster
et al. 2002). Commercial ultrasound backscatter
microscopes (UBMs) have been available, and further studies on high-frequency
ultrasound imaging are ongoing.
B. High-frequency Ultrasound Doppler
Similar to B-mode imaging, the detection limit of conventional frequency
Doppler ultrasound currently extends to only the level of arterioles. Both the spatial
13
and velocity resolutions of conventional Doppler systems remain inadequate for
studying microcirculations, for example, blood flows in human eyes and in skin
tumors (Christopher et al. 1996). By increasing the ultrasound frequency to 50 MHz,
the current limitations of conventional frequency Doppler for detecting blood flow in
the capillaries may be overcome from calculations with (1.1), (1.5) and (1.6).
Recent studies have shown that high-frequency Doppler systems were capable of
detecting the blood velocities in capillaries (0.5 mm/s) and arterioles (5 mm/s)
(Christopher et al. 1996; Christopher et al. 1997). High-frequency ultrasound
Doppler has been used for detecting blood flow in the anterior segment of the eye
(Pavlin et al. 1998). Moreover, high-frequency color Doppler provides useful
information of the microcirculation in tumor research. Therefore developing both
high-frequency B-mode imaging and high-frequency ultrasound Doppler can
significantly improve image quality and provide more information in dermatology,
ophthalmology and small animal studies.
C. High-frequency Ultrasonic Linear Arrays
At present, most of high-frequency ultrasound backscatter microscopes (UBMs)
obtain images by scanning a single element transducer mechanically (Foster et al.
1993). The mechanical nature and fixed focal point of UBMs limits the frame rate
and resolution of UBMs in the field of view. For example, the mouse heart is about
7~10 mm in long-axis and 4~8 mm in short-axis and the heart rate is greater than 400
beats per minute (Hedrich and Bullock 2004)
. Imaging modalities with high frame
rate capability (>200 Frames/sec) and wide field of view (10 mm both in axial and
14
lateral directions) are required for cardiovascular research utilizing mice (Li et al.
2005). At present, these specifications cannot be accomplished by using single
element transducers. As in conventional ultrasound, an array-based high-frequency
ultrasound imaging system can alleviate the mentioned limitations of UBMs.
Recently, high-frequency ultrasonic arrays (>30 MHz), which provide clinical
convenience, reduce imaging time, and offer dynamic focusing, have been developed
successfully (Ritter et al. 2002; Cannata et al. 2006). These arrays have more than
48-elements with a pitch ranging from 50 µm to 100 µm. With these arrays, higher
frame rate and better image quality can be achieved compared to current UBMs.
Moreover, the larger arrays with more than 128 elements and less than 50 µm pitch
are being developed and a superior image quality may be expected in the future.
In order to utilize the developed arrays, an imaging system supporting more than
64-channel electronics is required. However, imaging systems for these high-
frequency arrays are not yet commercially available although extensive
investigations are underway to develop them (Brown and Lockwood 2002; Stitt et al.
2002; Hu et al. 2006).
Dissertation Structure
The development of a high-frequency array-based imaging system with Doppler
features and high frame rate capabilities is the main focus of this dissertation.
In Chapter 2, a high-frequency high-frame rate imaging system is described.
Then several improvements are discussed in Chapter 3 for achieving better image
15
quality and detecting weak Doppler signals from blood. Chapter 4 introduces the
implementation of a high-frequency pulsed-wave Doppler system with intraocular
transducers for ophthalmological applications. By combining the improved imaging
system described in Chapter 2 and the Doppler system in Chapter 4, a high-
frequency array-based imaging system with Doppler features is presented in Chapter
5. In Chapter 6, initial studies on high-frequency color Doppler and power Doppler
with arrays are introduced. Finally, future work is proposed in Chapter 7, as well as a
summary of this dissertation.
16
CHAPTER 2
HIGH-FREQUENCY HIGH-FRAME RATE ULTRASOUND
SYSTEM FOR SMALL ANIMAL IMAGING WITH LINEAR
ARRAYS
Introduction
In current evaluations of gene and drug therapies, small animals like mice and
rats are preferred animal models (Foster et al. 2002). Due to their size, Micro-CT,
Micro-MRI and Micro-PET have been developed and used for small animal imaging
(Hwang et al. 2002). Other imaging modalities which can achieve high-resolution
images and visualize blood flows in vascular circulation, such as ultrasound, are in
high demand in small animal research.
High-frequency (HF) ultrasound imaging, capable of achieving good spatial
resolution at an affordable price, has been shown to be useful for imaging small
animals in biological and pharmaceutical research (Lockwood et al. 1996; Foster et
al. 2000). The frequency range of HF ultrasound has been extended to over 100 MHz.
State-of-the-art HF ultrasound systems can obtain a spatial resolution better than 15
µm (Hozumi et al. 2004) and detect blood velocities less than 0.5 mm/s in capillaries
as small as 20 µm in diameter (Christopher et al. 1996 and 1997).
To date, HF ultrasound imaging is almost exclusively performed with single
element transducer based ultrasound backscatter microscopes (UBMs). Commercial
UBMs used in small animal research provide high-resolution real-time imaging with
Doppler features (Foster et al. 2002; Gronros et al. 2006; Gan et al. 2007). These
17
UBMs have been utilized in preclinical cancer and cardiovascular research
(Visualsonics 2006; Perrino et al. 2006). The latest version can achieve a real time
resolution better than 30 µm, and have the ability to detect microvascular flows in
vivo. With special scanheads, the latest system can perform high frame rate (>240
frame/s) B-mode imaging in a limited field of view (<1mm) (Visualsonics 2006). In
addition, a frame rate of greater than 1000 frame/s can be realized, using certain
reconstruction algorithms (Cherin et al. 2006; Liu et al. 2006). The mechanical
nature and fixed focal point of UBM, however, limit the further improvement of
image quality and frame rate (>200 frame/s) in a wider field of view in small animal
cardiac studies (Xu et al. 2005; Li et al. 2005). As with clinical ultrasound systems, a
better solution would be a system that utilizes HF ultrasonic linear arrays (>30 MHz)
which have been successfully developed (Ritter et al. 2002; Cannata et al. 2006). In
order to take full advantages of these arrays, a complete imaging system utilizing
these HF arrays is desired. In this chapter, the development of a HF high frame rate
ultrasound B-mode imaging system with arrays is described along with in vivo and in
vitro experimental results.
Methods
A. Beamformers in Ultrasound System
Similar to conventional ultrasound systems with low-frequency arrays, the
beamformer is the critical part of this system. The beamformer, which includes
transmit beamformer and receive beamformer, is used to scan and focus the acoustic
18
beam of an array electronically. As shown in Fig. 2.1, by applying appropriate time
delays to transmitted pluses or received echoes of each element, and summing the
delayed echoes up, the acoustic beam of an array can be focused electronically
(Shung 2005).
Figure 2.1 Focused acoustic beam by applying appropriate time delays.
As shown in Fig. 2.2, the time delay between the two elements can be calculated
as:
c
a b b x a
c
r
t
n
n
n
2 2 2 2
) ( + − + +
=
∆
= ∆ (2.1)
where c is the sound velocity in a medium, x
n
is the pitch size between two elements,
P(a,b) is the intended focal point, and r is the path length from an element to the
focal point.
In this dissertation, HF linear arrays were used. Their acoustic beam can be
scanned and focused electronically as shown in Fig. 2.3. A group of elements in an
array forms a sub-aperture. All received signals in the sub-aperture elements are
delayed and summed up to obtain one scan line of an image. The sub-aperture moves
electronically from left to right by one element for each time. Thirty-three scan lines
Transmit /Receive
Beamformer
Delay
Delay
Delay
Delay
Delay
Delay
1
2
3
4
5
6
Focal
Point
Array Element
19
constitute a whole image with a 48-element array. This electronic scanning can be
realized by multiplexing and demultiplexing circuits, which will be discussed in the
later sessions.
Figure 2.2 Delay time calculations based on the geometry structures of an array.
Figure 2.3 Electronically scanned and focused acoustic beams in a linear array. N
B
is the
element number of a sub-aperture, x
n
is the pitch, and F is the focal distance in the
azithumal direction.
r
x
n
r+ ∆r
n
Z
P(a,b)
a
b
20
As shown in Fig. 2.3, the lateral resolution of an array or a sub-aperture is
determined by the sub-aperture size and focal distance on the azithumal plane:
λ
n B
L
x N
F
R
×
= (2.2)
where R
L
is the lateral resolution, N
B
is the element number of a sub-aperture, x
n
is
the pitch, λ is the wavelength in a medium, and F is the focal distance in the
azithumal direction.
B. Digital vs. Analog Beamformers
In array-based ultrasound systems, the receive beamformer is more complex and
critical to the performance of the whole system (Thomenius 1996). It can be
implemented either analogly or digitally. As shown in Fig. 2.4, multiple echo signals
are digitized first in the digital beamformer, while these signals are beamformed first
in the analog beamformer. Therefore many more analog-to-digital converters (ADCs)
and digital signal processors (DSPs) are required to reconstruct images in digital
beamformers than in analog beamformers. In current low-frequency ultrasound
systems, the cost-performance ratio of digital beamformers is much better than that
of analog beamformers since the ADCs and DSPs with low clock frequencies are
inexpensive. Therefore, digital beamformers dominate in clinical ultrasound systems.
Nevertheless analog beamformers still have the advantages of low cost, accurate
delays, zero beamforming time, and large dynamic range (Brunner 2002; Stitt et al.
2002). For example, analog beamforming are still performed before continuous-wave
Doppler in clinical ultrasound systems due to its large dynamic range (Brunner 2002).
21
Considering the frequency range of the HF arrays, ADCs with high sampling rate,
which are still expensive until now, are required. The digital beamformer also
requires powerful and expensive DSPs or Field Programmable Gate Arrays (FPGAs)
in order to process large amounts of sampled data. These ADCs and DSPs make HF
beamformers more difficult and expensive to be implemented digitally. Therefore, in
this dissertation, we chose the approach of having the acoustic beam focused with an
analog receiving beamformer.
(a)
(b)
Figure 2.4 Block diagrams of an analog beamformer (a) and a digital Beamformer (b).
Array Element
DSP/Digital Adder
A/D
6
5
4
3
2
1
Focal
Point
A/D
A/D
A/D
A/D
A/D
DSP+Delay
DSP+Delay
DSP+Delay
DSP+Delay
DSP+Delay
DSP+Delay
DSP
6
5
4
3
2
1
Focal
Point
Array Element
Analog Adder
Analog Delay Line
Analog Delay Line
Analog Delay Line
Analog Delay Line
Analog Delay Line
Analog Delay Line
A/D
DSP
22
C. High-frequency Beamformer Design
Major difficulties in high-frequency beamformer design are the requirements of
wider bandwidth signal and more accurate delays than low frequency versions
(Thomenius 1996). Furthermore, noise tolerance is stricter for high-frequency
beamformers due to the higher attenuation in this frequency range (Shung 2005).
More precise control signals are required for triggering and dynamic focusing. Since
high-speed circuits are involved, printed circuit board (PCB) layout is more critical
to the performance of high-frequency beamformers.
By carefully considering the difficulties mentioned previously, a 16-channel
miniaturized analog beamformer was designed and implemented. The system is
capable of acquiring high-frequency high frame rate (>100 frame/s) ultrasound
images with 48-element arrays.
System Description
Fig. 2.5 shows the block diagram of the system. The system included five circuit
boards: a microprocessor timing control board, a 64-channel transceiver board, a 16-
channel three-stage amplification board with a cross-point switch, a receive
beamformer board and a power supply board. Their performances will be described
in the following sections.
23
TGC
AMP
Transducer
Array
HV
AMP
64 Channel
Transceiver
64 To16
Multiplexers
Cross-point
Switches
Micro
Processor
Rx Analog
Beamformer
PC A/D
Labview
16 To 64 De-
Multiplexers
Figure 2.5 Block diagram of the HF imaging system with arrays.
The heart of the system was the microprocessor board supported by high-speed
buffers (74AC573, ON Semiconductor, Phoenix, AZ, USA). Upon each line trigger
signal given by the PC, the control board sent out over 70 control signals to the
16×16 video cross-point switch, multiplexers, demultiplexers, and programmable
analog delay circuits. After demultiplexing, sixteen –66 Vpp spikes were applied to
the appropriate elements of a transducer. The 16 received echoes were selected from
the 64 channel transceivers by the multiplexers. Then these echoes were band-pass
filtered and amplified by the Time-Gain-Compensation (TGC) amplifiers, and
rearranged by the cross-point switch. Then these 16 rearranged echoes were
beamformed on the receive beamformer board. Next, the one-channel beamformed
echo signals are digitized by a 500 MHz 8-bit PCI A/D card (Gage Applied
Technologies, Lachine, Canada) in the PC. Finally the acquired data was processed
and B-mode images were displayed in real time by Labview-based software
(National Instruments, Austin, USA).
24
A. Array Characteristics
The characteristics of three linear arrays used in this study are outlined in Table
2.1. The transducers were fabricated at the NIH Medical Transducer Resource Center
at the University of Southern California (Ritter et al. 2002; Cannata et al. 2006). The
lateral resolutions of the arrays were calculated based on 16 element sub-aperture at
the elevation focal point with both transmit and receive beamforming. In order to
achieve an image of better quality and wide field of view, it will be shown later that
an array with 96 or 128 elements is preferred as well as a sub-aperture with 32
elements or more.
TABLE 2.1 CHARACTERISTICS OF THE LINEAR ARRAYS USED IN THIS STUDY
Transducer 1 Transducer 2 Transducer 3
Element No. 48 64 64
Bandwidth 59% 55% 56 %
Element dimensions 1.5 mm×82 µm 3 mm×36 µm 2 mm×85 µm
Kerf width 18 µm 14 µm 15 µm
Pitch 100 µm 50 µm 100 µm
Focal distance 6.2 mm 9.5 mm 8 mm
Lateral resolution 200 µm 534 µm 250 µm
Field of view 3.3 mm 2.4 mm 4.8 mm
Note: Focal distance is in the azithumal plane. Lateral resolutions were calculated based on 16
element sub-aperture at the focal point with both transmit and receive beamforming. Sound
velocity in water was 1500 m/s.
B. Microprocessor Timing Control Board
The main component of the timing control board was an Atmel ISP (In-System
Programmable) microprocessor (Atmel Corp., San Jose, CA, USA). Supported by
high-speed buffers (74AC573, ON Semiconductor, Phoenix, AZ, USA) the board
provided over 70 control signals to the 16×16 video cross-point switch, multiplexers,
demultiplexers and analog delay circuits with reliable timing control and flexibility.
25
For a high performance imaging system both dynamic focusing in transmit and
receive at all locations are necessary, but considering the complexity of circuit
design and high frame rate or real time capability, the first generation system was
implemented with only focusing in receive. A maximum of 32 ns transmit delays
with increment of 0.5 ns (PDU16F-.5, Data Delay Device Inc., Clifton, NJ) or
maximum 64 ns delays with increment of 1 ns (PDU16F-1, Data Delay Device Inc.,
Clifton, NJ) can be achieved. Longer delays are also available by choosing
appropriate digital delay lines.
C. 64 channel Transceiver Board
The transceiver board included 16 1-to-4 demultiplexers, 64 pulsers, 64 T/R
switches, 64 preamplifiers and 16 4-to-1 multiplexers. The transceiver board
consisted of 16 sub-groups, each of which had a demultiplexer, 4 pulsers, 4 pre-
amplifiers and a multiplexer, As shown in Fig. 2.6, each multiplexer switched among
four transducer elements, and each demultiplexer switched among four trigger
signals. The multiplexer and the demultiplexer in the same sub-group had the
identical control signals which were used to select the corresponding activated
element out of the 4 transducer elements in the sub-group.
The pulser circuits, triggered by demultiplexed control signals from the control
board, generated -66 Vpp negative spikes with 10 ns duration and 80 MHz –6 dB
bandwidth for chosen elements. A typical pulse measured using a 50-ohm load (solid
line) and its spectrum (dash line) are shown in Fig. 2.7. Before each preamplifier, a
T/R switch including back-to-back diodes limited the preamplifier input voltage to
26
be less than ± 0.7 V. The preamplifiers were employed to compensate for the
insertion losses introduced by the T/R switches and the following multiplexers. Each
preamplifier included an ultra low-noise operational amplifier MAX4106 (Maxim
Integrated Products Inc., Sunnyvale, CA, USA) with a gain of 10 dB and a –6 dB
bandwidth of 100 MHz.
Figure 2.6 Interconnections among the transducer, the multiplexers, the cross-point
switch and the analog receive beamformer.
… …
……
……
E1 E17 E33 E49
MUX_1
E2 E18 E34 E50
MUX_2
E16 E32 E48 E64
MUX_16
Analog Receive Beamformer
Cross-point switch
… Beamformer
Channel l6
Beamformer
Channel l
Beamformer
Channel 2
Transducer Element
E1 E2 E16 E17 E18 E32 E33 E34 E48 E49 E50 E64
P1 P2 P16 P17 P18 P32 P33 P34 P48 P49 P50 P64
Pulser for Each Element
DEMUX_1
T1 T17 T33 T49
DEMUX_2
T2 T18 T34 T50
DEMUX_16
T16 T32 T48 T64
……
27
Figure 2.7 The unipolar pulse (solid line) and its corresponding spectrum (dash line).
28
D. 16-channel Amplification Board
A three-stage 16-channel amplification board provided another 30 to 70 dB gain
to the echoes. Each channel included a fixed gain amplifier, a TGC amplifier and a
variable gain amplifier. Two AD603 (Analog Devices Inc., Norwood, MA, USA)
were involved in the TGC and the variable gain amplifier. A passive band-pass LC
filter with the –3 dB cut-off frequency from 10 MHz to 50 MHz were used directly
after the second stage AD603 to remove the noise outside the transducer bandwidth
and the noise from the amplifier. An AD8009 (Analog Devices Inc., Norwood, MA,
USA) was used to provide another gain of 10 dB to compensate the 6 dB insertion
loss through the passive LC filter.
As shown in Fig. 2.6, each multiplexer switched among four transducer elements.
Therefore, each scan-line presented a different order of 16 elements, and the 16-
channel RF (Radio Frequency) data cannot be connected directly to the analog
receive beamformer. For example, the first line had the order of E1, E2,…E16, and
the second line became E17,E2,…E16 before being fed to the beamformer. Because
the receive beamformer was designed with delays based on the geometrical location
of each element, the second line must be rearranged into the proper order of E2,
E3, …E16, E17. A high-speed 16×16 video cross-point switch AD8114 (Analog
Devices Inc., Norwood, MA, USA) was used to perform this task.
29
E. Receive Beamformer Board
Appropriate delays were applied to each of the 16 rearranged RF signals by a 4-
bit programmable analog delay circuit, which was a combination of four fixed analog
delay lines FDC1005~FDD15005 (ELMEC Technology of America Inc., San Mateo,
CA) and four 2-to-1 multiplexers AD8182 (Analog Devices, Norwood, MA) in
cascade as shown in Fig. 2.8(a). By setting the 2-to-1 multiplexer, each delay line
can have two possible states: the delay line is used, or the delay line is bypassed. The
programmable analog delay circuit can achieve a maximum delay time of 60 ns (four
15 ns delay lines used), and minimal increment of 1 ns (at least one 1 ns delay line
used). By adding more delay lines and multiplexers in each analog delay circuit, the
6-bit or 8-bit programmable analog delay circuit can be implemented. Then the
sixteen delayed signals were summed to form the ultrasound beam as shown in Fig.
2.8(b). The theoretical and measured delays of each channel at differential focal
points are shown in Fig. 2.9. Burst signals at 30 MHz generated by the waveform
synthesizer AFG2020 (Tektronix Inc., Beaverton, OR) were used in the above
measurement. In order to test the bandwidth of the receive beamformer, a continuous
sinusoid signal from AFG2020 was inputted to IN 1 in Fig. 2.8(b), and the amplitude
at OUT was read on an oscilloscope. The CH 1 delay circuit was set to the maximal
delay time 60 ns. The –6 dB bandwidth of the delay circuit was measured to be over
90 MHz, by sweeping the input continuous sinusoid signal from 10 MHz to 100
MHz. This bandwidth is sufficient for the current 30~35 MHz arrays, and future
arrays with frequency as high as 90 MHz. Due to the problem caused by glitches
30
generated by the multiplexers (Stitt et al. 2002), a fixed focal zone is used in in vitro
and in vivo experiments.
(a)
(b)
Figure 2.8 Block diagrams of the HF programmable analog delay circuit (a) and the
summation circuit (b) in the analog receive beamformer.
31
Figure 2.9 Calculated (lines) and measured (solid triangle) time delays of each channel
in the analog beamformer.
F. The Complete System
For reducing noise and on-site in vivo experiments, all circuit boards were
packed in a metal box. The photos of the completed system are shown in Fig. 2.10.
32
(a)
(b)
Figure 2.10 Photos of the unpacked system (a) and packed system (b).
33
Experiments and Results
A. Phantom Tests
A tissue mimicking phantom and a wire phantom were used to evaluate the
performance of the system as shown in Fig. 2.11. Tissue mimicking phantom in Fig.
2.11(a) had a 1 mm diameter bore (Hall et al. 1997). The wire phantom consisted of
4 tungsten wires with diameter of 20 µm (California Fine Wire Co., CA, USA) and
was arranged diagonally with an axial distance of 1.55 mm and a lateral distance of
0.65 mm as shown in Fig. 2.11(b). The wire phantom can be used to determine the
lateral resolution of the system (Cannata 2004; Huang and Shung 2005). The lateral
resolution of the system becomes poorer as the penetration is increased as described
in (2.2). In Fig. 2.11(d), the –6 dB axial and lateral resolutions of Transducer 3 were
measured to be 53 µm and 370 µm approximately at the focal point 8 mm.
34
(a) (b) (c) (d)
Figure 2.11 Tissue and wire phantom images. The tissue mimicking phantom image with
Transducer 2 showing 30 dB dynamic range (a). The wire arrangement in the wire
phantom used in the experiment (b). The wire phantom image with Transducer 3
showing 20 dB dynamic range (c). The 16 channel receive beamformer focused at 4.8
mm, 6.4 mm, and 9.6 mm. The wire phantom image with Transducer 3 showing 20 dB
dynamic range and the receive beamformer focused at 8 mm (d).
B. In vitro Experiments
In vitro rabbit eye images were acquired by placing the eye in 0.9% sodium
chloride solution at room temperature. The array transducer was electronically
scanned across the eye to obtain one frame of image. With Transducer 1 listed in
Table 2.1, the width of the field of view in azithumal direction is 3.3 mm. After one
frame image was obtained, the linear array was mechanically moved 3.3 mm to
acquire the next adjacent frame for covering a wider field of view. The composite
eye images are shown in Fig. 2.12. The anterior region, consisting of structures such
as cornea and lens, are resolved.
Real time images of excised rabbit eye were obtained with Transducer 2 listed in
Table 2.1. For each frame, the field of view in azithumal direction (48 elements used)
2
3
4
5
6
7
8
Axial Distance (mm) Lateral Distance (mm)
-1 01
2
3
4
5
6
7
8
-1 01
Axial Distance (mm)
•
•
•
•
2
3
4
5
6
7
8
-1 0 1
Axial Distance (mm)
-1 0 1
2
3
4
5
6
7
8
Axial Distance (mm)
Lateral Distance (mm)
Water
Tissue
Bore
Tissue
35
is 1.6 mm. Two frame images are listed in Fig. 2.13. The cornea, the lens and the iris
are distinguished on the images.
(a) (b)
Figure 2.12 Rabbit eye images obtained by Transducer 1. The cross section image at the
center of the eye (a) and the cross section image at the side of the eye (b).
(a) (b)
Figure 2.13 Rabbit eye images obtained with Transducer 2. Cornea (A), iris (B) and lens
(C) are distinguished.
A
C
B
-1 01
Axial Distance (mm)
Lateral Distance (mm)
5
6
7
8
9
10
11
-1 01
Axial Distance (mm)
Lateral Distance (mm)
5
6
7
8
9
10
11
B
A
C
Lateral Distance (mm)
-8 0 -4 48
Axial Distance (mm)
2
6
10
14
Lateral Distance (mm)
-8 0 -4 48
Axial Distance (mm)
2
6
10
14
36
C. In vivo Experiments
In vivo real-time imaging and high frame rate imaging were conducted on mice
after anesthetizing and shaving. Ultrasound gel was used as a coupling medium on
the skin. To avoid shadowing, the acoustic beam must be transmitted through the gap
between two ribs. High frame rate images (>100 frames/s) were acquired after
positioning a transducer at the area of interest.
(a) (b) (c)
Figure 2.14 Mouse heart images. Showing atrium (A) and ventricles (B) (C). (a) and (b)
were obtained with Transducer 1 at 30 frames/s. (c) was obtained with Transducer 3 at
100 frames/s.
Due to the field of view in azithumal direction (3.3 mm/frame), only a part of the
mouse heart was displayed. By mechanically scanning the whole heart with a
transducer, structures such as atriums and ventricles were visible. In Fig. 2.14, heart
images were extracted from the acquired movies. Analyzing and counting the heart
beating cycles in the acquired movie, the heart rates in anesthetized mice were
estimated as 200-250 beats/min, which was approximately 45% of those of
conscious mice (Wang et al. 2004;Yang et al. 1999). In the high frame rate
3
4
5
6
7
8
9
Axial Distance (mm) Lateral Distance (mm)
-1 01
B
3
4
5
6
7
8
9
Axial Distance (mm) Lateral Distance (mm)
-1 01
C
3
4
5
6
7
8
9
Axial Distance (mm) Lateral Distance (mm)
-1 01
A
37
acquisition mode, only half-second movie (50 frames) was saved and played back
due to the limitation of on-board memory on the Gage A/D Card.
Summary
In this chapter, we described the development of a high-frequency high frame
rate ultrasound imaging system with 30~35 MHz linear arrays. This system is
capable of acquiring B-mode images at a frame rate higher than 100 frame/s. The
acquired movements of cardiac structures in mouse heart demonstrated the potential
of this system in biomedical applications.
However, the image quality of this system is not satisfactory due to the
unoptimized analog front-end electronics and 8-bit digitization card. Further
improvements of this system will be discussed in the following chapters.
38
CHAPTER 3
IMAGE QUALITY IMPROVEMENTS
Introduction
As discussed above, SNR is the most critical factor in achieving high quality
images, as well as detecting weak signals from blood. SNR improvements can be
made by either decreasing the system noise or increasing the system signal.
First of all, strong receiving signals can be achieved by using efficient pulsers.
The pulse generator is a critical component in all ultrasound systems. The pulsers
described in Chapter 2 were unipolar pulsers. These unipolar pulsers are frequently
used in high-frequency ultrasound applications because they are commercially
available and easy to implement (Brown and Lockwood 2002). In conventional
ultrasound systems, however, bipolar pulsers are commonly used to achieve a higher
sensitivity and for Doppler applications (Brunner 2002). Several recent papers
(Brown and Lockwood 2002; Brown and Lockwood 2005) have discussed both the
importance and the design of unipolar and bipolar pulsers for high-frequency (> 50
MHz) ultrasound applications. Based on these reports and because of the poor
performance of our current unipolar puslers, a low-cost simple bipolar pulser
allowing operation for frequencies greater than 60 MHz for current HF linear arrays
and future HF phased arrays (Ritter et al. 2002; Cannata et al. 2006) was proposed
and designed. With the new puslers, we are capable of increasing the amplitudes of
received echoes. However, this is still not enough due to the extremely large
dynamic range of these echoes. As pointed out by Brunner (2002), the received
39
amplitude from tissues may be as low as 10’s of µV
p-p
and as high as 1 V
p-p
. Hence,
the front-end electronics needs 100 dB dynamic range or even more. Therefore, in
the second part of this chapter, we discussed the strategy of reducing noise by using
new low-noise amplifiers, appropriate band-pass filters, grounding isolators and
PCBs with appropriate layouts. Finally, the increased dynamic range of the system
enables us to take full advantages of utilizing a Gage 14-bit A/D card. These
strategies of improving SNR will be discussed in detail in this chapter.
Methods
A. Efficient Bipolar Pulse Generator
This bipolar pulser is based on a high-speed level-shifter and a high-voltage
MOSFET pair with supporting TTL (Transistor-Transistor Logic) timing circuit. Fig.
3.1 and 3.2 are the schematics of the timing control circuit and the bipolar pulse
generator respectively. The timing control circuit consisted of a programmable clock
generator ECS-P83/P85 (ECS, Inc. International, Kansas City, KS), a high-speed 8-
bit TTL counter 74F269 (Fairchild Semiconductor, South Portland, ME), logic gates
74AC04 and 74AC00 (Fairchild Semiconductor, South Portland, ME), and two 12-
bit counters 74VHC4040 (TOSHIBA America, New York, NY). The timing circuit
provided matched trigger signals for the bipolar pulser as shown in Fig. 3.3. The
output of the clock generator was divided by two 12-bit counters 74VHC4040,
making it possible to obtain a variable pulse repletion frequency (PRF) from 100 Hz
to 100 KHz. The 8-bit counter 74F269 was used to adjust the pulse cycle number
40
from 1 to 255. The matched outputs C and D shown in Fig. 3.3(b) were connected to
the bipolar pulse generator. The pulse generator included two delay lines DS1100
(Dallas Semiconductor, Dallas, TX), two level-shifters EL7158 (Intersil Americas,
Milpitas, CA), and a high-voltage MOSFET pair TC6320 (Supertex Inc., Sunnyvale,
CA). The matched input signals C and D were delayed by DS1100. The selected
delay time of DS1100 was determined by the center frequency of the bipolar pulse
and the chip delay of 74AC04. The level-shifters EL7158 shifted the TTL delayed
signals to 12 V inverted outputs (Intersil 2003). Then the shifted outputs were fed to
the high-speed high-voltage MOSFET pair TC6320, which offers a 400 Vpp
breakdown voltage and a 2 A output peak current (Supertex 2003). After the diodes
PMBD7000 (Philips Semiconductor, Sunnyvale, CA), a RF transformer TMO-1-1+
(Mini Circuits, Brooklyn, NY) was inserted for isolation. Following the transformer,
two crossed diodes were used as an expander (Lockwood et al. 1991).
41
Figure 3.1 Schematic of the timing control circuit for the bipolar pulser.
Figure 3.2 Schematic of the bipolar pulser.
Figure 3.3 The timing signals and output of the bipolar pulser. (a) illustrates the
waveforms at the points A and B. (b) shows the matched outputs of the timing control
circuit and the inputs of the pulser. (c) displays the delayed outputs at the points E and F.
(d) is the output of the pulser. (The cycle number was set to 2 at the counter 74F269)
PRF
A
B
C
D
E
F
(a) (b)
(c) (d)
Pulse
42
The pulser generator shown in Fig. 3.2 was simulated by Pspice (Cadence Design
Systems, Inc., San Jose, CA). The simulation result is displayed in Fig. 3.4. The
spice models of the components (EL7158 and TC6320) were provided by the
vendors. As noted in the data sheet of EL7158, its clock speed is up to 40 MHz at a
capacitance load of 2000 pF. Its actual load in this design, the input capacitance of
TC6320, is less than 200 pF (Intersil 2003) and it is much lower than 2000 pF.
Therefore, the center frequency of the bipolar pulser design may exceed 100 MHz
based on simulation.
Figure 3.4 Simulated monocycle bipolar pulse. Its duration is approximately 10 ns.
43
Low ringing pulse generation was benefited from clean power supplies. Linear
power supplies HB48-0.5-A+ (Condor D.C. Power Supplies, Inc., Oxnard, CA) were
used in the bipolar pulse generator which was a part of a HF Doppler system
(Gottlieb et al. 2005). High-voltage commercial power supplies Agilent E3647A
(Agilent Technologies, Palo Alto, CA ) and Extech Instrument 382285 (Extech
Instruments Corp., Waltham, MA) were used together to obtain the high-voltage
bipolar pulses. Bypass capacitors 10 uF and 0.1uF provided further power supply
filtering, as did ferrite beads. All beads and capacitors were placed as close as
possible to the power pins of chips. The prototype bipolar pulser is shown in Fig. 3.5.
Figure 3.5 Photo of the prototype bipolar pulser.
44
The output of the bipolar pulser driving a 50 Ω load is shown in Fig. 3.6. The
PRF was set to approximately 1 KHz for the monocycle pulse generation and was set
to approximately 44 KHz for the N-cycle bipolar pulse generation. The chosen PRFs
can be determined by the requirements of either B-mode imaging or Doppler
imaging. The power supplies were set to ±120 V for high-voltage pulse generation.
Otherwise, the power supplies were switched to ±48 V setup for low-voltage pulse
generation, in order to address the transducer heating issues. The waveforms shown
in Fig. 3.6 were recorded by a digital oscilloscope LeCroy 9350AL (LeCroy Corp.,
Chestnut Ridge, NY). A 30 dB attenuator HAT-30 (Mini Circuits, Brooklyn, NY)
was used to protect the oscilloscope.
45
(a) (b)
(c) (d)
(e) (f)
Figure 3.6 Experimental results of the bipolar pulser with a 50 Ω load. (a) is a 35 MHz
monocycle bipolar pulse with amplitude over 160 V and its spectrum. (b) is the
expanded voltage scale of (a), showing the ringing of the pulse in (a) was less than 0.3
Vpp. (c) illustrates a 7-cycle 45 MHz bipolar pulse for Doppler applications and its
spectrum. (d) displays a 7-cycle 65 MHz bipolar for Doppler applications and its
spectrum. (e) is a 3-cycle pulse with amplitude over 160 Vpp and frequency over 60
MHz. (f) shows a relationship curve between the normalized output amplitude and the
frequency of the pulser.
46
Fig. 3.6 (a) shows a monocycle 35 MHz pulse with amplitude of approximately
160 Vpp and its spectrum. The pulse generator should satisfy the requirement of
30~35 MHz linear arrays discussed in Chapter 2. The voltage scale of Fig. 3.6(a) was
expanded and shown in Fig. 3.6 (b). The measured ringing was less than 0.3 Vpp.
Due to the non-identical electrical characteristics between P-channel and N-channel
MOSFETs in TC6320, the fall and rise times of the pulses shown in Fig. 3.4 and Fig.
3.6 (a) are slightly unsymmetric. In Fig. 3.6 (b), this unsymmetry is more evident.
An undischarged voltage about 2 V which is not visible in Fig. 3.6 (a) exists near the
tail of the pulse falling edge. This undischarged voltage is due to the existing
capacitances in TC6320. After expanding the voltage scale of Fig. 3.6 (a), it shows
that an extra 20 ns is needed for TC6320 to be completely discharged. No side
effects were found in experiments due to this extra discharging time. An active
damping circuit described in the datasheet published by Supertex in 2006 can be
used to reduce this discharging time. Fig. 3.6 (c) displays a 7-cycle 45 MHz pulse
and its spectrum. Fig. 3.6 (d) illustrates a 7-cycle 65 MHz pulse and its spectrum.
The 45 MHz and 65 MHz multicycle pulses were used in the HF pulsed-wave
Doppler system (Gottlieb et al. 2005). Fig. 3.6(e) demonstrates that the pulser with
this design can produce N-cycle pulses with amplitude over 160 Vpp and frequency
over 60 MHz [N=3 in Fig. 3.6 (e)]. The 2nd harmonics of the produced pulses in Fig.
3.6 were less than –19 dB. The achieved pulse frequency of this design (60 MHz) is
higher than the one of a new integrated pulser HV732 (40 MHz) that was recently
released by Supertex Inc. in April 2006 (Supertex 2006). The breakdown voltage of
47
this design is up to 400 Vpp as well, compared to the 220 Vpp of HV732. Fig. 3.6(f)
displays a relationship curve between the normalized output amplitudes and the
center frequencies of the bipolar pulser with the ±48 V power supplies and a
programmable clock generator Agilent 33250A (Agilent Technologies, Palo Alto,
CA). Fig. 3.6(f) shows that an increase of the pulse frequency does not significantly
decrease the pulser output amplitude in the operating frequency range of the pulser.
The trigger-out-delay of the pulser was approximately 58 ns. The measured jitter of
the pulser was less than 300 ps. Although a pulse of higher frequency could be
achieved, it had a distorted waveform and a lower amplitude. It appears that the
highest frequency that could be achieved was limited by the switching time of
EL7158 and TC6320, the highest operating frequency of TTL timing circuits, and
the load capacitance. The 2 A peak output current of TC6320 ensures the pulser’s
driving capability on a capacitive load that is due to transducer piezoelectric
materials and connection cables. In the above experiments, a 1 m RG-174 coax cable
(Belden CDT Inc., St. Louis, MO) was used.
Finally, the performance of the bipolar pulser was compared to the performance
of the Panametrics 5900 Pulser/Receiver (Olympus NDT Inc., Waltham, MA). The
bipolar pulser generated a 26 ns 65 Vpp monocycle pulse, and the 5900PR generated
a 63 Vpp 12 ns negative spike. The Pb(Mg
1/3
Nb
2/3
)O
3
-PbTiO
3
(PMN-PT) transducer
(Gottlieb et al. 2005) was excited by the bipolar pulser and the Panametrics 5900PR
respectively. Unamplified received echoes are shown in Fig. 3.7. The solid line
represents the echoes from the transducer excited by the Panametrics 5900PR, and
48
the dotted line represents the echoes from the transducer excited by the bipolar pulser.
It is clear that the stronger echoes were obtained when the transducer was excited by
the bipolar pulser. The received signal strength was increased by 5 dB, after
compensating for the amplitude difference between the negative spike and the
monocycle pulse. We assumed a linear relationship between the amplitude of echoes
and the amplitude of the excitation waveforms.
Figure 3.6 Comparison between the bipolar pulser and the Panametrics 5900PR.
Unamplified echoes from the PMN-PT transducer are shown in this figure. The solid
line is the echoes from the transducer excited by the 5900PR. The dotted line represents
the echoes from the same transducer excited by the bipolar pulser. The received echo
amplitude was increased by 5 dB, using the bipolar pulser.
More than 16 identical bipolar pulsers with this design have been duplicated and
used in the system described in Chapter 5. The output amplitude variation across the
49
16 bipolar pulsers was less than 0.3 dB at 30 MHz. As shown in later chapters, this
bipolar pulser design achieved satisfactory experiment performances and it is
suitable for applications in high-frequency ultrasound image systems.
B. Low-noise Front-end Electronics
Powerful pulses can be applied to transducer with the new pulser design,
however, the detection of week echoes is still depending on the performance of the
receiving circuits. Typically measuring the minimal detectable signal is a method to
evaluate the performance of an analog signal chain as Figure 3.8 shows:
Figure 3.8 Block diagram of measuring the minimal detectable signal.
A burst signal (10 mVp-p typically) from the function generator is attenuated by
40-60 dB. Then, the attenuated signal is amplified by front-end electronics, which
include a pre-amplifier, a TGC amplifier and a variable amplifier in one channel of
our system. Both the original signal and amplified signal are displayed on an
oscilloscope (or spectrum analyzer). Increasing the attenuation of the variable
attenuator to find the highest attenuation point, where the amplified signal still can
be recognized as a burst signal on the oscilloscope. Based on the amplitude of the
input signal and the attenuator readings, the minimal detectable signal of the front-
end electronics can be measured. The minimal detectable signal identifies the
Function
Generator
Variable
Attenuator
Front-End
Electronics
Oscilloscope/
Spectrum Analyzer
50
sensitivity of a system. The lower value demonstrates the better sensitivity of the
system.
The output noise of the front-end electronic described in Chapter 2 was in range
of 100 mV
p-p
at 58 dB gain, and the minimal detectable signal was approximately
130 µv, which was not enough for acquiring satisfactory images or detecting weak
scattering signals from blood. A low-noise front-end electronics is a must for the 2
nd
-
generation system with Doppler features.
Recently, an ultra low-noise differential pre-amplifier with programmable R
in
AD8332/31 has been designed particularly for ultrasound applications (Analog
Device 2003(a)). The effects from on-board noise (i.e. clock noise etc.) can be
minimized with its differential structures. Furthermore, it has already embedded with
wideband variable gain amplifiers for TGC. Each chip can handle two-channel
amplifications with a maximum gain of 55.5 dB. All of these features enable us to
build a compact ultra low-noise system. Since it has differential outputs, either a
differential receiver amplifier AD8129 (Analog Devices, Norwood, MA) or a
passive transformer should be followed if its post-circuits are with single-end input
as Fig. 3.9 shows.
Evaluation boards with AD8332 and AD8219 from the vender were employed to
construct a test circuit which followed the setup shown in Fig. 3.9(b). Fig. 3.10(a)
illustrates that the gain control interface of AD8332/31 provides the linear-in-dB
scaling, which is ideal for compensating for the attenuation in tissues. At the
maximum achievable gain of 50 dB, the output background noise was about 10
51
mVp-p in Fig. 3.10(b) and the minimal detectable signal was approximately
measured as 50 µV
p-p
. Finally the performance of AD8332 was compared to the
commercial amplifier Miteq AU-1313 which is being used in our current UBM
system. In vitro images illustrated in Fig. 3.11 were obtained by using the AD8332
and AD8129 based amplifier and Miteq AU-1313 respectively. These images
demonstrate that these two amplifiers have comparable performance.
(a)
(b)
Figure 3.9 Block diagrams of AD8332 applications.
52
(a) (b)
Figure 3.10 Measured characteristics of AD8332. The relationship between the control
voltage and gain of AD8332 (a) and the output noise of AD8332 at 50dB gain (b).
(a) (b)
Figure 3.11 Comparison between AD8332 and Miteq AU-1313. In vitro rabbit eye
images obtained by AD8332 with 42 dB gain (a) and AU-1313 with 45 dB gain (b).
Further noise reduction can be achieved by limiting the signal bandwidth,
because the major amplifier noise is proportional to the bandwidth of signals as
expressed below (Horowitz and Hill 1989)
:
kTRB rms V
noise
4 ) ( = (3.1)
where k is Boltzmann’s constant, T is the absolute temperature in degrees Kelvin, B
is the bandwidth in hertz and R is the input resistance of an amplifier. Since the
53
bandwidth of a transducer is always limited to a certain range, there is no reason to
use an amplifier with a bandwidth wider than the transducer bandwidth. Appropriate
band-pass filters should be selected for different applications. Due to narrow-band
processing in Doppler applications, filters with fairly narrow bandwidth were
selected, such as FN-2181 (Filtronetics Inc., Kansas City, Missouri) with –3 dB
bandwidth of 5% and BBP-30 (Mini-Circuits, Brooklyn, NY) with –3 dB bandwidth
of 30%. On the other hand, in order to achieve good axial resolutions in B-mode
imaging applications, PBP-35W (Mini-Circuits, Brooklyn, NY) with –3 dB
bandwidth of 60% was selected. Experiments show that the utilization of band-pass
filters can improve SNR further.
C. Grounding Isolation and PCB Layout Considerations
Due to the existence of both high-voltage pulsers and low-voltage amplifiers in
ultrasound systems, isolating them is an important issue for achieving better SNR.
For pulsers, the high-voltage switch power supply is regulated from a low-
voltage power supply. The switching noise contaminates the low-amplitude received
signals. Completely isolating different grounds is a good solution (Kester and Bryant
1996). Thus, different isolators have to be involved depending on the frequencies
and the types of these signals. A typical isolator includes RF transformers for analog
signals and optical isolators for low-speed (<10 MHz) digital signals (Agilent
Technologies 2004). In our system, however, some digital signals are high-speed
signals (>10 MHz). Both the mentioned transformers and optical isolators are not
applicable.
54
A new digital isolator ADuM1100 was introduced by Analog Devices in 2003(b).
This isolator combines high-speed CMOS and monolithic air-core transformer
technology and it supports data rates up to 100 Mbps, which is enough for our
system with 30~35 MHz arrays. Furthermore, separating digital and analog power
supplies may improve noise reduction further (Kester and Bryant 1996). The second
generation system described in the following chapters has separate digital, analog,
and pulse ground. ADuM1100 were used between the digital ground and the analog
ground. RF transformers separated the pulse ground and the analog ground.
The PCB layout is also critical to the performance due to the high frequency
range of the whole system. Six-layer printed circuit boards can provide better ground
coupling and power distribution. All power pins of chips were decoupled locally
with 0.1uF and 10uF capacitors. Each power pin of chips was connected to a ferrite
bead, which filters out high-frequency noise from power supplies. All lengths of
corresponding traces for different channels were equalized most since the
beamformed signals are affected by delays introduced by traces. Surface mount
components were used (where possible) to decrease phase errors. Noisy parts, such
as microcontrollers, switching power supplies and logic ICs, were physically kept as
far as possible from amplifiers and delay units.
D. High-resolution A/D
By going over all strategies for improving SNR, it is meaningful to use the 200
MHz 14-bit PCI acquisition card (CS14200), which can provide a 66 dB dynamic
range (Gage Applied Technologies 2004).
55
Figure 3.12 illustrates the advantage of using an acquisition card with more bits.
The original data was sampled by the UBM with CS14200 (Sun et al. 2006b), and
then the original data was resampled with 8-bit resolution by Matlab to simulate an
8-bit A/D card. In Figure 3.12, (a) and (b) are the zebrafish images reconstructed
from the original data and the resample data respectively. The dynamic range of (a)
and (b) are 66 dB and 44 dB respectively. Significant image quality improvements,
such as less saturation and more image details, are noticed with comparing these two
images. However, the utilization of 14-bit A/D card decreases the frame rate due to
transferring and processing twice amount of data.
56
Figure 3.12 Zebra fish images with different dynamic ranges. (a) shows 66dB dynamic;
(b) shows 44dB dynamic range.
Summary
In this chapter, we proposed and evaluated several strategies of improving the
performance of the system, including the efficient bipolar pulser, low-noise front-end
electronics, better PCB layouts, and 14-bit digitization card. Upon utilizing these
strategies, significant improvements are expected in the new systems which will be
described in the following chapters.
(a)
(b)
57
CHAPTER 4
HIGH-FREQUENCY PLUSED-WAVE DOPPLER SYSTEM
WITH PMN-PT NEEDLE TRANSDUCERS FOR
OPHTHALMOLOGY APPLICATIONS
Introduction
In order to implement the ultrasound Doppler system with high-frequency arrays,
we first developed a high-frequency pulsed-wave Doppler system which can be
connected to either a single element transducer or the analog beamformer with arrays.
This Doppler system was initially used to evaluate the treatment of retinal vein
occlusion.
Retinal vein occlusion whose incidence is over 0.5% in the population over 60-
year old (Klein et al. 2000) is a common eye vascular disease. It is the second most
common disease next to diabetic retinopathy among all retinal vascular diseases,
causing visual loss in elder population. (Williamson et al. 2003). Retinal vein
occulsion may occur in the central vein (central retinal vein occlusion: CRVO) or in
one of its main branches (branch retinal vein occlusion: BRVO) (Green et al. 1981)
as shown in Fig. 4.1. Several surgical procedures have been devised to remove the
obstruction and reestablish the blood flow (Arciniegas et al. 1984; Sergott et al. 1991;
Opremcak et al. 2006;Weiss and Bynoe 2001). In these surgical procedures, a
vitrectomy is often performed. During the vitrectomy procedure, several tiny
incisions are first made on the sclera. Then microsurgical instruments, for example a
19-gauge metal cannula, are inserted through the incisions as shown in Fig. 4.2
(Tameesh et al. 2004). Then 27~30-gauge needles can be used to inject tissue
58
plasminogen activator (t-PA) (Weiss and Bynoe 2001; Tameesh et al. 2004) or
triamcinolone acetonide (Ip
et al. 2004) into branch veins to reestablish blood flow.
Therefore, quantitative analysis is needed to evaluate blood flow during this surgery.
At present, fluorescein angiography in which retina photos are taken after
injecting dye into retinal veins is commonly used for both and evaluating treatments
of retinal vein occlusion (Tameesh et al. 2004). Due to the medial opacities, however,
this method is inadequate for both diagnosing and evaluating central retinal vein
occlusion. Furthermore, in the fluorescein angiography procedure, multiple photos
are taken by a camera with special filters, and they are analyzed subjectively by
ophthalmologists (Williamson, 1997). As a result, this method may not be suitable
for instantaneously evaluating blood flow reestablishment during surgery.
On the other hand, ultrasound Doppler is a widely used technology for flow
velocity measurements. Some researchers have demonstrated the feasibility to
measure the blood flow velocities from the central retinal vein and artery under the
optical disk, using commercial color Doppler systems (Baxter and Williamson 1993).
These studies have shown that the significant flow velocity reduction in the central
retinal veins correspond to CRVO. These velocity variances may be used to diagnose
CRVO. Nevertheless, this method may not be suitable during an ophthalmologic
surgery either. Because the achievable velocity resolution 1~2 cm/s is insufficient for
both the accurate clinical diagnosis and treatment evaluation, and the bulky
ultrasound probe size introduces some setup complexities for the instantaneous flow
evaluation during surgeries. Consequently, a simple and reliable approach with
59
minimized patient injuries is desired for real-time evaluating blood flow in retinal
vessels during retinal occlusion removal surgeries (Williamson 1997; Tameesh et al.
2004).
Figure 4.1 Eye anatomy.
Figure 4.2 Overview of microsurgerical instruments inserted through sclera. Labels 1
and 3 indicate microforceps, label 2 the microcatheter, and label 4 the infusion cannula.
CRVO
Location
BRVO
Location
Optical
Disk
Lens
Cornea
Sclera
12 3
4
60
A better velocity resolution of an ultrasound Doppler system can be achieved by
increasing the operational ultrasound frequency. Pioneering high-frequency
ultrasound Doppler studies with 20 MHz transducers dates back to 1974 (Hartley and
Cole 1974). Recent studies have shown that HF ultrasound Doppler can be used to
detect blood flow in the microcirculation (Foster et al. 2002). Blood flow less than
0.6 mm/s was detected in the eye (Silverman
et al. 1999). The measured blood flow
velocities can provide valuable information for diagnosing glaucoma, hypotony,
tumors, and other ophthalmology diseases. Hence, HF ultrasound Doppler can
provide sufficient velocity resolutions for evaluating blood flow in retinal vessels.
In HF ultrasound applications, the large tissue attenuation at high frequency
limits the penetration depth of acoustic wave. For example, the penetration depth is
less than 1 cm at 50 MHz (Lockwood et al. 1996). In order to measure the flow
under the optical disk which locates about 2 cm away from the cornea, a HF
intraocular transducer should be inserted through the sclera in the same way as other
microsugerical instruments mentioned above. Evidently intraocular transducers as
small as possible are preferred, considering minimizing patient injuries. Another
advantage of using small transducers is that the transducer can be located at the
measuring point precisely by its user. However most transducers used in previous HF
Doppler studies were Pb(Zr
1-x
Ti
x
)O
3
[PZT] or Poly(vinylidene fluoride) [PVDF]
single element transducers whose outer diameter were greater than 2 mm (Foster et
al. 2002; Silverman
et al. 1999; Hartley and Cole 1974; Reddy et al. 2005a, 2005b).
These transducers may not be appropriate for intraocular ultrasound applications due
61
to their sizes. Therefore fabricating small diameter transducers (less than 1 mm) is
necessary.
In addition to the size of the intraocular transducer, the sensitivity of the
transducer also affects the flow evaluation, since the backscattered signals from
blood are much weaker than those from tissues. Moreover, the diameters of retinal
veins are less than 200 µm (Feke et al. 1989). The transducer sensitivity is primarily
determined by electromechanical coupling coefficient k
t
(Shung 2005). Thus a
material with high k
t
should be chosen. Recent studies have shown that single crystal
materials such as PZN-PT and PMN-PT, which have been commercially available
for medical ultrasonic transducer applications, can significantly improve the
bandwidth and sensitivity of transducers (Ritter et al. 2002; Chen et al. 2005).
Therefore, we decided to utilize PMN-PT single crystal for fabricating intraocular
needle transducers.
In this chapter, we introduced a directional high-frequency pulsed-wave Doppler
system with PMN-PT needle transducers for in vivo measuring blood flows in the
central retinal vein and branch retinal veins during occlusion removal surgeries.
Transducer fabrication, system implementation and evaluation, in vivo animal studies,
and experiment analysis will be discussed.
62
System Description
A. Transducer Fabrication
As shown in Fig. 4.3, the central vein is under the optical disk and is
perpendicular to the retina. The branch veins travel parallel to the retina. In order to
evaluate both central vein occlusion and branch vein occlusion with needle
transducers, the angles between the ultrasound propagation axis and the needle axis
should be 0° and 45° respectively as illustrated in Fig. 4.4. Therefore, the needle
transducers with either a straight 0° tip or a 45° tip were developed respectively. The
fabrication procedures are described as follows.
A 700 µm thick (001) poled PMN-PT (HC Materials Corp., Urbana, IL) single
crystal was used as the active material of the transducer. A brief list of the material
properties for PMN-PT appears in Table 4.1. The data showed that PMN-PT single
crystal displays good electromechanical coupling coefficient, high piezoelectric
constant (d
33
) and lower dielectric loss. Firstly, the sample was lapped to 50 µm. A
matching layer made of Insulcast 501 and Insulcure 9 (American Safety
Technologies, Roseland, NJ) and 2-3 µm silver particles (Sigma-Aldrich Inc., St.
Louis, MO) was cured over the PMN-PT and lapped to 10 µm. A conductive backing
material, E-solder 3022 (VonRoll Isola, New Haven, CT), was cured over the
opposite side of the PMN-PT and lapped to 3 mm.
63
(a) (b)
Figure 4.3 Vein locations in CRVO (a) and BRVO (b).
(a) (b)
Figure 4.4 Diagrams of different probes for CRVO (a) and BRVO (b) measurements.
TABLE 4.1 CHARACTERISTICS OF PMN-PT SINGLE CRYSTAL.
Characteristic Value
Electromechanical coupling coefficient (k
t
) 0.58
Piezoelectric constant d
33
(pC/N) 1430
Relative clamped dielectric constant ( ε
S
/ ε
0
) 797
Dielectric loss 0.0036
Density (g/cm
3
) 8.0
Longitudinal wave velocity (m/s) 4608
Acoustic impedance (MRayl) 36.8
Curie temperature (
o
C) 131
Optical
Disk
Central Vein under
Optical Disk
Branch Veins
on the Retina
Optical
Disk
64
Since the needle transducer with either a 0° or a 45° tip, different dicing
strategies were used next. Based on simulation, the aperture size of active element
plugs should be 0.4 mm by 0.4 mm. For both 0° and 45° tips, the lapped PMN-PT
disk with backing was first diced to long and thick rods with width of 0.4 mm. Then
straight dicing was used for preparing the active elements with 0° tips. For those with
45° tips, the wafers were aligned with width direction facing up, and then diced into
0.4 × 0.4 mm posts, along 45
o
angle to the thickness direction.
Then, the diced active element was housed using Epotek 301 (Epoxy
Technology Inc., Billerica, MA) within a polyimide tube with inner diameter of 0.57
mm (MedSource Technologies, Trenton, GA). An electrical connector was fixed to
the conductive backing using the conductive epoxy. The polyimide tube provided
electrical isolation from the 20-gauge needle housing with inner diameter of 0.66
mm. An electrode was sputtered across the silver matching layer and the needle
housing to form the ground plane connection. A vapor-deposited parylene layer with
a thickness of 14 µm was used to coat the aperture and the needle housing. The
electrical connector was fixed to the conductive epoxy backing. The device was
electrically shielded by a stainless steel needle housing connecting a sputtered
chrome/gold (50/100 nm) layer to silver particle matching layer. The finished
transducers are shown in Fig. 4.5. Their outer diameters are 0.9 mm.
After the fabrication procedures, the characteristics of the needle transducer
were measured. Fig. 4.6 shows the electrical impedance magnitude (solid line) and
phase (dash line) of one needle transducer after housing. At the resonant frequency,
65
the electrical impedance was 47 Ω which was very close to 50 Ω required by the
electrical impedance matching of system. The center frequency of the transducer was
around 44 MHz. The bandwidth at -6 dB was about 45 %. The two-way insertion
loss of the transducer was -15 dB. Using PR5900 Pulser/Receiver with 1 uJ energy
setup, the maximum output voltage (Vpp) of the unamplified echo from a quartz
target was 1.5 V. These results show that the PMN-PT needle transducer has a high
sensitivity due to excellent piezoelectric properties of this material. Finally a wire
phantom with 20 µm tungsten wires was used to measure the lateral resolution of this
transducer. The wire phantom was placed in a water tank filled with deionized water.
The needle transducer was mounted on a 3-axis positioning system (Burleigh Inc.,
Fishers, NY). The needle transducer was moved laterally by a step of 20 µm. The RF
data was digitized by the Gaga A/D card with sampling frequency of 200 MHz. The
measured lateral profile is illustrated in Fig. 4.7. The –6 dB lateral resolution is about
300 µm at 2 mm depth which is a typical distance between the transducer and
measured retinal vessels in experiments.
(a) (b)
Figure 4.5 Photos of the needle transducer (a) and detailed 0° and 45° tips (b).
66
(a)
(b)
Figure 4.6 Measured needle transducer properties. (a) shows the impedance (solid line)
and phase (dash line);(b) shows the pulsed-echo (solid line) and spectrum (dash line).
67
Figure 4.7 Measured lateral profile of the needle transducer at 2 mm depth.
68
B. Micro-flow Phantom
In order to mimic small vessels in microcirculation, a micro-flow phantom was
fabricated. The micro-flow phantom consisted of polyimide tubes with inner
diameters from 127 µm to 574 µm (5707K11~5707K16, McMaster-Carr Supply
Company, Atlanta, GA). The blood-mimicking fluid was made by diluting 1-5 µm
spheriglass solid spheres 10004E (Potters Industries Inc., Carlstadt, NJ) in deionized
water. The blood-mimicking fluid container was connected to the polyimide tubes
through infusion lines with 27~30-gauge needles. The flow velocity in the micro-
flow phantom can be adjusted by height of the reservoir. A photo of the micro-flow
phantom is illustrated in Fig. 4.8.
Figure 4.8 Photo of the micro-flow phantom. Consisting of polyimide tubes with inner
diameters from 127 µm to 574 µm.
Micro-tubes with inner
diameters 127 µm to 574 µm
27~30 gauge
needles
69
C. System Design
Due to the low backscatter coefficient of blood, the SNR of the Doppler system
plays a significant role in blood flow measurements. Therefore, a custom high-
frequency pulsed-wave Doppler system with high SNR was built. Both the high-
efficiency bipolar pulser described in Chapter 3 and a low-noise commercial
amplifier AU-1114 (MITEQ Inc, Hauppauge, NY) were used. The system block
diagram (a) and photo (b) are shown in Fig. 4.9.
The whole system was controlled by a clock which was generated by a
programmable clock generator ECS-P83/P85 (ECS, Inc. International, Kansas City,
KS). The clock was duplicated to several identical clocks by a clock driver CD328
(Texas Instruments, Dallas, TX). The timing circuits synchronized by the clocks
from CD328 consisted of a high-speed 8-bit TTL counter 74F269 (Fairchild
Semiconductor, South Portland, ME), logic gates 74AC04, 74AC00 and 74AC573
(Fairchild Semiconductor, South Portland, ME), and two 12-bit counters
74VHC4040 (TOSHIBA America, New York, NY). One 45 MHz clock from the
clock driver CD328 was divided by two 12-bit counters 74VHC4040, making it
possible to obtain a PRF clock (171 Hz~175 KHz), a PRF filter clock (10.9
KHz~11.3 MHz), and a wall filter clock (171 Hz~21.9 KHz). A sample volume gate
(0.1-2 µs) at a specific sample depth (0-50 µs) was produced by a PRF trigged
monostable multivibrator 74HCT221 (Texas Instruments, Dallas, TX), and its
waveform is shown in Fig. 4.10 (B). As shown in Fig. 4.10(D, E), the timing circuits
70
(a)
(b)
Figure 4.9 Block diagram (a) and photo (b) of the HF directional pulsed-wave Doppler
system.
71
provided a N-cycle (N=1~255) trigger signals for the bipolar pulsers. Then the
bipolar pulses with 20~70 Vpp in Fig. 4.10(F) were produced and applied to the
needle transducer. The received echoes from the needle transducer were limited and
then amplified by the Miteq-1114. The amplified signals were first band-pass filtered
by a 45 MHz custom band-pass filter FN-2181 (Filtronetics Inc., Kansas City,
Missouri) and then fed to the in-phase and quadrature demodulator MIQC-60WD
(Mini-Circuits, Brooklyn, NY). The reference clock of the IQ demodulator was a
low-pass filtered output of the clock driver CD328. The demodulated intermediate
frequency (IF) signals I
IF
and Q
IF
were low-pass filtered to remove harmonics and
noise by BLP-10.7 (Mini-Circuits, Brooklyn, NY). Then they were sampled and held
by AD783 (Analog Devices, Norwood, MA). Followed by the 8th-order PRF filters
which consisted of MAX291 switch-capacitor filters (Maxim Integrated Products,
Sunnyvale, CA), the sampled-and-held signals were cleaned by removing sample-
and-hold harmonics. Also the optional 4th-order wall filters MAX7490 (Maxim
Integrated Products, Sunnyvale, CA) were followed to remove low-frequency clutter
signals. The cut-off frequencies of both the PRF filters and wall filters were adjusted
by their clocks from the timing circuits. The ratio between the clock frequency and
the cut-off frequency is 100. Finally, the amplified Doppler signals Q
A
and I
A
were
played by the stereo speakers and digitized by either a sound card, or a Gage A/D
card CompuScope-14200 (Gage Applied Technologies Inc., Montreal, Quebec,
Canada), depending on the bandwidth of the Q
A
and I
A
. The digitized Doppler
signals were converted into a directional spectrogram by a Labview software in real
72
time. Further off-line analysis was conducted using Matlab based software.
Figure 4.10 Timing sequences in the HF directional pulsed-wave Doppler system.
PRF A
C
Sample Depth
Sample Volume
B
D
E
Expanded
F
73
D. Flow Estimation
As described in the book (Evans and McDicken 2000), the flow velocity can be
estimated by:
c
f
f
v
θ cos 2
0
∆
= (4.1)
where f ∆ is the measured Doppler shift frequency, f
0
is the center frequency of the
transducer, c is the sound velocity in a medium, and θ is the angle between the
ultrasound beam and the flow. In ultrasound Doppler measurements, f ∆ , f
0
and c are
all known variables, and θ is often estimated from the corresponding B-mode images.
In our system, however, no B-mode images were obtained for estimating θ.
Therefore the estimated flow velocities must be interpreted appropriately.
As discussed above, the absolutely correct flow velocity may not be obtained
using this pulsed-wave Doppler system. Nevertheless it is still acceptable in our
application since the factor of interest to ophthalmologists is the blood flow velocity
variance before and after blood flow reestablishment in an occlusion removal surgery.
In order to detect the Doppler signals from a same vessel, the needle transducer must
be placed at the same position before and after blood flow reestablishment. Therefore
the θ should keep a constant theoretically for velocity estimation. In real cases,
however, the θ may vary slightly, depending on human positional accuracy. If the θ
variance is less than the blood flow velocity variance before and after blood flow
reestablishment, this Doppler system still can provide quantitative blood flow
information for ophthalmologists to determine the success of an occlusion removal
74
surgery. In the later sessions, the effects of the angle variance will be evaluated and
discussed based on both the in vitro and in vivo studies.
Eqn. (4.1) also indicates that flow velocity is proportional to Doppler shift
frequency. The signal bandwidth of the integrated sound card on the PC motherboard
is approximately 20 Hz~20 KHz, and its sampling frequency is up to 44.1 KHz.
These sound card specifications meet the requirements of acquiring the Q
A
and I
A
data in most cases. However, in order to detect extremely low flow velocities, or
extremely high flow velocities, neither the signal bandwidth of the sound card is low
enough, nor is its sampling frequency high enough. Consequently, the 14-bit Gage
A/D card was used. The signal bandwidth of the Gage card can be as low as 0 Hz,
and its sampling frequency can range from 50 KHz to 200 MHz with 14-bit
resolution.
Finally, proper temporal and velocity resolution can be determined as noted by
Christopher (1997), when each cardiac cycle contains at least 25 discrete time
intervals, and the peak-to-peak velocity range also contains similar numbers of
velocity intervals. Appropriate temporal and velocity resolutions were selected in
each experiment.
75
Experiments and Results
A. Minimal Detectable Velocity Measurements
As pointed out in the references (Ferrara et al. 1996; Silverman et al. 1999;
Foster et al. 2002), blood velocities as low as 0.6 mm/s exist in the microcirculation.
The minimally detectable velocity of a Doppler system demonstrates the system’s
sensitivity and accuracy. The same wire phantom, used in the lateral resolution
measurement, was used to calibrate and evaluate the Doppler system. Both the
needle transducer and the wire phantom were merged into a water tank filled with
deionized water. The Doppler angle θ was 45
○
. The distance between the tungsten
wire and the transducer was adjusted to be equal to the sample depth 1.5 mm. The
needle transducer mounted on a step motor (Model RFT-57-102-B11, Parker
Hannifin Corp., Rohnert Park, CA) traveled at a constant speed 100 µm/s up and
down. In the experiment, the PRF was 343 Hz, the sample volume was 7 cycles, and
no wall filters were used. The audio Doppler signals I
A
and Q
A
were digitized by the
Gage A/D card at sampling frequency of 50 KHz. Then the signals were down
sampled and processed by Matlab. The temporal resolution was 100 ms with a
temporal overlap of 80%, and the velocity resolution was 4.6 µm/s. As shown in Fig.
4.11, the measured Doppler shift frequency was approximately ±5 Hz, and its
corresponding velocity was ±100 µm/s which was the preset motor speed. The
positive and negative velocities indicated the needle transducer traveling toward to
the wire phantom, or away from the wire phantom respectively. The minimal
76
detectable velocity of 100 µm/s of this system should be sufficient for most
microcirculation studies.
Figure 4.11 Minimal detectable velocity measured from the wire phantom.
77
B. Micro-flow Phantom Evaluation
In order to evaluate the feasibility of studying the microcirculation with this
system, the micro-flow phantom was used to evaluate the system. The Doppler
signals were measured from blood-mimicking fluids in the polyimide tubes with
inner diameter from 127~574 µm. Both the tubes and the needle transducer were
submerged in deionized water. By adjusting the height of the blood-mimicking
solution, the flow velocity in the tube can be manipulated. The PRF was set to 44
KHz, the sample volume was 7 cycles at a depth of 2 mm, the Doppler angle θ was
estimated as 60
○
, and the sound card was used. A software high-pass filter with a cut-
off frequency of 5 Hz was applied to the digitized Doppler signals.
Fig. 4.12(a) shows the Doppler signals of the flow in the tube with diameter of
127 µm. The temporal resolution was 20 ms with a temporal overlap of 50%, and the
velocity resolution was 190 µm/s. As shown in Fig. 4.12(a), the velocity lower than
10 mm/s was measured from the tube, and its corresponding Doppler shift frequency
was 250 Hz. This figure also shows some strong Doppler signals which may be
caused by air bubbles in the solution.
Fig. 4.12(b) shows that the flow in the tube with a diameter of 320 µm varies
from low to high by gradually adjusting the height of the solution. The temporal
resolution was 23 ms with a temporal overlap of 50%, and the velocity resolution
was 413 µm/s.
78
(a)
(b)
Figure 4.12 Detected Doppler signals from the micro-flow phantom. The blood-
mimicking fluids in the tube with diameters of 127 µm (a) and 320 µm(b). In (b), the
fluid velocity was varying.
Doppler signals
from air bubbles
79
These figures demonstrate the capability of this system to study microcirculation.
However, one requirement for evaluating an occlusion removal surgery is that the
system with human positioning should be capable of making several flow velocity
measurements at the same position within acceptable errors. For this requirement, we
performed a repeating flow evaluation experiment, using a 3-D mechanical stage and
the micro-flow phantom. The needle transducer mounted on the 3-D mechanical
stage. It can be manually moved around and placed at the same position to simulate
the situation of inserting the transducer into the eye before and after the blood flow
reestablishment. The obtained Doppler spectrum is illustrated in Fig. 4.13. When the
transducer located at the micro-tube with inner diameter of 574 µm, Doppler signals
were detected. Otherwise, black spots were found on the spectrum. Fig. 4.13
indicates that the velocity variance caused by relocating the transducer can be
ignored, since the measuring position can be determined precisely by the 3-D
mechanical stage. This experiment shows that the Doppler system can measure the
flow velocities robustly. In the experiment, the PRF was set to 5.5 KHz, the sample
volume was 7 cycles at a depth of 2 mm, the Doppler angle θ was estimated as 45
○
,
and the sound card was used. A software high-pass filter with a cut-off frequency of
5 Hz was applied to the digitized Doppler signals. The temporal resolution was 23
ms with a temporal overlap of 50% and the velocity resolution was 126 µm/s.
80
Figure 4.13 Velocity variance measurements on the micro-flow phantom.
81
C. In vivo Rabbit Experiments
The preliminary experiments for measuring subretinal blood flow were carried
out in vivo on New England Cottontail rabbits at the Doheny Eye Institute of USC.
Under anesthesia, an incision was made on the side of the rabbit eye. Both the needle
transducers with 0° and 45° tip were used.
Under optical microscopic guidance, the needle transducer with 0° tip was
inserted through the incision and placed near the retina, where a branch retinal vein
and a branch retinal artery were close to each other shown in Fig. 4.14(a). Blood
flows in both the retinal vein and retinal artery with diameter less than 200 µm were
measured in real time. The PRF was set to 44 KHz, the sample volume was 7 cycles
at a depth of 2 mm, the Doppler angle θ was estimated as 30
○
.The temporal
resolution was 6 ms with a temporal overlap of 50%, and the velocity resolution was
200 µm/s. On the spectrogram shown in Fig. 4.14(b), the Doppler signal from the
artery can be separated from the one from the vein based on the flow directions and
the flow patterns. The vein Doppler signal shows the vein blood flow with low and
relatively constant velocity, while the artery Doppler signal shows the artery blood
flow with high and pulsatile velocity. The average maximum velocity in the retinal
vein was approximately 1 cm/s and the average maximum velocity in the retinal
artery was approximately 4 cm/s as shown in Fig. 4.14(b).
82
(a)
(b)
Figure 4.14 Measured retinal blood flows with the 0° needle transducer. (a) is the
captured photo by the video recorder. (b) is the Doppler spectrogram showing vein and
artery flows.
Artery
Vein
Vein
Artery
Transducer
83
Fig. 4.15(a) shows that the needle transducer with 45° tip was inserted through
the incision and placed on the top of the optical disk where the central retinal artery
(CRA) and central retinal vein (CRV) were coming out. The angle between the
ultrasound beam and the blood flows in CRA and CRV should be close to 0°. Since
the optical disk is untransparent, both the CRA and CRV were invisible on the
guidance optical microscope. The locations of CRA and CRV can only be
distinguished based on the patterns of the obtained audio Doppler signals. The
obtained Doppler spectrum is illustrated in Fig. 4.15(b) in which the average
maximum velocity in the retinal vein was approximately 1 cm/s and the average
maximum velocity in the retinal artery was approximately 6 cm/s. Fig. 4.15(b) also
shows that stronger Doppler signals were detected in the CRV than the CRA due to
the larger diameter of CRV. In the experiment, the PRF was set to 44 KHz, the
sample volume was 7 cycles at a depth of 2 mm, the Doppler angle θ was estimated
as 20
○
. The temporal resolution was 6 ms with a temporal overlap of 50%, and the
velocity resolution was 207 µm/s. During the above in vivo experiments, both the
audio and video signals were recorded by a Sony digital recorder.
84
(a)
(b)
Figure 4.15 Measured the blood flows in CRV and CRA with the 45° needle transducer.
(a) shows the tip of the transducer facing to the optical disk. (b) is the Doppler spectrum
from both the CRV and CRA.
Optical Disk
Central Retinal Artery
Central Retinal Vein
85
In order to evaluate the measured velocity error at the same position caused
human positional accuracy, a blind test was performed. Several measuring points
were selected in each rabbit eye by an ophthalmologist. The ophthalmologist
randomly placed the needle transducer at each position to pick up Doppler signals.
An engineering staff recorded the peak velocity of each measurement and predicted
the position of each measurement based on the real-time spectrogram. Then each
predicted position was verified by the ophthalmologist. After each measurement, the
needle transducer was removed from the eye to simulate the real surgical situation.
At least 10 measurements were made at each position. Total of 10 positions were
selected in 4 rabbit eyes. The mean velocity and standard deviation of the
measurements at each position were calculated and plotted in Fig. 4.16 (a). Then the
measurement error which is the ratio between the standard deviation and the mean
velocity is illustrated in percentage scale in Fig. 4.16 (b). The measurement errors
caused by human positional accuracy were less than 10 % in most cases. Lower
measurement errors are expected if the ophthalmologist performs more trials. The
same blind test was performed by another ophthalmologist as well, and the similar
measurement errors were obtained. As reported in (Baxter and Willisamson 1993),
the maximal CRV blood flow velocities in the normal eye and abnormal eye were
6.4±1.6 cm/s and 3.3±1.9 cm/s respectively. The difference between them is much
greater than 10%. Therefore, valuable clinical information can still be provided by
our system if blood flow is reestablished successfully.
86
(a)
(b)
Figure 4.16 Blood flow velocity estimation error caused by human positioning. (a)
shows the relationship between Doppler shift frequency and flow velocity, assuming the
Doppler angle of 20°. The error bars correspond to the standard deviation of
measurements at each position. (b) illustrates the estimated measurement error in
percentage scale caused human positional accuracy. The error is the ratio between
standard deviation and mean velocity at each measuring position.
87
Summary
In this chapter, we described the implementation of a high-frequency pulsed-
wave Doppler system with intraocular needle transducers. This system is capable of
detecting velocities as low as 100 µm/s with appropriate temporal and velocity
resolutions.
This system can provide valuable clinical information during retinal occlusion
removal surgeries in real time. The in vivo studies show that the system is capable of
detecting blood flow from retinal veins and arteries with diameters less than 200 µm.
The measured blood flow velocity errors which were caused by human positing the
intraocular transducer at the same position were less than 10%. Furthermore, the
insertion of the intraocular transducers through sclera doesn’t cause extra injuries to
patients since the transducer size is smaller than those of microsurgerical instruments
which are used in the occlusion removal surgeries. Human trials will be pursued in
the near future.
88
CHAPTER 5
HIGH-FREQUENCY ULTRASOUND PULSED-WAVE
DOPPLER SYSTEM FOR BIOMEDICAL APPLICATIONS
WITH 30 MHZ LINEAR ARRAY
Introduction
Preliminary studies of B-mode imaging with HF linear arrays were described in
Chapter 2. Mouse heart movements were captured at 100 frame/s using that system.
Cardiovascular and tumor research utilizing mice requires not only B-mode imaging,
but also ultrasound Doppler to evaluate blood flows (Reddy et al. 2005(a)(b); Goertz
et al. 2002). To our knowledge, the development of Doppler systems with HF linear
arrays has not yet been reported.
Several ultrasound Doppler techniques can be used to evaluate blood flow,
including continuous-wave Doppler, pulsed-wave Doppler, and color Doppler
(Jensen 1996). Continuous-wave Doppler detects average flow velocities of a large
sample volume, while pulsed-wave Doppler can determine flow velocities of a
smaller sample volume at a specific depth. Color Doppler imaging can be
implemented by further processing acquired B-mode image data and demodulated
Doppler data. Considering the limitations of large sample volume in continuous-
wave Doppler, and the design complexities of color Doppler, pulsed-wave Doppler is
suitable as an initial attempt for detecting blood flow in individual vessels with HF
arrays.
The implementation of pulsed-wave Doppler with ultrasonic arrays can be done,
after either analog receive beamforming, or digital receive beamforming (Brunner
89
2002). Although digital beamforming has been used in most clinical ultrasound
systems (Thomenius 1996; Brunner 2002), analog beamforming still has the
advantages of low cost, accurate delays, zero beamforming time, and large dynamic
range in HF ultrasound (Stitt et al. 2002). For example, analog beamforming is still
performed before continuous-wave Doppler in clinical ultrasound systems due to its
large dynamic range (Brunner 2002). The bandwidth of the analog beamformer
described in Chapter 2 exceeded 90 MHz which is much higher than the bandwidth
of the digital beamformer reported recently (Hu et al. 2006). Considering the above
advantages of analog beamformers, as an initial attempt, the pulsed-wave Doppler
after analog beamforming was implemented. Therefore combining the Doppler
system described in Chapter 3 with part of the improved B-mode image system, the
HF pulsed-wave Doppler system with a 30 MHz linear array was successfully
implemented and tested. The development of this system is described in this chapter.
System Description
A block diagram and a photo of the pulsed-wave Doppler system with HF arrays
are illustrated in Fig. 5.1.
As shown in Fig. 5.1(a), the system included two separated blocks: a 16-channel
analog beamformer block in the dash-dot line box, and a HF directional pulsed-wave
Doppler block in the dash line box. The main clock of the system was generated by a
programmable clock generator ECS-P83/P85 (ECS, Inc. International, Kansas City,
KS). The two blocks were synchronized by clocks which were distributed by a clock
90
driver CD328 (Texas Instruments, Dallas, TX). The 16-channel analog beamformer
block consisted of: a 30MHz array, timing circuits, 16-channel transmit beamformer,
N-cycle bipolar pulsers, analog front-ends, and a 16-channel programmable analog
receive beamformer. The beamformed echoes were directly fed to the HF directional
pulsed-wave Doppler block. The Doppler block included an in-phase and quadrature
(IQ) demodulator, low-pass filters, sample-and-hold circuits, PRF filters, wall filters,
and audio amplifiers. Doppler audio signals were played by two stereo speakers.
Meanwhile, the Doppler audio signals were digitized by either an integrated sound
card, or a Gage A/D card, depending on the bandwidth of the audio signals. Then the
digitized signals were processed and converted into a directional spectrogram by
Labview-based software (National Instruments Corp., Austin, TX) in real time.
Further off-line analysis was conducted using Matlab (The MathWorks, Inc., Natick,
MA) based software. Most commercial ultrasound systems operate in duplex mode
(i.e. displaying B-mode images and Doppler images simultaneously). However, this
system currently can only display either the Doppler spectrogram, or the B-mode
images in real time at a time. After expanding the analog front-end channel number
and upgrading the Labview software, the duplex image mode can be realized readily
by feeding the RF data to the Gage A/D card and feeding the audio Doppler signals
to the sound card simultaneously. The electronic circuits and array characteristics are
discussed in detail in the following.
91
(a)
(b)
Figure 5.1 Block diagram (a) and photo (b) of the HF ultrasound Doppler with array. A
is the 30 MHz array, B is the pulsed-wave Doppler block, C is the transmit beamformer
board with the microprocessor, D is the bipoloar pulser and analog front-ends board, E is
the analog beamformer board, F is the power supplies, and G is the PC.
92
A. Array Characteristics
A 30 MHz piezo-composite ultrasound array was used. It was fabricated at the
NIH Resource on Medical Ultrasonic Transducer Technology, using the same
mechanical dicing method described by Cannata et al. in 2006. The characteristics of
the array are listed in Table 5.1. Element 1 to Element 16 were used in the
experiments described in the following sections.
TABLE 5.1 MEASURED CHARACTERISTICS OF THE 30 MHZ ARRAY.
Characteristic Value
Number of elements 64
Pitch 100 µm
Element dimensions 2 mm×85µm
Average center frequency 29.8 MHz
Highest/lowest center frequency 30.6/28.8 MHz
Average bandwidth 56 %
Highest/lowest bandwidth 61%/52%
Average sensitivity 1.51 V
Highest/lowest sensitivity 1.63/1.35 V
-20 dB pulse length 92 ns
Insertion Loss 16 dB
B. Timing Circuits
The timing circuits consisted of a high-speed 8-bit TTL counter 74F269
(Fairchild Semiconductor, South Portland, ME), logic gates 74AC04, 74AC00 and
74AC573 (Fairchild Semiconductor, South Portland, ME), two 12-bit counters
74VHC4040 (TOSHIBA America, New York, NY), a microprocessor
89S8252(Atmel Corp., San Jose, CA), and digital delay lines PDU16F-1 (Data Delay
Device Inc., Clifton, NJ). The 30 MHz clock from the clock driver CD328 was
divided by two 12-bit counters 74VHC4040, making it possible to obtain a PRF
93
clock (114 Hz~117 KHz), a PRF filter clock (7.3 KHz~7.5 MHz), and a wall filter
clock (114 Hz~14.6 KHz). A sample volume gate (0.1-2 µs) at a specific sample
depth (0-50 µs) was produced by a PRF trigged monostable multivibrator 74HCT221
(Texas Instruments, Dallas, TX), and its waveform is shown in Fig. 5.2(B). As
shown in Fig. 5.2(C, D, E), the timing circuits provided a N-cycle (N=1~255) trigger
signal for the bipolar pulsers. Then the N-cycle trigger signal was duplicated to 16
identical triggers by two high-speed buffers 74AC573, since the driving capability of
74 series chips is insufficient. The 16 triggers were then sent to the 16-channel
transmit beamformer which consisted of programmable digital delay lines PDU16F-
1 and the microprocessor 89S8252. The delay time of PDU16F-1 can be
programmable from 0 ns to 63 ns with an increment of 1 ns. The appropriate delay
time for each channel, which was calculated by equation (2.1) before experiments,
was stored in the microprocessor 89S8252.
94
Figure 5.2 Timing sequences in the HF Doppler system with 30 MHz array.
C. Bipolar Pulsers and Analog Front-ends
After the transmit beamformer, the delayed trigger signals shown in Fig. 5.2 (D,
E) were applied to the sixteen bipolar pulsers which were described in Chapter 3.
PRF A
C
Sample Depth
Sample
B
D
E
Expanded
F
G
95
These sixteen pulsers were connected to Element 1 to Element 16 of the 30 MHz
array. Fig. 5.2 (F) and (G), for example, show the undelayed pulse for element 1 and
the delayed pulse for element 8. A 30 MHz 7-cycle bipolar pulse used in the
following experiments is shown in Fig. 5.3, as well as its spectrum.
Figure 5.3 Measured 7-cycle 30 MHz bipolar pulse and its spectrum.
The analog front-end consisted of 16 channel independent circuits. In order to
prevent high voltage pulses from damaging the amplifiers, a limiter including a pair
of back-to-back diodes was added before the pre-amplifier of each channel as shown
in Fig. 5.4 (Lockwood et al. 1991). Then the received echoes from each element
were amplified. Each channel included a variable gain amplifier AD8331 (Analog
Devices, Norwood, MA), an optional band-pass filter PBP-35W (Mini Circuits,
Brooklyn, NY), and an optional second amplifier MAX4107 (Maxim Integrated
-0.1 0 0.1 0.2 0.3
-30
-15
0
15
30
Time ( µs)
Measured Amplitude (V)
15 30 45 60 75
-60
-40
-20
0
Frequency (MHz)
Normalized Amplitude (dB)
96
Products, Sunnyvale, CA). The total achievable amplification was 70 dB. After the
amplification, the 16-channel echoes were beamformed by the analog beamformer
described in Chapter 2.
Figure 5.4 Block diagram of one channel analog front-end circuit.
D. HF Pulsed-wave Doppler Block
The beamformed echoes were first filtered by a band-pass filter BBP-30 (Mini-
Circuits, Brooklyn, NY), and the echoes were fed to an in-phase and quadrature (IQ)
demodulator MIQC-60WD (Mini-Circuits, Brooklyn, NY). The reference clock of
the IQ demodulator was generated from one output clock of the clock driver CD328,
after a low-pass filter PLP-30 (Mini-Circuits, Brooklyn, NY) and an attenuation
network. The demodulated intermediate frequency (IF) signals I
IF
and Q
IF
were low-
pass filtered to remove harmonics and noise by BLP-10.7 (Mini-Circuits, Brooklyn,
NY). Then they were sampled and held by AD783 (Analog Devices, Norwood, MA).
Followed by 8
th
-order PRF filters which consisted of MAX291 switch-capacitor
filters (Maxim Integrated Products, Sunnyvale, CA), the I
SH
and Q
SH
signals were
cleaned by removing sample-and-hold harmonics. Before the I
A
and Q
A
were
amplified by the audio amplifiers LM386 (National Semiconductor Corp., Santa
97
Clara, CA), optional 4
th
-order wall filters MAX7490 (Maxim Integrated Products,
Sunnyvale, CA) were used to remove low frequency clutter signals. The cut-off
frequencies of both the PRF filters and wall filters were adjusted by their clocks from
the timing circuits. The ratio between the clock frequency and the cut-off frequency
is 100. The amplified Doppler signals Q
A
and I
A
were played by the stereo speakers
and digitized by either a sound card, or a Gage A/D card CompuScope-14200 (Gage
Applied Technologies Inc., Montreal, Quebec, Canada), depending on the bandwidth
of the Q
A
and I
A
. The digitized Doppler signals were converted into a directional
spectrogram by a Labview software in real time. Further off-line analysis was
conducted using Matlab based software.
Experiments and Results
Both in vitro and in vivo experiments were carried out. First, the lateral
resolutions of the array with 16 channel transmit and receive beamformer were
measured, using a wire phantom with five 20 µm tungsten wire arranged diagonally
(California Fine Wire, Grove Beach, CA) (Cannata 2004; Huang and Shung 2005 ).
Then the Doppler system with HF array was evaluated with the wire phantom to
determine its accuracy and minimal detectable velocity. Thirdly, the micro-flow
phantom with inner diameter of 127 µm described in the previous chapter was used
to evaluate the system. The flow velocities were measured as the blood-mimicking
fluid passed through the 127 µm tube. The flow velocity in the micro-flow phantom
can be adjusted by gravity. Next in vivo studies were carried out on mice. In all
98
experiments, an oscilloscope was used to monitor the sample gate and the
beamformed echoes, ensuring that the sample gate located at the correct positions of
the wire, tube and vessels. Finally several in vitro images were obtained in order to
demonstrate the imaging capability of the system.
A. Lateral Resolution Measurements
The array and the wire phantom were mounted on a 3-axis positioning system
(Burleigh Inc., Fishers, NY), and placed in a water tank filled with deionized water.
The wire phantom was moved laterally by a step of 50 µm. The 16-channel transmit
and receive beamformers were electronically focused at 4.8 mm, 6.4 mm, 8.0 mm
and 9.6 mm respectively at each step. The one channel beamformed RF data was
digitized by the Gage A/D card with sampling frequency of 200 MHz. The lateral
beam profile of 16 elements of the array at different depths was measured and
illustrated in Fig. 5.5. In Table 5.2, the measured lateral resolutions satisfied those
calculated by (2.2). The axial resolution of the array was determined by the
transmitted pulse length. Seven cycle pulses were used to obtain enough signal-to-
noise ratio (SNR) in our Doppler experiments.
99
Figure 5.5 Measured lateral resolutions of the 30 MHz array. The 16-channel transmit
and receive beamformers focused at the depths of 5.0 mm, 6.6 mm, 8.0 mm, and 9.5 mm.
TABLE 5.2 LATERAL RESOLUTIONS OF THE ARRAY WITH 16-CHANNEL BEAMFORMER.
Focal Distance Calculated Measured
5.0 mm 154 µm 160 µm
6.6 mm 204 µm 220 µm
8.0 mm 247 µm 275 µm
9.5 mm 293 µm 310 µm
Note: A 1485 m/s sound velocity in water was assumed (Christopher et
al. 1997).
B. Minimal Detectable Velocity Measurement
As pointed out in (Pavlin et al. 1998; Silverman
et al. 1999), blood velocities as
low as 0.6 mm/s exist in microcirculation. The minimally detectable velocity of a
Doppler system demonstrates the system’s sensitivity and accuracy. The same wire
phantom, used in the lateral resolution measurements, was used to calibrate and
100
evaluate the Doppler system with the 30 MHz array. Both the array and the wire
phantom were submerged in a water tank filled with deionized water. The Doppler
angle θ was 0
○
. The distance between the tungsten wire and the array was adjusted to
be equal to the sample depth 8.0 mm. The array mounted on a step motor (Model
RFT-57-102-B11, Parker Hannifin Corp., Rohnert Park, CA) traveled at a constant
speed 100 µm/s up and down. The PRF was 228 Hz, the sample volume was 7 cycles,
and no wall filters were used. The audio Doppler signals I
A
and Q
A
were digitized by
the Gage A/D card at sampling frequency of 50 KHz. Then the signals were down
sampled and processed by Matlab. The temporal resolution was 100 ms with a
temporal overlap of 80%, and the velocity resolution was 5 µm/s. As shown in Fig.
5.6, the measured Doppler shift frequency was approximately ±4 Hz, and its
corresponding velocity was ±100 µm /s which was the preset motor speed. The
positive and negative velocities indicated the array traveling toward to the wire
phantom, or away from the wire phantom respectively. The up and down traveling
velocities of the step motor were slightly different due to gravity. Therefore, the
measured velocities in Fig. 5.6 are slightly asymmetric. The system’s minimum
detectable velocity of 100 µm/s should be sufficient for most microcirculatory
studies.
101
Figure 5.6 Measured minimal detectable velocity with the wire phantom
C. Micro-flow Phantom Tests
The micro-flow phantom was used to mimic small vessels in the
microcirculation. The polyimide tube with inner diameter of 127 µm was submerged
in water. The flow velocity in the tube was varied by adjusting the height of the
blood-mimicking solution. The PRF was set to 7.3 KHz, the sample volume was 7
cycles at a depth of 8 mm, the Doppler angle θ was estimated as 45
○
, and the sound
card was used. A software high-pass filter with a cut-off frequency of 5 Hz was
applied to the digitized Doppler signals. The temporal resolution was 25 ms with a
temporal overlap of 50% and the velocity resolution was 190 µm/s. As shown in Fig.
5.7, the velocity lower than 7 mm/s was measured from the tube, and its
102
corresponding Doppler shift frequency was 200 Hz. This figure also shows some
strong Doppler signals which may be caused by air bubbles in the solution.
Figure 5.7 Measured Doppler signals from the micro-flow phantom. Doppler
spectrogram showing detectable velocity lower than 7 mm/s in a 127 µm tube filled with
blood-mimicking fluid.
Doppler signals from
micro air bubbles
Doppler signals from
blood-mimicking fluid
103
D. In vivo Studies
In both clinical studies and small animal studies, the parameters of blood flows,
such as the peak velocities, mean velocities and peak velocity intervals, are often
used to identify hemodyanmic diseases( Reddy et al. 2005b). Therefore, evaluating
blood flow on animal models with this array-based HF Doppler system is necessary
and meaningful.
At the Health Science Campus at USC, in vivo Doppler studies were carried out
on mice. After anesthetizing and shaving, the mice were fixed on a custom-designed
stage. Water was used as a coupling medium. Two abdominal superficial vessels and
several major arteries close to the heart were selected as the measuring points. Since
only 16-channel front-end electronics were built, the system is not capable of
producing real-time B-mode images. Therefore a 50 MHz sector scanner developed
at the Resource Center was used to produce real-time B-mode images and locate
some vessels. In all cases, 7-cycle bipolar burst pulses were applied to the elements 1
to 16, and this array sub-aperture was electronically focused at 8 mm. The audio
Doppler signals were digitized by the sound card and then processed off-line further
by Matlab-based software.
Fig. 5.8 is the B-mode image obtained by the 50 MHz sector scanner (Sun et al.
2006a), and two abdominal superficial vessels with diameters about 200 µm can be
identified. We adjusted the positions of the array in order to detect the Doppler
signals from these vessels. Fig. 5.9 shows the spectrograms of the detected Doppler
signals. The spectrogram in Fig. 5.9(a) indicates that the detected Doppler signals
104
were from both the artery and vein. Therefore, the positive pulsatile velocity
corresponds to the Doppler signals from the artery, and the negative relatively
constant velocity corresponds to the Doppler signals from the vein. The heart rate of
the mouse can also be approximately estimated as 300 beat/min based on the
intervals between two velocity peaks. Fig. 5.9(b) illustrates the detected Doppler
signals from the vein only.
Figure 5.8 B-mode image of abdomen superficial vessels. This image was obtained by
the sector scanner, showing the vessels with diameters about 200 µm.
Vessels
105
(a)
(b)
Figure 5.9 Measured Doppler signals from the abdomen superficial vessels. (a) shows
the signals from both the artery and vein. (b) shows the signals only from the vein.
Vein
Artery
106
As shown in Fig. 5.10(a), multiple vessels exist close to the mouse heart. As
reported by Reddy et al. in 2005b, the Doppler spectrograms from these vessels are
different due to the different hemodynamics in different vessels. These
hemodynamics may be used to diagnose cardiovascular diseases. Therefore, we
measured the Doppler signals from several major arteries which were closed to the
mouse heart. Due to the limitation of our system, we are not able to obtain the B-
mode images of these vessels simultaneously. Nevertheless, the obtained
spectrograms illustrated in Fig. 5.10 are still valuable for our future studies. Based on
the parameters of the Doppler spectrograms in Fig. 5.10, the measuring points may
include the carotid, mitral and aorta.
These in vivo studies demonstrated that this array-based Doppler system is able
to detect flow in small vessels with diameters less than 200 µm and detect flow with
velocities greater than 1 m/s. These specifications are fairly promising for carrying
out future small animal and skin cancer studies with this system.
107
(a) (b)
(c) (d)
(e) (f)
Figure 5.10 In vivo Doppler studies on mice with the 30 MHz array. (a) illustrates the
cardiac vessels. (b)~(f) illustrate the different hemodynamics of the Doppler signals
from different arteries. The measured arteries may include the aorta, mitral and carotid.
(a) was approved for redrawing and reprinting by Reddy (2005b).
108
E. In vitro B-mode Imaging Studies
As mentioned in Chapter 2, the previous B-mode imaging system with HF array
was not able to obtain satisfactory images due to the noisy front-end electronics and
8-bit digitization card.
As described in this chapter, both the low noise analog front-end electronics and
14-bit digitization card have been utilized in the array-based Doppler system, and
significant improvements have been noticed based on both the in vitro and in vivo
studies. However, the current system only has 16-channel bipolar pulsers and front-
end electronics, and it is unable to obtain real-time B-mode images with the acoustic
beam electronically scanned. Nevertheless, it is still possible to obtain B-mode
images with the acoustic beam mechanically scanned.
Therefore, in vitro B-mode imaging studies were conducted using the array-
based Doppler system. The wire phantom, pig eye and rabbit eye were imaged. We
mounted the transducer on a step motor and the transducer can be moved by a step of
50 µm or 100 µm. The sub-aperture of the array included element 1 to 16, which
were excited by monocycle pulses in order to achieve better axial resolution. The
received echoes were amplified by the amplification board with AD8331 and then
band-pass filtered. At each position, the supaperture of the array was electronically
focused at 4.8 mm, 6.4 mm, 8 mm, and 9.6 mm respectively with both the transmit
and receive beamformers. The imaging lines were digitized by the Gage 14-bit AD
card, and then the data was processed offline by the Matlab based software.
109
Fig 5.11~13 shows the B-mode images of the wire phantom, pig eye and rabbit
eye. Significant image quality improvement can be noticed, compared to the images
illustrated in Fig. 2.11~13. In addition, a wire phantom image was also obtained by
the UBM with a 35 MHz single element transducer as Fig. 5.11(b) shows.
Comparing the two wire phantom images in Fig. 5.11, the image obtained by the
array-based imaging system can achieve more constant lateral resolutions, while the
image obtained by the UBM can only obtain a good lateral resolution at its focal
point. This comparison demonstrated the advantages of the HF ultrasound imaging
system with arrays. In addition, the lateral resolution of this array-based system can
be further improved by expanding the subaperture from 16 elements to 32 elements.
Satisfactory real-time images may be expected when the array-based Doppler system
is expanded to support the whole 64-element array.
110
(a)
(b)
Figure 5.11 Wire phantom images obtained by the array Doppler system (a) and the 35
MHz UBM (b). 50 µm step was used in (a).
111
(a)
(b)
Figure 5.12 Pig eye images in vitro. The array was mounted on a step motor moving at a
step of 100 µm.
112
(a)
(b)
Figure 5.13 Rabbit eye images in vitro. The array was mounted on a step motor moving
at a step of 100 µm.
113
Summary
In this chapter, we described the development of the first HF directional pulsed-
wave Doppler system with a 30 MHz linear array. The system has a variable PRF
from 114 Hz to 117 KHz which corresponds to maximal detectable velocities from 3
mm/s to 3 m/s, assuming a Doppler angle of 60
○
. The system has been shown to be
capable of detecting the motion velocities of the wire phantom as low as 0.1 mm/s,
and detecting the blood-mimicking flow velocity in the 127 µm tube lower than 7
mm/s with suitable temporal resolution (25 ms to 100 ms) and velocity resolution (5
µm/s to 190 µm /s). The in vivo studies demonstrated the applications of such a
Doppler system in biomedical applications. This system provided additional blood
flow measurement capabilities to the imaging systems with HF arrays.
Furthermore, the improved SNR of this system, compared to the system
described in Chapter 2, enables us to obtain comparable B-mode images to current
UBM images. It is anticipated that this ultrasound system with both Doppler and B-
mode imaging capabilities can be used in small animal and cancer research in the
future, achieving better performance and higher frame rate than current UBMs with
single element transducers.
114
CHAPTER 6
HIGH-FREQUENCY COLOR DOPPLER AND POWER
DOPPLER WITH ARRAYS
Introduction
As discussed in the previous chapters, both pulsed-wave Doppler and
continuous-wave Doppler are only able to detect flows in a limited sample volume.
However, color flow mapping (i.e. color Doppler) can display the instantaneous
velocity in a vessel. It can provide valuable clinical information about the occlusion
of vessels, heart valve deficiencies, and other cardiovascular diseases.
To date, color Doppler is a basic function on most clinical ultrasound systems,
and it is also being studied with HF UBMs (Goertz et al. 2000). Usually, the color
Doppler images which show both the flow velocities and directions are
superimposed onto B-mode images. Therefore, both the blood flow and anatomical
information can be obtained from the superimposed images. For example, using the
color Doppler function embedded in the HF UBMs, HF color flow imaging provides
useful information of the microcirculation in tumor research. Recent studies have
also demonstrated that color Doppler imaging may be used in therapeutic
applications (Goertz et al. 2002).
In the previous chapters, HF ultrasound imaging systems with arrays have been
developed and used for small animal research. To our best knowledge, however, HF
color Doppler using arrays has not been reported yet. It is anticipated that HF color
flow imaging capability will become an integral part of future HF array-based
115
ultrasound systems.
As discussed in Chapter 1, both the autocorrelation and cross-correlation
algorithms are being used in color Doppler imaging systems. Since the early 1980s’,
these two algorithms have been studied extensively and shown distinct advantages
and disadvantages. Although the time-domain cross-correlation method achieves
superior noise immunity, no velocity aliasing, and better spatial resolution, the
phase-domain autocorrelation is still used in most color Doppler machines due to its
cost effectiveness. Therefore, we decided to use the autocorrelation-based color
Doppler to visualize blood flows in an initial attempt. Because only 16 channels
were implemented in the array-based Doppler system. M-mode color Doppler (i.e.
successive image lines are collected at the same position) was first studied in this
chapter.
Methods
Figure 6.1 illustrates the basic block diagram of color Doppler system. After the
beamformer and amplifiers, IQ demodulation is performed. The IQ demodulated
signals of each image line are sampled and saved to the on-board memory using the
A/D card. Then clutter filters are applied to the digitized signals in order to remove
the stationary signals from tissues or vessel walls. The filtered signals enter the
following autocorrelation-based velocity estimator. Finally, an appropriate color-
mapping table, which may be set up based on both the estimated velocity and power
of each image pixels, is used to obtain color Doppler images. Finally, the color
116
Doppler images is superimposed onto the B-mode images. The details of each
function block are described in this chapter.
Figure 6.1 Block diagram of a color Doppler imaging system.
A. Clutter Signal Rejection
As shown in Figure 6.1, clutter filters, which perform the same function as those
in pulsed-wave Doppler systems, are first studied to remove the signals from vessel
walls or surrounding tissues in the digitized image line data. These signals have very
high amplitudes and contain only low-frequency components, compared to the
signals from blood. Based on these characteristics, a variety of algorithms, including
finite impulse response (FIR) filters, infinite impulse response filters (IIR), and
regression filters, can be utilized (Tysoe and Evans 1995). Depending on the
different clutter rejection requirements, different wall filters must be selected based
on their performances which are mainly due to their frequency responses.
Among the above methods, the simplest FIR wall filter, single echo canceller, is
described in this chapter. It is implemented by subtracting two successive image
lines as shown in Fig. 6.2. Since the movements of tissues are fairly slow, the echoes
corresponding to tissues in two successive image lines remain approximately the
same. Consequently, only echoes from blood are conserved after the single echo
Q I
I
Q
I
Q
Beamformer &
Amplifiers
IQ
Demodulator
A/D with
Memory
Clutter
Filters
Velocity
Estimator
Color
Table
Superimposed
Images
B-mode Image
Processing
117
canceller. The transfer function of the single echo canceller is derived in (6.1), and
its corresponding frequency response is illustrated in Fig. 6.3 as well.
Figure 6.2 Block diagram of the single echo canceller.
1
1 ) (
−
− = Z Z H (6.1)
Figure 6.3 Frequency response of the single echo canceller.
Although the single echo canceller can be applied by simple hardware shown in
Fig. 6.2, its performance is insufficient for most real cases. In Fig. 6.3, it is shown
that the single echo canceller has a wide transition band and its frequency roll-off is
DELAY
(1/PRF)
OUTPUT INPUT
118
only 6 dB/octave. Based on the frequency response in Fig. 6.3, the single echo
canceller has a –3 dB cut-off frequency at the half normalized frequency PRF/4. That
means that all signals whose frequency is lower than PRF/4 are dramatically
attenuated. Therefore, the single echo canceller is not suitable for filtering out the
noise in the blood signals which are from the slow blood flow. The single echo
canceller may achieve satisfactory performance, when the detected signals only
contain high Doppler shift frequencies, for example, the frequencies are higher than
PRF/4.
Consequently, higher order FIR filters or IIR may be used to provide narrow
transition bands. However, the implementation of high order filters requires powerful
signal processors. Furthermore, high-order FIR filters requires more sample data for
flow velocity estimation. The increasing of sample data may decrease the frame rate
of color flow imaging. Similarly, high-order IIR filters may cause bias of frequency
estimation due to its poor transient response. Several new algorithms supported by
latest powerful DSPs are being studied with the hope of achieving better clutter
signal rejections (Cloutier et al. 2003; Shamdasani et al. 2004).
B. Autocorrelation Based Flow Velocity Estimation
After clutter rejection, the autocorrelation based flow velocity estimator is
performed in phase domain. One derivation of this velocity estimator was given by
Evans and McDicken (2000). Due to blood flow, the angular Doppler shift frequency
is defined as:
119
T dt
d
i i 1 −
Φ − Φ
≈
Φ
= ω (6.2)
where Φ is the phase shift relative to the system clock at time t, and T is the time
between two successive image lines( i.e. PRF in our system). Using the
trigonometric relationships, (6.2) can be rewritten as:
1 1
1 1
1
1
1
sin sin cos cos
sin cos cos sin
) cos(
) sin(
) tan(
− −
− −
−
−
−
Φ Φ + Φ Φ
Φ Φ − Φ Φ
=
Φ − Φ
Φ − Φ
= Φ − Φ
i i i i
i i i i
i i
i i
i i
(6.3)
Since the quadrature signals are acquired as shown in Fig. 6.1, the cos and sin terms
in (6.3) can be replaced by the quadrature signals I and Q, which are illustrated in
Figure 6.4. The i and i-1 correspond to two successive image lines. Therefore, (6.2)
can be derived as:
) 1 ( ) ( ) 1 ( ) (
) 1 ( ) ( ) 1 ( ) (
) tan(
1
− + −
− − −
= Φ − Φ
−
i Q i Q i I i I
i Q i I i I i Q
i i
(6.4)
In color Doppler imaging, n image lines are usually used to estimate the average
angular Doppler frequency ω :
}
) 1 ( ) ( ) 1 ( ) (
) 1 ( ) ( ) 1 ( ) (
{ tan
1
1
1
1
∑
∑
=
=
−
− + −
− − −
=
n
i
n
i
i Q i Q i I i I
i Q i I i I i Q
T
ω (6.5)
Finally, the estimated flow velocity v is:
}
) 1 ( ) ( ) 1 ( ) (
) 1 ( ) ( ) 1 ( ) (
{ tan
cos 2 2 cos 2
2
1
1
1
0 0
∑
∑
=
=
−
− + −
− − −
= =
n
i
n
i PRF
i Q i Q i I i I
i Q i I i I i Q
f
c f
f
c
v
θ π θ
π
ω
(6.6)
where c is the sound velocity in a medium, f
0
is the center frequency of transducer, θ
120
is the Doppler angle between the acoustic beam and flow, and f
PRF
is the PRF of the
image lines. After the velocity estimation, an appropriate color table should be
chosen to indicate the flow velocities and directions.
Figure 6.4 IQ signals used for calculating the angular Doppler shift frequency. (a) shows
the IQ signals of successive image lines in phase domain. (b) shows the real
demodulated IQ. I(i), Q(i), I(i-1), and Q(i-1) in (a) can be calculated by measuring the
amplitudes of IQ signals in (b) at the specific depth (dash lines).(note: in order to show
the significant amplitude difference between Line i and Line i-1, the I and Q image lines
in (b) are not the two successive lines.)
I(i-1)
I(i)
Q(i-1)
Q(i)
In-phase signals Quadrature signals
(b)
(a)
Q(i)
Q(i-1)
I(i-1)
Φ
i-1
Φ
i
I(i)
121
C. Velocity Mapping and Power Doppler
Different user-defined color tables can be used to encode the flow velocities,
with blue shades typically indicating negative flow velocities (i.e. blood moving
away from the transducer), and reds indicating positive flow velocities (i.e. blood
moving towards the transducer). The intensities of the reds or blues show the
magnitudes of positive or negative velocities.
Based on the chosen color table, color Doppler images can be obtained. If these
images are superimposed onto the B-mode images, both anatomical and flow
velocity information can be found. Because of the fact that blood flow only exists in
certain area of anatomical structures, color Doppler images are only superimposed
onto the B-mode images in those areas where blood flow exists. Whether to display
color Doppler images or B-mode images in a certain area is usually determined by
the estimated Doppler power of that area. B-mode images are displayed where the
Doppler power is insignificant; otherwise color flow images are displayed. Simply,
the Doppler power can be calculated using the amplitudes of the demodulated
quadrature signals I and Q as:
∑
−
=
+
−
=
1
1
2 2
)] ( ) ( [
1
1
n
i
d
i Q i I
n
P (6.7)
where I(i) and Q(i) are the amplitudes of the demodulated quadrature signals of
successive image lines at the sample volume, and n is the number of image lines.
Using the estimated Doppler power with (6.7), we also are able to generate
power Doppler images to show the existence of flow. In clinical systems, both the
122
color Doppler images and the power Doppler images are interchangeable depending
on the user requirements. In our experiments, both the color Doppler images and
power Doppler images were studied. These images are presented in the later sections.
D. System Setup
Using the array-based pulsed-wave Doppler system, we are able to study HF
color Doppler and power Doppler with arrays. The system block diagram is shown
again in Fig. 6.5.
Figure 6.5 Block diagram of the system with 30 MHz array for color Doppler studies.
The demodulated I
F
and Q
F
were acquired by the PRF triggered Gage A/D card
at a sample frequency of 200 MHz. The Gage A/D card synchronized by PRF can
acquire more than 40,000 I
F
and Q
F
signals per second. This specification of the
Gage A/D card ensures that no aliasing would occur in most cases. The acquired
123
image line numbers are limited by the on-board memory of the Gage card. The
acquired I
F
and Q
F
demodulated signals were post-processed offline using Matlab.
Experiments and Results
At the Health Science Campus at USC, initial color Doppler studies were carried
out on mice. After anesthetizing and shaving, the mice were fixed on a custom-
designed stage. Water was used as a coupling medium. Seven-cycle bipolar burst
pulses were applied to the array sub-aperture which was electronically focused at 8
mm. The pulsed-wave Doppler system was first used to locate a vessel and obtain
clear audio Doppler signals which were heard from the stereo speakers. Then, the I
F
and Q
F
were simultaneously digitized at 200 MHz. A total of 16000 image lines were
stored in the on-board memory of Gage A/D card, and then post-processed by
Matlab.
The flow velocity was estimated using the autocorrelation algorithm without
clutter signal rejections. The sample volume used to generate the color Doppler
image was determined by the transmitted pulse length. Every forty successive image
lines were used to estimate the flow velocity (i.e. temporal resolution was 40/PRF).
In the commercial real-time velocity estimators, the temporal resolution is usually
determined by the calculation power of DSPs and the required frame rate of color
Doppler imaging (Evan and McDicken 2000 ). In general, a coarser temporal
resolution may achieve more accurate velocity estimation.
In the in vivo studies, the system was used to identify a single artery close to the
124
mouse heart and two abdominal superficial vessels.
A. Single Vessel Detection
A major artery close to the mouse heart was identified by the pulsed-wave
Doppler system. The demodulated quadrature signals were digitized by the Gage
A/D card. Then the digitized IQ signals were processed by a software pulsed-wave
Doppler algorithm, and the audio Doppler signals at the depth of 8 mm were
obtained (Sun et al. 2006b). Using the autocorrelation velocity estimator, the
obtained color Doppler image is shown in Fig. 6.6 (a), as well as its corresponding
Doppler spectrogram at 8 mm in Fig. 6.6(b). A color table in which the brightest red
and blue shadows present the maximal positive velocity of 12 cm/s and negative
flow velocity of –12 cm/s respectively was used to map the flow velocities. Fig.
6.6(b) is the pulsed-wave Doppler spectrogram at the depth of 8 mm, and it indicates
that the flow velocity varies gradually from positive to negative. The color Doppler
image in Fig. 6.6(a) shows the similar flow velocity trend at the depth between 8 mm
and 9 mm. The skin, moving at low velocities, locates at the depth between 5 mm
and 6 mm. Then, the power Doppler plot and the power Doppler image are displayed
in Fig. 6.7, in which the Doppler power is displayed in the logarithm scale. Fig. 6.7(b)
indicates that the Doppler power from skin after clutter rejection is still about 15 dB
higher than that from blood. From this figure, it illustrates the vessel diameter of
approximate 1 mm. In this case, the single echo canceller (i.e. 1
st
order FIR filter) is
not suitable for removing clutter signals due to its poor frequency response.
125
(a)
(b)
Figure 6.6 M-mode color Doppler image of a mouse vessel (a) and its corresponding
pulsed-wave Doppler spectrogram at the depth of 8 mm.
126
(a)
(b)
Figure 6.7 Power Doppler image (a) and plot (b) at the time of 0.33s. The figures show
that the Doppler power from skin is 15 dB higher than that from blood. The vessel
diameter can also be calculated as 1 mm from (b).
Skin
Vessel
127
B. Multiple Vessels Detection
As discussed in the previous chapters, traditional pulsed-wave Doppler systems
are not capable of identifying vessels at multiple depths simultaneously. In tumor
research, however, the density of blood vessels and capillaries are usually used to
monitor the growth of tumor (Goertz et al. 2002). Therefore, the array-based color
Doppler imaging system, capable of detecting blood flows from multiple sample
depths, can be extremely useful for small animal research.
Consequently, initial studies for detecting multiple vessels using the array-based
color Doppler imaging system were carried out on mice. Two abdominal superficial
vessels were chosen in this experiment. The experimental setup was almost same as
that in the previous experiment, except the PRF was 7 KHz due to the low flow
velocities in the measured vessels. In Fig. 6.8 (a), two vessels exist at the depths of 8
mm and 10 mm respectively. The existence of the two vessels was verified based on
the B-mode image shown in Fig. 6.8 (b). The estimated vessel diameters are about
200 µm.
128
(a)
(b)
Figure 6.8 M-mode color Doppler image (a) and B-mode image of two vessels. The
vessels locate at the depths of 8 mm and 10 mm respectively based on (a) and (b).
Vessels
Vessels
129
Summary
In this chapter, we described the development of HF color Doppler and power
Doppler system with arrays. The initial studies show the feasibility of acquiring color
Doppler and power Doppler images using the current Doppler system with HF linear
arrays. The obtained experimental results show that this color Doppler imaging
system is capable of detecting both the single vessel and multiple vessels. These
results indicate the system’s potential application areas in small animal research and
skin cancer research.
130
CHAPTER 7
FUTURE WORK AND CONCLUSION
Future Work
In order to achieve the goal of visualizing tissue movements and blood
circulations better, further studies still need to be done.
A. Better Lateral Resolution and Wider Field of View
The field of view of the system with a 64-element array was less than 4.8 mm in
the azithumal direction, and the lateral resolution of the system was about 160 µm.
These specifications are still inadequate, compared to the 30 µm lateral resolution
and the 20 mm field of view obtained by the 80 MHz commercial UBM system
(Visualsonics 2006). No doubt, more efforts must be made in developing an array-
based system with a better lateral resolution and wider field of view.
The field of view and lateral resolution of an ultrasound system with arrays
depend on the element number of the arrays and the channel number of the
beamformer. In conventional ultrasound systems, linear arrays with more than 128
elements, supported by the 32-channel or 64-channel beamformer, are required to
obtain a satisfactory resolution and field of view (Shung 2005; Brunner 2002).
Therefore, increasing the channel number of our beamformer and utilizing 96 or 128
element linear arrays will alleviate the mentioned limitations. In the Resource center,
both an array with 128 elements and the 32-channel analog beamformer are being
developed.
131
B. Duplex Imaging System with Higher Frame Rate
Most commercial ultrasound systems operate in duplex mode (i.e. displaying B-
mode images and Doppler images simultaneously). Therefore developing a duplex
HF imaging system with arrays should be the logical next step.
Initial studies have shown that the Doppler system with arrays is capable of
acquiring both satisfactory Doppler images and static B-mode images. Thus,
expanding this system should be relatively straightforward. The current Doppler
system with arrays has 16 independent analog channels, each of which included a
bipolar pulser, a limiter, and two-stage amplifiers. In addition, each analog channel
can be enabled or disabled. These specifications allow us to implement software
multiplexers and demultiplexers in order to scan the sub-aperture of a linear array
electronically. After the analog channels, a new analog receiving beamformer has
been designed and tested. Each analog receiving beamformer board included 8-
channel 6-bit programmable delay circuits. Therefore either 32-channel or 64-
channel receiving beamformer can be constructed easily by using multiple these
beamformer boards. Finally, after upgrading the imaging software, the duplex image
mode can be realized readily by feeding the RF data to the Gage A/D card and
feeding the audio Doppler signals to the sound card simultaneously. The block
diagram of the proposed new system is shown in Fig. 7.1(a), as well as its mother-
daughter boards’ structure in Fig. 7.1(b).
132
(a)
(b)
Figure 7.1 Block diagram of the proposed new system (a) and its mother-daughter
boards’ structure (b).
m
Daughter Board
8-CH Bipolar Pulsers
8-CH Limiters
8-CH TGC Amplifiers
Daughter Board
8-CH Analog
Rx Beamformer with
RAMs
Timing Circuits, Tx Beamformer, and Doppler Processor
(Software Multiplexers, Demultiplexers and Cross-point Switches)
Mother Board
133
In addition, the achieved 100 frame/s of the current system is still insufficient for
imaging the mouse heart beating at 500~600 beats/min. More than 400 frame/s is
expected in small animal cardiovascular research (Liu et al. 2006; Cherin et al. 2006).
Basically, the frame rate depends on both the field of view (i.e. image depth and
width) and the configuration time before acquiring each image line. Since the field of
view is always determined before experiments, reducing the configuration time is the
only way to increase the system frame rate. This configuration time includes the time
of selecting appropriate analog channels for electronically scanning and the time of
programming the appropriate delay time to each delay circuit.
In the current design, the control board sent out the configuration data byte by
byte. In consequence, total of 140 µs were used to finish the configuration. In order
to reduce this configuration time, RAMs will be used to distribute all of these
configuration data as Fig. 7.2 shows. Before imaging, all line-by-line configuration
data are written to RAMs from the microcontroller. During imaging, the
microcontroller only provides an image line trigger which drives an image line
counter. Based on the image line number at the counter, corresponding data saved in
RAM is distributed to the destination circuits, such as the analog channels and the
programmable delay circuits. From the timing analysis diagram Fig. 7.2(b), the total
time consumed for configuration is only the time delays of chips, which are around
several nano seconds compared to the current 140 µs. Initial tests on using RAM
based receiving beamformer board have been conducted successfully.
134
(a) (b)
Figure 7.2 Block diagram of RAMs based control structure (a) and corresponding timing
sequences (b).
Therefore the frame rate of the new system relies only on the field of view. This
frame rate can be calculated as:
dN
c
FR
2
= (7.1)
where FR is the frame rate, c is the sound velocity in a medium, d is the image depth,
and N is the image line number of each frame. This frame rate can reach to more
than 400 frame/s in a wider field of view (>6.4 mm laterally and >20 mm axially).
C. 2-D Color Doppler
Using the duplex imaging system with higher frame rate, the current 1-D color
Doppler imaging system can be upgraded to a 2-D system easily. More attention
should be paid to the clutter filter design. Because the PRF of the high frame rate
system is high and the measured Doppler shift frequencies from slow flows is low,
the frequency response of a clutter filter must be very sharp as discussed in Chapter 6.
Therefore, high-order FIR or IIR filters need to be designed carefully instead of the
single-echo canceller filter. In addition, recent studies have shown that inter-frame
Trig
RAMs
Image Line
Counter
Destination
Circuits
WR
RD
Micro-
Controller
Trig
Counter
RAMs
2 1 0
D
2
D
1
D
0
135
clutter filters can be also used for decreasing a system’s minimal detectable velocity
at high frame rate (i.e. high PRF). These studies show that the minimal detectable
velocity of fluids can be as low as 0.5 mm/s at 20 frame/s, using the inter-frame
cluttering filter. Although these inter-frame filters were currently only utilized on
mechanical scanners, they may also be applicable for array-based imaging systems
(Needles et al. 2006). Nevertheless, all filters mentioned above must be supported by
appropriate signal processing hardwares in order to accomplish filtering in real time.
D. In vivo Microcirculation Studies Enhanced by Contrast Agents
Although the Doppler systems discussed in Chapter 4 and 5 were able to detect
the velocity as low as 0.1 mm/s from a wire phantom, it may be problematic for the
system to detect those low velocities from blood in capillaries, considering the
backscattering of blood is much lower than that of the wire phantom (Shung 2005).
Therefore, improving the SNR of the system would be critical both for B-mode
imaging and blood flow measuring. Carefully designed matched filters for RF
signals, determined by both bandwidths of transducers and transmitted pulses, can
optimize the SNR of the system (Jensen 1996). Increasing transmit voltage is another
way to increase SNR, but it is not always feasible due to the transducer heating
issues.
Strong Doppler signals were detected in the micro flow phantom experiments,
due to the existence of air-bubbles in fluid. These strong Doppler signals were the
undesired signals in our experiments, but they may be beneficial for improving SNR
of images. Contrast agent studies (de Jong et al. 1992; Burns et al. 1994; Shi and
136
Forsberg 2000) have shown that contrast agents containing micro air-bubbles are an
efficient way to improve SNR of ultrasound images. HF pulsed-wave Doppler
enhanced by contrast agents has shown its advantages in recent studies (Goertz et al.
2005), as well as the contrast agent enhanced color Doppler imaging (Foster et al.,
2000).
As a result, the contrast agent enhanced HF ultrasound imaging with arrays will
be conducted in future microcirculation studies, such as visualizing and quantifying
microcirculations in tumor tissues.
E. Digital Beamforming and Doppler Processing
The system with arrays reported in the dissertation was based on the analog
beamformer and the analog Doppler processing circuits which are a low-cost and
high-performance option. Digital beamformers using FPGAs and Doppler processors
using DSPs dominate in conventional clinical ultrasound systems, due to the high
performances and strong reprogrammabilities of FPGAs and DSPs at affordable
prices. It is anticipated that digital beamformers and their post-processing units with
latest DSPs or FPGAs will achieve better performance than the current analog
system (Hu et al. 2006; Brown and Lockwood 2005).
Furthermore, the color Doppler imaging using the autocorrelation algorithm is
not real time in our studies. All the images shown in Chapter 6 were obtained after
post-processing the acquired data using Matlab based software. Although current
PCs are capable of processing a large amount of data in a short time, they may still
not be able to meet the requirements of real-time color Doppler imaging. For
137
example, 3 MHz color Doppler system requires two million calculations per second
(Evans and McDicken 2000). Hence, DSPs are still being used for real-time color
Doppler processing in most commercial ultrasound systems. More calculations per
second are definitely expected in the system with 30 MHz arrays. Therefore DSPs or
FPGAs are required in order to implement real-time color Doppler imaging.
In summary, the development of high-frequency digital beamformers and
Doppler processors may become more cost-effective as the cost-performance ratios
of DSPs and FPGAs continue to decrease.
Conclusion
This dissertation has presented the investigation of the high-frequency ultrasound
systems for biomedical application using the 30~35 MHz linear arrays and the
analog beamformer.
The imaging system with the 48 element linear array, which can be used for
imaging the mouse heart, is capable of displaying 30 images per second and
acquiring over 100 images per second. The high-frequency pulsed-wave Doppler
systems with both the PMN-PT needle transducers and 30 MHz array can detect and
measure the blood velocities in vessels with diameters less than 200 µm, providing
valuable information for evaluating microcirculation. Initial color Doppler and
power Doppler studies with the HF array were used to visualize the mouse
cardiovascular blood flows. Both the in vitro and in vivo results described in this
dissertation have demonstrated that the high-frequency array-based ultrasound
138
imaging system with Doppler features has a promising future in the research fields of
small animal studies, ophthalmology and dermatology.
139
BIBLIOGRAPHY
Agilent Technologies. 6N135/6 Single Channel, High Speed Optocouplers Technical
Data. Agilent Technologies; 2004.
Analog Device. AD8332: Ultralow noise VGAs with preamplifier and
programmable Rin. Analog Device; 2003(a).
Analog Devices. ADuM1100: iCoupler digital isolator. Analog Device; 2003(b).
Arciniegas A. Treatment of the occlusion of the central retinal vein by section of the
posterior ring. Ann Ophthalmol 1984; 16: 1081-1086.
Aydin N, Fan L and Evans DH. Quadrature-to-directional format conversion of
Doppler signals using digital methods. Physiol Meas 1994; 15: 181-199.
Baxter GM and Williamson TH. Color Doppler flow imaging in central retinal vein
occlusion: a new diagnostic technique?. Radiology 1993; 187: 847-850.
Brown JA, Lockwood GR. A low-cost, high-performance pulse generator for
ultrasound imaging. IEEE Trans Ultrason Ferroelec Freq Contr 2002; 49: 848-
851.
Brown JA and Lockwood GR. A digital beamformer for high-frequency annular
arrays. IEEE Trans Ultrason Ferroelec Freq Contr 2005; 52: 1262-1269.
Brunner E. Ultrasound system considerations and their Impact on front-end
components. Analog Devices, Norwood, MA, 2002.
Burns PN, Powers JE, Simpson DH, Brezina A, Kolin A, Chin CT, Uhlendorf V and
Fritzsch T. Harmonic power mode Doppler using contrast agents: An improved
method for small vessel flow imaging. J Ech Med Ultra 1994; 16: 132-142.
Cannata JM, Ritter TA, Chen WH, Silverman RH and Shung KK. Design of efficient,
broadband single-element (20-80 MHz) ultrasonic transducers for medical
imaging applications. IEEE Trans Ultrason Ferroelec Freq Contr 2003; 50: 1548-
1557.
Cannata JM. High frequency (>20 MHz) ultrasonic arrays for medical imaging
applications. Ph.D. dissertation, Pennsylvania State University, University Park,
PA, 2004.
140
Cannata JM, Williams JA, Zhou Q, Ritter TA and Shung KK. Development of a 35-
MHz piezo-composite ultrasound array for medical imaging. IEEE Trans
Ultrason Ferroelec Freq Contr 2006; 53: 224-236.
Chen J and Panda R. Review: Commercialization of piezoelectric single crystals for
medical imaging applications. IEEE Ultrason Symp Proc 2005; 235-240.
Cherin E, Williams R, Needles A, Liu G, White C, Brown AS, Zhou YQ and Foster
FS. Ultrahigh frame rate retrospective ultrasound microimaging and blood flow
visualization in mice in vivo. Ultrasound Med Biol 2006; 32: 683-691.
Christopher DA, Burns PN, Armstrong J, Foster FS. A high-frequency continuous-
wave Doppler ultrasound system for the detection of blood flow in the
microcirculation. Ultrasound Med Biol 1996; 22:1191-1203.
Christopher DA, Burns PN, Starkoski BG, Foster FS. A high-frequency pulsed-wave
Doppler ultrasound system for the detection and imaging of blood flow in the
microcirculation. Ultrasound Med Biol 1997; 23:997-1015.
Cloutier G, Chen D, and Durand LG. A new clutter rejection algorithm for Doppler
ultrasound, IEEE Trans Medical Imaging 2003; 22: 530-538.
de Jong N, Ho L, Skotland T and Bom N. Absorption and scatter of encapsulated
microspheres: Theoretical considerations, measurements. Ultrasonics 1992; 30:
90-103.
Evans DH and McDicken WN. Doppler Ultrasound, John Wiley & Sons, LTD,2000.
Feke GT, Tagawa H, Deupree DM, Goger DG, Sebag J, Weiter JJ. Blood flow in the
normal human retina. Investigative Ophthalmology & Visual Science 1989; 30:
58-65.
Ferrara KW, Zager BG, Sokil-Melgar JB, Silverman RH and Aslanidis IM.
Estimation of blood velocity with high frequency ultrasound. IEEE Trans
Ultrason Ferroelec Freq Contr 1996; 43: 149-159.
Foster FS, Pavlin CJ, Lockwood GR, Ryan LK, Harasiewicz KA, Berube L, Rauth
AM. Principles and application of ultrasound backscatter microscopy. IEEE
Trans Ultrason Ferroelec Freq Contr 1993; 40: 608-617.
Foster FS, Burns PN, Simpson1 DH, Wilson SR, Christopher DA and Goertz DE.
Ultrasound for the visualization and quantification of tumor microcirculation.
Cancer and Metastasis Reviews 2000; 19: 131-138.
141
Foster FS, Zhang MY, Zhou YQ, Liu G, Mehi J, Cherin E, Harasiewicz KA,
Starkoski BG, Zan L, Knapik DA and Adamson SL. A new ultrasound
instrument for in vivo microimaging of mice. Ultrasound Med Biol
2002;28:1165-72.
Gage Applied Technologies. CompuScope 14200: General-purpose digitizer for the
PCI bus. Gage Applied Technologies, 2004.
Gan LM, Gronros J, Hagg U, Wikstrom J, Theodoropoulos C, Friberg P and
Fritsche-Danielson R. Non-invasive real-time imaging of atherosclerosis in mice
using ultrasound biomicroscopy. Atherosclerosis 2007; 190:313-320.
Goertz DE, Christopher DA, Yu JL, Kerbel RS, Burns PN and Foster FS. High
frequency color flow imaging of the microcirculation. Ultrasound Med Biol 2000;
26: 63-71.
Goertz DE, Yu JL, Kerbel RS, Burns PN and Foster FS. High-frequency Doppler
ultrasound monitors the effects of antivascular therapy on tumor blood flow.
Cancer Res 2002; 62: 6371–6375.
Goertz DE, Needles A, Burns PN and Foster FS.High-frequency, nonlinear flow
imaging of microbubble contrast agents. IEEE Trans Ultrason Ferroelec Freq
Contr 2005; 52: 495- 502.
Gottlieb E, Lai B, Xu X, Shung KK. PMN-PT High frequency ultrasonic needle
transducers for pulsed wave Doppler measurements in the eye. IEEE Ultrason
Symp Proc 2005;4: 2227- 2230.
Green WR, Chan CC, Hutchins GM, Terry JM. Central retinal vein occlusion: a
prospective histopathologic study of 29 eyes in 28 cases. Retina 1981; 1: 27-55.
Gronros J, Wikstrom J, Hägg U, Wandt B, and Gan L. Proximal to middle left
coronary artery flow velocity ratio, as assessed using color Doppler
echocardiography, predicts coronary artery atherosclerosis in mice.
Arteriosclerosis, Thrombosis, and Vascular Biology 2006; 26:1126-1131.
Hall TJ, Bilgen M, Insana MF,Krouskop TA.Phantom Materials for Elastography.
IEEE Trans Ultrason Ferroelec Freq Contr 1997; 44: 1355-1365.
Hartley CJ, Cole JS. An ultrasonic pulsed Doppler system for measuring blood flow
in small vessels. J Appl Physiol 1974; 37: 626-629.
Hedrich HJ, Bullock G. The Laboratory Mouse (Handbook of Experimental
Animals). San Diego: Elsevier Academy Press, 2004.
142
Horowitz P, Hill W. The Art of Electronics. New York: Cambridge University Press,
1989.
Hozumi N, Yamashita R, Lee CK, Nagao M, Kobayashi K, Saijo Y, Tanaka M,
Tanaka N and Ohtsuki S. Time-frequency analysis for pulse driven ultrasonic
microscopy for biological tissue characterization. Ultrasonics 2004; 42: 717-722.
Hu CH, Xu XC, Cannata JM, Yen JT and Shung KK. Development of a real-time,
high-frequency ultrasound digital beamformer for high-frequency linear array
transducers. IEEE Trans Ultrason Ferroelec Freq Contr 2006; 53: 317-323.
Huang B and Shung KK. Characterization of high-frequency single-element focused
transducers with wire target and hydrophone. IEEE Trans Ultrason Ferroelec
Freq Contr 2005; 52: 1608- 1612.
Hwang AB, Iwata K, Sakdinawat AE, Wu MC, Hasegawa BH. Gantry specifications
for a dual modality imaging system for small animals. Proceeding of IEEE
Nuclear Science Symposium 2002; 2:1303-1307.
Interisl. Datasheet: EL7158 ultra-high current pin driver. Interisl Americas, 2003.
Ip MS, Gottlieb JL, Kahana A, Scott IU, Altaweel MM, Blodi BA, Gangnon RE,
Puliafito CA. Intravitreal triamcinolone for the treatment of macular edema
associated with central retinal vein occlusion. Arch Ophthalmol 2004;122:1131-
1136.
Jensen JA, Estimation of Blood Velocities Using Ultrasound. Cambridge University
Press, New York, 1996.
Kasai C,Koroku N,Koyano A and Omoto R. Real-time two-dimensional blood flow
Imaging using an autocorrelation technique. IEEE Trans on Sonics Ultrason 1985;
32:458-464.
Kester W and Bryant JS. High speed design techniques: Grounding in high speed
system. Technical References, Analog device, 1996.
Klein R,Klein BE, Moss SE, and Meue SM. The epidemiology of retinal vein
occlusion: the beaver dam eye study. Trans Am Ophthalmol Soc 2000; 98: 133-
143.
Larsson J, Carlson J, Olsson SB. Ultrasound enhanced thrombolysis in experimental
retinal vein occlusion in the rabbit. Br J Ophthalmol 1998; 82:1438-1440.
143
Li Y, Garson CD, French BA and Hossack JA. High resolution quantification of
myocardial motion in mice using 2d speckle tracking. IEEE Ultrason Symp Proc
2005; 1: 369- 372.
Liu JH, Jeng GS, Wu TK and Li PC. ECG triggering and gating for ultrasonic small
animal imaging. IEEE Trans Ultrason Ferroelec Freq Contr 2006;53: 1590-1596.
Lockwood GR, Hunt JW, and Foster FS. The design of protection circuitry for high-
frequency ultrasound imaging systems. IEEE Trans Ultrason Ferroelec Freq
Contr 1991; 38:48-55.
Lockwood GR, Turnbull DH, Christopher DA, Foster FS. Beyond 30 MHz:
Applications of high-frequency ultrasound imaging. IEEE Engineering in
Medicine and Biology Magazine 1996; 15: 60-71.
Needles A, Cheung AM, Foster FS, Goertz DE. High frequency colour flow imaging
of mouser tumors using inter-frame clutter filtering. IEEE Ultrason Symp Proc,
Vancouver, Canada, 2006 (in press).
Opremcak EM, Rehmar AJ, Ridenour CD, et al. Radial optic neurotomy with
adjunctive intraocular triamcinolone for central retinal vein occlusion: 63
consecutive cases. Retina 2006; 26: 306-313.
Ott HW. Noise reduction techniques in electronic systems. 2nd edition, Wiley
Publisher, New York , 1988.
Pavlin CJ, Christopher DA, Burns PN, Foster FS. High-frequency Doppler
ultrasound examination of blood flow in the anterior segment of the eye. Am J
Ophthalmol 1998; 126:597-600.
Perrino C, Prasad SV, Mao L, Noma T, Yan Z, Kim HS, Smithies O and Rockman
HA. Intermittent pressure overload triggers hypertrophy-independent cardiac
dysfunction and vascular rarefaction. J Clin Invest 2006;116: 1547-1560.
Reddy AK, Jones AD, Martono C, Caro WA, Madala S and Hartley CJ. Pulsed
Doppler signal processing for use in mice: design and evaluation. IEEE Trans
Biomed Eng 2005a; 52: 1764-1770.
Reddy AK, Taffet GE, Li YH, Lim SW, Pham TT, Pocius JS, Entman ML, Michael
LH and Hartley CJ. Pulsed Doppler signal processing for use in mice:
applications. IEEE Trans Biomed Eng 2005b; 52: 1771-1783.
144
Ritter TA, Geng X, Shung KK, Lopath P, Park SE, and Shrout TR. Single crystal
PZN/PT-Polymer composites for ultrasound transducer applications. IEEE Trans
Ultrason Ferroelec Freq Contr 2000; 47: 792-800.
Ritter TA, Shrout TR, Tutwiler R, Shung KK. A 30 MHz piezo-composite
ultrasound array for medical imaging applications. IEEE Trans Ultrason
Ferroelec Freq Contr 2002; 49: 217-230.
Sergott RC. Optic nerve sheath decompression: history, techniques, and indications.
Int Ophthalmol Clin 1991; 31: 71-81.
Shamdasani V, Managuli R, Sikdar S, Kim YM. Ultrasound color-flow imaging on a
programmable system. IEEE Trans. Information Technology in Biomedicine
2004; 8: 191- 199.
Shi WT and Forsberg F. Ultrasonic characterization of the nonlinear properties of
contrast microbubbles. Ultrasound Med Biol 2000; 26: 93-104.
Shung KK, Smith MB, and Tsue B. Principles of Medical Imaging. San Diego:
Academic Press, 1992.
Shung KK and Zippuro M. Ultrasonic transducers and arrays. Engineering in
Medicine and Biology Magazine IEEE 1996; 15: 20-30.
Shung KK, Diagnostic Ultrasound: Imaging and Blood Flow Measurements. Boca
Raton: CRC Press, 2005.
Silverman RH, Kruse DE, Coleman DJ and Ferrara KW. High-resolution ultrasonic
imaging of blood flow in the anterior segment of the eye. Investigative
Ophthalmology & Visual Science, 1999; 40: 1373-1381.
Stitt JP, Tutwiler RL, Shung KK. An improved four-focal zone high-frequency
ultrasound analog beamformer. IEEE Ultrason Symp Proc 2002; 617-620.
Sun L, Feng C, Cannata JM, Johnson JA, Yen JT and Shung KK. High frame rate
high frequency ultrasound system for small animal cardiac imaging. IEEE
Ultrason Symp Proc 2006a (in press).
Sun L, Sangkatumvong S, and Shung KK. A high resolution digital ultrasound
system for imaging of zebrafish. IEEE Ultrason Symp Proc 2006b (in press).
Supertex. TC6320 N- and P-channel enhancement-mode MOSFET pair. Supertex
Inc.,2003.
145
Supertex. Initial release: High speed ±100V 2A integrated ultrasound pulser.
Supertex Inc., 2006.
Tameesh MK, Lakhanpal RR, Fujii GY, Javaheri M, Shelley T, D'anna S, Barnes
AC, Margalit E, Farah M, Juan ED, Humayun MS. Retinal vein cannulation with
prolonged infusion of tissue plasminogen activator (t-PA) for the treatment of
experimental retinal vein occlusion in dogs. American Journal of
Ophthalmology 2004; 138: 829-839.
Thomenius KE. Evolution of ultrasound beamformers. IEEE Ultrason Symp Proc
1996; 1615-1622.
Tysoe C and Evans DH. Bias in mean frequency estimation of Doppler signals due to
wall clutter filters. Ultrasound Med Biol 1995; 21: 671-677.
Visualsonics: www.visualsonics.com, 2006.
Wang G, Liu YN, Ye YB, Zhao SY, Hsieh J, Ge SP. Top-level design and
preliminary physical analysis for the first electron-beam Micro-CT scanner,
Journal of X-Ray Science and Technology 2004; 12:251-260.
Weiss JN, Bynoe LA. Injection of tissue plasminogen activator into a branch retinal
vein in eyes with central retinal vein occlusion. Ophthalmology 2001; 108: 2249-
2257.
Williamson TH, Poon W, Whitefield L, Strothoudis N and Jaycock P. A pilot study
of pars plana vitrectomy, intraocular gas, and radial neurotomy in ischaemic
central retinal vein occlusion. British Journal of Ophthalmology 2003; 87: 1126-
1129.
Williamson TH. Central retinal vein occlusion: what’s the story? Br. J. Ophthalmol.
1997; 81: 698-704.
Wong TY, Klein R, Sharrett AR, Duncan BB, Couper DJ, Klein BEK. Retinal
arteriolar diameters and risk of hypertension. Ann Intern Med 2004; 140: 248-55.
Xu XC, Hu CH, Sun L, Yen JT and Shung KK. High-frequency high frame rate
ultrasound imaging system for small animal imaging with linear arrays. IEEE
Ultrason Symp Proc; 1: 1431- 1434.
Xu X, Yen JT and Shung KK. A low-cost bipolar pulse generator for high frequency
ultrasound applications. IEEE Trans Ultrason Ferroelec Freq Contr 2006 (in
press).
146
Yang XP, Liu YH, Rhaleb NE, Kurihara N, Kim HE, Carretero OA.
Echocardiographic assessment of cardiac function in conscious and anesthetized
mice. Am J Physiol 1999; 277:1967-1974.
Yang XP. High frequency ultrasound imaging processing system design. Master
thesis, Penn State University, State College, USA 2002.
Abstract (if available)
Abstract
Small animal research is gaining more attention since mice and rats are preferred animal models for gene and drug therapy. High-frequency (HF) ultrasound imaging, capable of achieving superior spatial resolution at an affordable price, has been shown to be useful for imaging and visualizing blood flow in small animals for biological and pharmaceutical research. The utilization of HF arrays in a HF imaging system can alleviate the limitations of the current systems with single element transducers, such as the uneven image quality and frame rate.
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
High-frequency ultrasound array-based imaging system for biomedical applications
PDF
Array transducers for high frequency ultrasound imaging
PDF
A high frequency array- based photoacoustic microscopy imaging system
PDF
Development of high-frequency (~100mhz) PZT thick-film ultrasound transducers and arrays
PDF
High frequency ultrasonic phased array system and its applications
PDF
High frequency ultrasound array for ultrasound-guided breast biopsy
PDF
Novel beamforming with dual-layer array transducers for 3-D ultrasound imaging
PDF
Development of back-end processing system for high frequency ultrasound b-mode imaging
PDF
Single-cell analysis with high frequency ultrasound
PDF
High frequency ultrasonic imaging for the development of breast biopsy needle with miniature ultrasonic transducer array
PDF
Fresnel beamforming for low-cost, portable ultrasound systems
PDF
Development of front-end circuits for high frequency ultrasound system
PDF
High-frequency ultrasonic transducers for photoacoustic applications
PDF
Intravascular imaging on high-frequency ultrasound combined with optical modalities
PDF
High frequency ultrasound elastography and its biomedical applications
PDF
Highly integrated 2D ultrasonic arrays and electronics for modular large apertures
PDF
Microfluidic cell sorting with a high frequency ultrasound beam
Asset Metadata
Creator
Xu, Xiaochen
(author)
Core Title
High-frequency ultrasound imaging system with Doppler features for biomedical applications using 30~35 mHz linear arrays and an analog beamformer
School
Viterbi School of Engineering
Degree
Doctor of Philosophy
Degree Program
Biomedical Engineering
Publication Date
02/14/2007
Defense Date
11/16/2006
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
high-frequency beamformer,high-frequency Doppler,high-frequency linear arrays,high-frequency ultrasound,OAI-PMH Harvest,retinal vein occlusion
Language
English
Advisor
Shung, K. Kirk (
committee chair
), Cannata, Jonathan Matthew (
committee member
), Leahy, Richard M. (
committee member
), Yen, Jesse T. (
committee member
)
Creator Email
xiaochex@usc.edu
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-m248
Unique identifier
UC1156483
Identifier
etd-Xu-20070214 (filename),usctheses-m40 (legacy collection record id),usctheses-c127-163727 (legacy record id),usctheses-m248 (legacy record id)
Legacy Identifier
etd-Xu-20070214.pdf
Dmrecord
163727
Document Type
Dissertation
Rights
Xu, Xiaochen
Type
texts
Source
University of Southern California
(contributing entity),
University of Southern California Dissertations and Theses
(collection)
Repository Name
Libraries, University of Southern California
Repository Location
Los Angeles, California
Repository Email
cisadmin@lib.usc.edu
Tags
high-frequency beamformer
high-frequency Doppler
high-frequency linear arrays
high-frequency ultrasound
retinal vein occlusion