University of Southern California Dissertations and Theses

Lesion enhancement for three dimensional rectilinear ultrasound imaging

(USC Thesis Other)

Lesion enhancement for three dimensional rectilinear ultrasound imaging

PDF

Download

Open document

Flip pages

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content LESION ENHANCEMENT FOR THREE DIMENSIONAL RECTILINEAR ULTRASOUND IMAGING by Samer Awad 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 2009 Copyright 2009 Samer Awad ii DEDICATION With genuine thanks to my beloved parents, Izzat Awad and Sabah Awad; my precious sisters, Rana and Rola and their families for their endless love and unflagging prayer and support iii ACKNOWLEDGEMENTS During my graduate years at the University of Southern California, I have been blessed to be accompanied and supported by many people. Their assistance and perceptive comments have considerably helped me improve this work. I would like to thank them all. This list is ordered, if at all, only by a rough combination of significance and alphabetical last name: Advisor Jesse Yen (Assistant Professor of Biomedical Engineering, USC) Thesis Committee Jesse Yen (Assistant Professor of Biomedical Engineering, USC) K. Kirk Shung (Professor of Biomedical Engineering, USC) Todd Brun (Associate Professor of Electrical Engineering, USC) Jonathan Cannata (Research Assistant Professor of Biomedical Engineering, USC) Brent Liu (Assistant Professor of Radiology & Biomedical Engineering, USC) Qualifying Committee Jesse Yen (Assistant Professor of Biomedical Engineering, USC) K. Kirk Shung (Professor of Biomedical Engineering, USC) Jonathan Cannata (Research Assistant Professor of Biomedical Engineering, USC) Linda Hovanessian- Larsen (Associate Professor of Radiology) Brent Liu (Assistant Professor of Radiology & Biomedical Engineering, USC) Colleagues Jay Mung (Ph.D Student of Biomedical Engineering, USC) Chi Seo (Postdoctoral Research Associate of Biomedical Engineering, University of Washington) iv In addition to the specific research contributions acknowledged above, special thanks goes to all my family and friends. They have made the years both more exciting and pleasurable. Each, in their own way, has added to the personal importance of this work to me. Most especially thanks to my mom and dad for unconditionally always being there for me; to my lovely two sisters Rola and Rana who have been always supportive; to my best friend Yousef and college friends Hazeem and Nael for these years of caring and all we have shared; to cousins Waseem and Muna and the Sunday dinners that made me feel home away from home; to my friend Abdullah for helping me cope with this new life and for the good times we spent together; to my friends Pooja, Tejas, Sasa and Reena and the great vacations we spent together. From the bottom of my heart I thank you all. v TABLE OF CONTENTS Dedication ........................................................................................................................... ii Acknowledgements ............................................................................................................. iii List of Tables ....................................................................................................................... vii List of Figures ..................................................................................................................... viii Abstract ............................................................................................................................... xiv Chapter 1: Introduction ....................................................................................................... 1.1 Background ............................................................................................................ 1.2 Hypothesis .............................................................................................................. 1.3 Overview ................................................................................................................ 1 1 4 6 Chapter 2: Ultrasound imaging ........................................................................................... 2.1 Ultrasound Physics ................................................................................................ 2.2 Transducer Fundamentals ...................................................................................... 2.3 Transducer Arrays ................................................................................................ 2.4 Beamforming .......................................................................................................... 2.5 Beam Patterns ......................................................................................................... 2.6 Spatial Resolution ................................................................................................ 9 9 10 11 13 15 18 Chapter 3: 3-D Ultrasound Strain Imaging ......................................................................... 3.1 Introduction ............................................................................................................ 3.2 Strain Imaging Principles ....................................................................................... 3.2.1 Time Delay Estimation ................................................................................ 3.2.2 3-D vs. 2-D Strain Imaging .......................................................................... 3.3 Methods .................................................................................................................. 3.3.1 System Description ...................................................................................... 3.3.2 Phantom Fabrication .................................................................................... 3.3.3 Data Collection and Signal Processing ........................................................ 3.3.4 Performance Evaluation ............................................................................... 3.4 Results .................................................................................................................... 3.5 Discussion .............................................................................................................. 20 20 22 24 27 29 29 31 31 34 35 43 vi Chapter 4: A Dual-Layer Transducer Array for 3-D Rectilinear Imaging .......................... 4.1 Chapter Overview ................................................................................................ 4.2 Introduction ............................................................................................................ 4.3 Methods .................................................................................................................. 4.3.1 3-D Rectilinear Scanning ............................................................................. 4.3.2 Data Acquisition .......................................................................................... 4.3.3 Beamforming, Signal Processing, and Display ............................................ 4.4 Results .................................................................................................................... 4.5 Discussion .............................................................................................................. 45 45 46 50 50 53 54 56 59 Chapter 5: 3-D Spatial Compounding Using a Row-Column Array ................................ 5.1 Chapter Overview ................................................................................................ 5.2 Strain Imaging Using the Row-column 2-D Array ................................................ 5.3 Introduction ............................................................................................................ 5.4 Methods .................................................................................................................. 5.4.1 Simulations ................................................................................................ 5.4.2 System Description ...................................................................................... 5.4.3 Phantom fabrication ..................................................................................... 5.5 Results .................................................................................................................... 5.5.1 Simulations ................................................................................................ 5.5.2 Experiments ................................................................................................ 5.6 Discussion .............................................................................................................. 61 61 62 63 66 66 70 75 76 76 80 88 Chapter 6: Summary and Future Work ............................................................................... 6.1 Summary ................................................................................................................ 6.2 Previous Research Objectives ................................................................................ 6.3 3-D Spatial Compounding For 3-D Strain Imaging ............................................... 6.4 3-D Companding ................................................................................................ 6.5 In vitro Experiments ............................................................................................... 6.5.1 In vitro Murine Breast Tumor Strain Imaging ............................................. 6.5.2 In vitro Ultrasound Imaging of Excised Tissue Specimens ......................... 91 91 93 97 99 102 102 102 Bibliography ........................................................................................................................ 103 Appendix A: Interfacing The Transducer Arrays With The Ultrasound System ................ 110 Appendix B: C++ Code For Data Acquisition With Ultrasonix ......................................... 113 vii LIST OF TABLES Table 1.1: The roles of 3-D strain imaging, 3-D B-mode imaging using the dual- layer array and 3-D spatial compounding in improving lesion detectability in 3-D rectilinear ultrasound images .......................................... 5 Table 4.1: Recipes for the background and inclusion materials in the gel phantoms used in the experiments ................................................................ 56 Table 5.1: Simulations parameters ................................................................................... 67 Table 5.2: Recipes for the background and inclusion materials in the gel phantoms used in the experiments ................................................................ 75 Table 5.3: Mean cross-correlation, mean speckle SNR values and the theoretical and experimental speckle SNR improvement values for the cyst and tumor 3-D images ........................................................................................... 82 Table 6.1: A comparison between the three transducer arrays used in this research ........................................................................................................... 96 Table A.1: A list of the parts used to interface the dual-layer and the row-column transducer arrays to the ultrasound system ..................................................... 110 Table B.1: Imaging parameters used in the sample C++ code ................................ 113 viii LIST OF FIGURES Figure 2.1: In 1-D ultrasonic imaging, echo signals correspond to tissue boundaries. Boundary depths are linearly related to echo arrival time. Calculating the envelope of echoes RF data yields the A-mode image .......................................................................................................................... 11 Figure 2.2: Ultrasonic arrays and their fields of view (a) 1-D linear array. (b) 2-D rectilinear array ............................................................................................... 12 Figure 2.3: Beamforming in transducer arrays. Because the distances from the focal point are different, the reflected echoes from focal point #1 arrive at different times. If certain time delays where applied to signals according, all the echoes will line up and can be summed together to form an image line ................................................................ 14 Figure 2.4: The near field and the far field of a circular transducer. In the near field, the magnitude of pressure is oscillating. In the far field, the pressure is more uniform and decreases approximately according to 1/Z (Christensen 1988) ................................................................................... 16 Figure 2.5: The far field angular beam profile of a typical transducer array .................... 17 Figure 2.6: Spatial resolution: (a) The echoes from two targets can be easily resolved with full beam width separation between them (b) Less separation between the targets makes them more difficult to resolve (c) If the separation is much less than the beam width, the two echoes become indistinguishable ................................................................................ 18 Figure 3.1: Measurement of strain in a one dimensional spring system of different spring constant by applying compression, calculating displacements and differentiating to find strain values ................................ 23 Figure 3.2: First step in stain imaging is to acquire a pre-compression image (D1) and a post-compression image (D2). A window W1 is selected from D1 and compared to several windows W2 selected from D2 within a search region S to calculate displacements ..................................................... 24 ix Figure 3.3: 1-D TDE using SAD: SAD is applied to the signal S1(x) and parts of S2(x) of equal size. The minimum value of the SAD vector corresponds to the best match and its location corresponds to the time delay ................................................................................................................ 26 Figure 3.4: Axial compression and lateral expansion in strain imaging ........................... 28 Figure 3.5: Sparse 2-D array. (a) Schematic of the entire array. (b) Schematic of a single cell with 0.6 × 0.6 mm receive elements and surrounding transmitters ................................................................................................ 29 Figure 3.6: A diagram that shows the interconnection between the receive elements and the Ultrasonix system ................................................................ 30 Figure 3.7: Collage of slices in azimuthal, elevational and axial slices of the 3-D axial strain image of the 12 mm, 9 mm and 6 mm (left to right) cylindrical inclusion phantoms generated using 3-D beamforming and 3-D TDE. Top row: azimuthal slices. The arrows indicate where the elevational and axial slices were taken. Middle row: elevational slices. The arrows indicate where the azimuthal slices were taken. Bottom row: Axial slices. The size of the images is 27 mm axially and 30 mm in azimuth and in elevation. The transducer is located on top of the azimuthal and elevational slices and parallel to the axial slices ................................................................................................................ 36 Figure 3.8: Perpendicular slices of the 9 mm cylindrical inclusion phantom 3-D strain image ................................................................................................ 37 Figure 3.9: Collage of slices in azimuthal, elevational and axial slices of the 3-D axial strain image of the 12 mm, 9 mm and 6 mm (left to right) spherical inclusion phantoms generated using 3-D beamforming and 3-D TDE. Top row: azimuthal slices. The arrows indicate where the elevational and axial slices were taken. Middle row: elevational slices. The arrows indicate where the azimuthal slices were taken. Bottom row: Axial slices. The size of the images is 27 mm axially and 30 mm in azimuth and in elevation. The transducer is located on top of the azimuthal and elevational slices and parallel to the axial slices ................................................................................................................ 38 Figure 3.10: Perpendicular slices of the 9 mm spherical inclusion phantom 3-D strain image ................................................................................................ 39 x Figure 3.11: A bar graph of the CNR values of the axial strain images of the 6 mm, 9 mm and 12 mm cylindrical and spherical inclusion phantoms. The strain images were generated by applying: 2-D beamforming followed by 2-D TDE (2-D BF, 2-D TDE), 3-D beamforming followed by 2-D strain imaging (3-D BF, 2-D TDE), 2-D beamforming followed by 3-D strain imaging (2-D BF, 3-D TDE) and 3-D beamforming followed by 3-D strain imaging (3-D BF, 3-D TDE) ............................................................................................................... 40 Figure 3.12: Collage of the same elevational slice of the axial strain images of the 9 mm spherical inclusion phantom generated using the four signal processing combinations: 2-D BF, 2-D TDE, 3-D BF, 2-D TDE, 2-D BF, 3-D TDE and 3-D BF, 3-D TDE. The white rectangles inside the inclusion and on the top left corner of the images are the cross section of the 3-D kernels used to calculate CNR. The size of the images is 23.7 mm axially and 26.1 mm in azimuth ...................................................... 41 Figure 3.13: A bar graph of the SNR values of the axial strain images of the 6 mm, 9 mm and 12 mm cylindrical and spherical inclusion phantoms. The strain images were generated by applying: 2-D beamforming followed by 2-D TDE (2-D BF, 2-D TDE), 3-D beamforming followed by 2-D strain imaging (3-D BF, 2-D TDE), 2-D beamforming followed by 3-D strain imaging (2-D BF, 3-D TDE) and 3-D beamforming followed by 3-D strain imaging (3-D BF, 3-D TDE) ............................................................................................................... 42 Figure 3.14: A bar graph of the strain contrast values of the axial strain images of the 6 mm, 9 mm and 12 mm cylindrical and spherical inclusion phantoms. The strain images were generated by applying: 2-D beamforming followed by 2-D TDE (2-D BF, 2-D TDE), 3-D beamforming followed by 2-D strain imaging (3-D BF, 2-D TDE), 2- D beamforming followed by 3-D strain imaging (2-D BF, 3-D TDE) and 3-D beamforming followed by 3-D strain imaging (3-D BF, 3-D TDE) ............................................................................................................... 43 Figure 4.1: 3-D scanning process of the dual-layer transducer array in (a) transmit and (b) in receive. Shaded elements indicate the active subaperture ................................................................................................ 51 xi Figure 4.2: Simulated on-axis beamplots of the dual-layer transducer with focus (x,y,z) = (0,0,30) mm: a) 3-D beamplot, b) Contour plot with lines at -10, -20, -30, -40, and -50 dB, c) azimuthal beamplot, and d) elevational beamplot ....................................................................................... 52 Figure 4.3: Simulated off-axis beamplots of the dual-layer transducer with focus (x,y,z) = (15,15,30) mm: a) 3-D beamplot, b) Contour plot with lines at -10, -20, -30, -40, and -50 dB, c) azimuthal beamplot, and d) elevational beamplot ........................................................................................ 53 Figure 4.4: Experimental axial wire target images with short-axis in azimuth (A- C) and short axis in elevation (D-F). All images are log-compressed and shown with 20 dB dynamic range ............................................................ 57 Figure 4.5: Azimuthal and elevational lateral wire target responses ................................ 58 Figure 4.6: Experimental cyst images with the cyst short-axis in azimuth (A-C) and the cyst short axis in elevation (D-F). All images are log- compressed and shown with 30 dB dynamic range ................................ 59 Figure 5.1: A GCF map and a slice of a 3-D strain image of a 10 mm cylindrical inclusion phantom that were generated at the same location .......................... 62 Figure 5.2: Ultrasound images of Fibroadenoma in a 44-year-old woman using (a) conventional ultrasound and (b) spatial compounding (Kwak and You 2004) ................................................................................................ 64 Figure 5.3: Apertures used to create the first scan line that have (a) 0.5 and (b) 1.5 fractional aperture separation. The apertures were walked to create on-axis scan lines whenever it is possible. Both apertures are indicated by braces and shading ................................................................ 68 Figure 5.4: The three apertures used to create the first scan line for the speckle reduction simulation. The apertures were walked to create on-axis scan lines whenever it is possible. All apertures are indicated by braces and shading .......................................................................................... 69 xii Figure 5.5: Row-column addressing scheme. The transmit side consists of 8 horizontal electrodes. The receive side consists of 8 vertical electrodes which serve as ground in transmit. Transmit beamforming is done in the vertical or elevational direction while receive beamforming is done horizontally or azimuthally .......................................... 71 Figure 5.6: Transmit and Receive Beamforming for 3-D rectilinear imaging .................. 72 Figure 5.7: The three apertures used to create the first scan line for (a) the transmit side and (b) the receive side of the row-column 2-D array. The apertures were walked to create on-axis scan lines whenever it is possible. All apertures are indicated by braces and shading ........................... 74 Figure 5.8: B-mode images of (a) the -11 dB cyst phantom and (b) the +11 dB tumor phantom using Ultrasonix L14-5/38 linear array. Both images are log-compressed and shown with 30 dB dynamic range ............................ 76 Figure 5.9: Speckle decorrelation curve for 0 to 2 fractional aperture translation with step size = 0.25 using Field II simulation. The error bars indicate ± 2 standard deviations ................................................................................... 77 Figure 5.10: Images of one of the realizations of the -11 dB cyst phantom simulated in Field using: (a) Standard beamforming. (b) Spatial compounding of three images. Both images are log-compressed and shown with 30 dB dynamic range ................................................................ 78 Figure 5.11: Images of one of the realizations of the +11 dB tumor phantom simulated in Field using: (a) Standard beamforming. (b) Spatial compounding of three images. Both images are log-compressed and shown with 30 dB dynamic range ................................................................ 79 Figure 5.12: Mean of CNR values for 1 to 9 compounded volumes for: (a) -11 dB cyst phantom (b) +11 dB tumor phantom. Solid lines show experimental CNR values and dashed lines show theoretical CNR values .............................................................................................................. 81 Figure 5.13: Mean of speckle SNR values for 1 to 9 compounded volumes for: (a) -11 dB cyst phantom (b) +11 dB tumor phantom. Solid lines show experimental speckle SNR values and dashed lines show theoretical speckle SNR values.......................................................................................... 83 xiii Figure 5.14: Experimental cyst 3-D images with the -11 dB cyst short-axis in azimuth using: (A-C) Standard 3-D beamforming. (D-F) Compounding of nine volumes. All images are log-compressed and shown with 30 dB dynamic range ................................................................ 84 Figure 5.15: Isosurface renderings of the -11 dB cyst 3-D image generated using: (a) standard beamforming and (b) compounding of nine volumes. In each case, the isosurface level was adjusted for optimal display of the cyst .................................................................................................................. 85 Figure 5.16: Experimental tumor 3-D images with the +11 dB tumor short-axis in azimuth using: (A-C) standard 3-D beamforming and (D-F) compounding of nine volumes. All images are log-compressed and shown with 30 dB dynamic range ................................................................ 86 Figure 5.17: Isosurface rendering of the +11 dB tumor 3-D image using: (a) Standard 3-D beamforming. (b) Compounding of nine volumes. In each case, the isosurface level was adjusted for optimal display of the tumor ............................................................................................................... 87 Figure 5.18: A short axis slice of the 3-D wires phantom image generated using: (a) standard beamforming and (b) Compounding of nine volumes. Both images are log-compressed and shown with 20 dB dynamic range ................................................................................................................ 88 Figure 6.1: The planar 4.8 × 4.8 mm transmit beam used in the sparse 2-D array allows for a narrow range of steering angles for the focused receive beam ................................................................................................................ 94 Figure 6.2: A flow chart summarizing strain image generation using companding .......................................................................................................................... 101 Figure A.1: A diagram that illustrates the interface between the ultrasound machine and the transducer array ................................................................ 112 xiv ABSTRACT 3-D ultrasound has been found useful for diagnostic imaging. It has several advantages over 2-D ultrasound including viewing planes that are usually inaccessible using 2-D ultrasound, more accurate volume measurements and volume-rendered images. These capabilities can improve the detectability of cancerous tumors in soft tissue such as the breast and prostate. Additionally, real-time 3-D ultrasound can be utilized in guiding procedures such as breast needle biopsy. This research explores different methods for improving the detectability of near field lesions in 3-D rectilinear ultrasound. The first part of this research investigates ultrasound 3-D strain imaging, an imaging technique that detects the stiffness of tissue. I present a 3- D strain imaging system using a prototype 2-D array and a data acquisition system. Using several offline data processing techniques, I examined the benefits of 3-D strain imaging using 3-D ultrasound over 2-D strain imaging. 3-D strain images of gelatin/agar phantoms with cylindrical and spherical stiff inclusions of various diameters were acquired. 3-D strain imaging has shown a better performance in all case. Knowing that cancerous tumors are naturally stiffer than normal tissue, 3-D strain imaging could improve the detection of stiffer malignant tumors in breast. In the second part of this research I present testing and 3-D imaging results of a dual- layer array. The 4 × 4 cm array uses PZT and P[VDF-TrFE] layers for transmit and receive respectively. This dual-layer transducer was interfaced with a commercial ultrasound system to acquire data. I acquired 3-D data sets and applied offline delay-and- xv sum beamforming to generate image volumes. 3-D images of an 8 mm diameter anechoic cyst phantom and nylon wire targets are presented. The measured -6 dB beamwidths are 0.65 mm and 0.67 mm in azimuth and elevation respectively compared to the theoretical value of 0.52 mm. Speckle is an artifact that degrades contrast resolution in ultrasound images and thereby reduces the diagnostic value of the images. It is caused by the interference of energy from randomly distributed scatters, too closely spaced to be resolved by the imaging system. In the third part of this research I investigated 3-D spatial compounding, a technique used to reduce speckle noise. Using a row-column 2-D array, 3-D images of - 11 dB cyst, +11 dB tumor and nylon wire targets were acquired synthetically. Using aperture translation technique with offline beamforming, nine uncorrelated volumes were generated. Compounding the nine volumes resulted in an average increase in contrast of about 150%. Using a smaller aperture for generating the compounded images resulted in an expected loss in resolution. 1 CHAPTER 1 INTRODUCTION 1.1 Background Ultrasound imaging has been in widespread clinical use. It is a useful way of examining many of the body's internal organs, including the heart and blood vessels, eyes, breast and abdominal organs. In many parts of the body, suspected tumors are monitored with ultrasound. It is also used to guide procedures such as needle biopsy of breast cancer. Doppler ultrasound images can help the physician see and evaluate blockages and narrowing of blood vessels. Ultrasound not only complements the more traditional diagnostic imaging technologies such as magnetic resonance imaging (MRI), positron emission tomography (PET), and X-ray but also possesses unique and advantageous characteristics. With up to a 100 frames per second, ultrasound can capture the dynamic movement of organs and details of blood flow in real time (Szabo 2004). Its relativly low cost and nonionizing radiation makes it more attractive than other imaging modalities of similar capabilities. The portability of ultrasound machines makes it possible to use them outside of the radiology and cardiology departments. In recent years, researchers have developed 3-D ultrasound imaging technology. It offers several advantages over 2-D ultrasound including improved elevational focusing, viewing planes that are usually inaccessible using 2-D ultrasound, volume-rendered images, more accurate volume measurements and improved detection of cystic or cancerous masses (Fenster and Downey 1996). Visualization using 3-D ultrasound can 2 improve the detection of breast lesions, prostate cancer and monitoring cardiac function (Moskalik et al. 1995, Fenster et al. 1998, Schmidt et al. 1999, Belohlavek et al. 1993). When a breast biopsy becomes necessary, 3-D ultrasound-guided procedures may help doctors by giving 3- D spatial location of the applied needle since the needle no longer has to be aligned within a single scan plane (Smith et al. 2001). Lesions might be sometimes difficult to detect using ultrasound. In some cases, tumors can appear isoechoic compared to surrounding healthy tissue and therefore indiscernible to the viewer under normal ultrasound. The small size and deep location of the lesion and the lack of acoustic backscatterers inside the lesion are other factors that may reduce lesion detectability. Strain imaging is a technique that detects variations in tissue stiffness which generally correlate with pathological phenomena (Ophir et al. 1991). Ultrasound is the most common imaging technology used to produce strain images. It is based on capturing tissue movement after a mechanical compression. Strain imaging overcomes the aforementioned limitations in conventional ultrasound as it does not directly depend on the acoustic properties of tissue. Recent studies have reported promising results for breast imaging using real-time 2-D strain imaging systems. In another study, elasticity imaging was found to have comparable sensitivities to mammography in lesions classified as BI-RADS 3 with good interobserver agreement (Thomas et al. 2006). However, the compression process is inherently three-dimensional and these strain images are mostly two-dimensional. Image quality can suffer if out-of- plane tissue displacements are larger than the elevational beamwidth. 3-D strain imaging has the potential to capture virtually all tissue movements, therefore improving strain 3 image quality by minimizing out-of-plane decorrelation noise. It can also provide additional data about the tumor such as the shape and volume that could help in characterizing the tumor as benign or malignant. Speckle is a common artifact in coherent imaging modalities that degrades target contrast against the surrounding background. Spatial compounding is one of the methods that have been proposed to reduce speckle noise (Burckhardt 1978, Trahey and Smith et al. 1986, Trahey and Allison et al. 1986, Li and O’Donnell 1994, Claudon et al. 2002). It involves averaging images of the same target that have uncorrelated speckle patterns. These images are acquired by viewing the target from different angles using steering or aperture translation techniques (Li and O’Donnell 1994, Trahey and Smith et al. 1986, Dahl et al. 2005, Claudon et al. 2002). Besides speckle noise reduction, spatial compounding offers additional advantages such as reduced acoustic shadowing and refraction correction (Claudon et al. 2002). Clinically, real-time 2-D spatial compounding has shown usefulness in the evaluation of breast lesions, peripheral blood vessels, and musculoskeletal injuries (Seo et al. 2002, Lin et al. 2002, Entrekin et al. 2001). Besides speckle noise reduction, spatial compounding offers additional advantages such as reduced acoustic shadowing and suppressed refraction artifacts. 3-D spatial compounding has the advantages of providing a greater number of independent images and reducing out-of-plane refraction artifacts (Krucker et al. 2000). 4 1.2 Hypothesis The hypothesis of this work is that techniques such as 3-D strain imaging, 3-D ultrasound imaging and 3-D spatial compounding implemented using rectilinear 2-D arrays can improve the detection of near field lesions. Using experimental results, I demonstrate that 3-D ultrasound imaging with elevational focusing followed by 3-D tissue movement tracking can enhance the detectability of stiff lesions in 3-D strain images. The testing results and 3-D ultrasound images acquired using a dual-layer array are presented. The experimental results of the 2-D array with improved axial resolution and reduced number of channels indicate the feasibility of this array design. Finally, I evaluate the feasibility of 3-D spatial compounding with aperture translation by simulation and experiment. Results indicate a significant gain in contrast and speckle noise reduction accompanied with an expected loss in resolution. Factors that can reduce lesion detectability in ultrasound images include: isoechoic lesions, small size, deep location and speckle noise. Table 1.1 shows the role of 3-D strain imaging, 3-D B-mode imaging using the dual-layer array and 3-D spatial compounding in improving lesion detectability in 3-D rectilinear ultrasound images. 3-D strain imaging can improve detectability for lesions that have all of the aforementioned factors that can reduce lesion detectability. It requires that the lesion is stiffer than the surrounding tissue which generally correlates with pathological phenomena. The dual- layer array with enhanced bandwidth can improve the detectability of small sized or deeply located lesions by offering finer axial resolution and a wider range of imaging 5 frequencies. 3-D spatial compounding can reduce speckle noise which improves the contrast of lesions. 3-D strain imaging Dual-layer Array 3-D Spatial Compounding Isoechoic lesions Small size Deep location Speckle noise Table 1.1: The roles of 3-D strain imaging, 3-D B-mode imaging using the dual-layer array and 3-D spatial compounding in improving lesion detectability in 3-D rectilinear ultrasound images. Previous objectives of this research included applying 3-D companding and 3-D spatial compounding using a sparse 2-D array to improve 3-D strain image quality. Due to limitations in the prototype 2-D array used, attempts to implement these ideas were not successful. 2-D companding was implemented successfully using a commercial 1-D array. The 2-D companding experimental results agree with the literature that states that the benefits of companding can mostly be seen for compression rates in the range of about 3-5% (Chaturvedi 1998). Pre- and post- compression images acquired using the prototype 2-D arrays had low correlation for compression rates higher than 2%. This resulted in more errors in time delay estimations (TDE) that companding could not fix. The sparseness of the array which gives rise to clutter is the main reason for the TDE errors when relatively low compression rates were applied (2-3%). 6 It is well established that a key factor in the success of spatial compounding is low correlation between the compounded images (Trahey and Smith et al. 1986). Decorrelation between images increases with increasing the steering angle. The synthetic aperture approach that was adapted for acquiring data using the sparse array involved using an unfocused transmit beam with the size of 4.8 × 4.8 mm. This allows for a maximum steering angle of about ±2° at the depth of 40 mm. This resulted in relatively high correlation between the strain images generated using different steering angles from -2° to +2°. This in turn resulted in an insignificant improvement in the quality of strain imaging. Sections 6.2 to 6.4 explore in more detail previous research objectives and how lesion detectability in 3-D strain images can be improved using the new 2-D array prototypes with improved performance. 1.3 Overview Chapter 2 describes fundamental ultrasound imaging and transducer concepts that are used throughout the rest of the thesis. Chapter 3 starts with discussing the fundamentals of 2-D and 3-D strain imaging. Next it describes a 3-D strain imaging system using a prototype rectilinear sparse 2-D array. It also presents 3-D strain images of gelatin/agar phantoms with cylindrical and spherical stiff inclusions of various diameters. In addition to the 3-D strain images generated using 3-D beamforming followed by 3-D strain imaging, all other signal processing combinations of 2-D or 3-D beamforming followed by 2-D or 3-D strain imaging were used to generate 3-D strain images. Implementing the different signal 7 processing combinations allowed isolating and identifying the benefits of 3-D beamforming with elevational focusing and 3-D TDE for strain imaging. The quality of all images was compared qualitatively first, then quantitatively to evaluate the performance of the 3-D strain imaging system. In Chapter 4, the testing results and 3-D ultrasound images acquired using a dual- layer array are presented. 3-D images of an 8 mm diameter anechoic cyst phantom and five pairs of nylon wires embedded in a clear gelatin phantom are displayed. I also present simulation and experimental beamplots. Chapter 5 describes a 3-D ultrasound imaging system using a prototype 2-D array with row column addressing used for 3-D spatial compounding. In this work, I use subaperture translation to compound nine uncorrelated volumes. Using this technique eliminates the need for registration. I present results of several simulations to demonstrate the feasibility of this idea. Finally, I present images of gelatin phantoms with a cylindrical -11 dB cyst and a cylindrical +11 dB tumor and quantify the improvement in CNR and speckle SNR for the compounded images over standard B-mode images. Using five pairs of nylon wires embedded in a clear gelatin phantom, the decrease in lateral resolution caused by this spatial compounding technique was evaluated. In Chapter 6, the results presented in each of the previous chapters are summarized. Previous research objectives are illustrated. The implications of this work for future research are also discussed. Initial investigations indicates that 3-D spatial compounding 8 with aperture translation could be successfully applied to 3-D strain imaging to increase contrast and minimize artifacts in strain images. 3-D companding is another method that can be considered to minimize decorrelation noise in strain images. Future work will also include in vitro animal experiment to test 3-D strain imaging and 3-D spatial compounding more rigorously. 9 CHAPTER 2 ULTRASOUND IMAGING 2.1 Ultrasound Physics Ultrasound is a sound wave with frequencies higher than the upper limit of human audible range of 20 kHz. Unlike electromagnetic waves, ultrasound requires a medium to travel and it cannot propagate in vacuum. When an acoustic wave reaches a boundary between two materials or objects with different acoustic impedance, part of the acoustic energy gets reflected and the remainder gets transmitted. If the size of the second object is smaller than or comparable to the wavelength and the wavefront of the ultrasound wave, then the reflection phenomenon is termed scattering. The amount of reflected energy can be calculated using the reflection coefficient. Assuming that the direction of acoustic propagation is perpendicular to the boundary between the two materials, the reflection coefficient R is defined as: 1 2 1 2 Z Z Z Z R + − = (Eq. 2.1) where Z 1 , Z 2 are the specific acoustic impedances of the material the acoustic signal is traveling in and the next material respectively. R varies between -1 to 1. Negative values for R means the reflected signal is 180° out of phase with the incident signal. If Z 2 = Z 1 (the wave is traveling in the same medium) then R=0 and there will be no reflected wave. If Z 2 >> Z 1 then R≈1 which means most of the signal will be reflected back when there is 10 a big mismatch between the impedance of the media. For plane waves, the specific acoustic impedance is defined as: c Z ρ ± = = u p (Eq. 2.2) where p is the acoustic pressure in a medium and u is the associated particle speed. ρ and c are the density of the material and speed of sound in the material respectively. The plus or minus sign depending on whether the propagation is in the plus or minus direction (Kinsler et al. 1982). In diagnostic ultrasound imaging, sound waves are transmitted through biological tissue. The reflected acoustic signals are collected and converted to electric signals. The amplitude of the electric signals cause a corresponding increase or decrease in the brightness of the displayed image and the total transit time from the initial pulse transmission to the reception of the echo is proportional to the depth of the boundary. 2.2 Transducer Fundamentals Ultrasonic transducers made of piezoelectric materials are commonly used in medical ultrasonography. An electric excitation applied to a piezoelectric material will emit an acoustic wave in the surrounding media. Similarly, when an acoustic wave arrives at the piezoelectric material surface, an electric signal proportional to amplitude is generated. The most common forms of ultrasound transducers are circular disk single element transducer, 1-D transducer array and 2-D transducer array. Figure 2.1 shows a single element transducer and the 1-D RF data that it generates for different depths of material 11 boundaries M1 and M2. Calculating the envelope of 1-D RF data yields the “A-mode” image. 2-D images can be generated using a single element transducer by mechanical translation. 2-D images are commonly displayed in B-mode format. In B-mode images, the amplitude of the envelope of the RF data is mapped to pixel brightness in a gray scale image. Figure 2.1: In 1-D ultrasonic imaging, echo signals correspond to tissue boundaries. Boundary depths are linearly related to echo arrival time. Calculating the envelope of echoes RF data yields the A-mode image. 2.3 Transducer Arrays Transducer arrays are sets of individually controlled piezoelectric material elements used to give sets of 1-D image lines that are processed and grouped together to give a 2-D or 3-D image. Transducer arrays have the advantage of electronic steering of the acoustic beam which gives a wider view far away from the transducer and focusing the acoustic Time (depth) Single element ultrasound transducer M1 M2 A-mode image RF data Time (depth) Amplitude 12 beam which provides a higher lateral resolution at the focal point. A 1-D linear sequential array has transducer elements along one direction. Consequently, focusing of the ultrasound beam can only be done in this direction. Figure 2.2a shows a typical 1-D linear array and its field of view where line A is created using the on-axis or center subaperture, indicated by the gray shading, and line B is creating using an off-axis subaperture. The acoustic lens of a 1-D array focuses the beam at a single, predetermined depth. The beam is well focused in elevation at this depth, but diverges at depths away from the focus degrading image quality due to the increase in slice thickness. Figure 2.2: Ultrasonic arrays and their fields of view (a) 1-D linear array. (b) 2-D rectilinear array (Yen and Smith 2001). 2-D arrays, on the other hand, have equal focusing capability in both lateral directions (Figure 2.2b), which make them capable of producing a thinner slice thickness since dynamic elevation focusing is possible. Line A is created using the on-axis or center subaperture, indicated by the gray shading, and line B is creating using an off-axis 13 subaperture. Overall, 2-D arrays have advantages of higher speed, more reliability, and better elevational resolution through electronic focusing. For breast ultrasound, a 1-D linear sequential array may contain 256 elements. Analogously, a fully sampled 2-D array would have up to 256 × 256 = 65,536 elements. At this time, constructing such a large fully sampled 2-D array would be an unrealistic goal due to difficulty in fabricating such a large number of elements, high electric impedance of the small array elements and interconnection problems. One well-known method to overcome these problems is to use a very sparse array design where a small portion of the elements is used (Yen et al. 2000). 2.4 Beamforming In transducer arrays, electronic focusing and steering the acoustic field is done in a process called beamforming. Figure 2.3 shows a 14-element 1-D array of which 7 elements are focusing at focal point #1. The echo signals S 1-7 represent a point target at the focus. 14 Figure 2.3: Beamforming in transducer arrays. Because the distances from the focal point are different, the reflected echoes from focal point #1 arrive at different times. If certain time delays where applied to signals according, all the echoes will line up and can be summed together to form an image line. Since the distance from the focal point to the elements is different, the echo signals reflected from the target arrive to the elements at different times. Applying time delays to the signals according to Equation 2.3 (in which τ 4 =0 and τ 1 =τ 7 >τ 2 =τ 6 >τ 3 =τ 5 >τ 4 ) will arrange all the echoes in phase. The lined up echoes are summed to form an image line. c z c z z y y x x f f ei f ei f ei i − − + − + − = 2 2 2 ) ( ) ( ) ( τ (Eq. 2.3) 15 where ei ei ei z y x , , define the location of the element i in 3-D space, f f f z y x , , define the location of the focus, and c denotes the speed of sound in the imaging medium. The 7 elements form a portion of the entire aperture otherwise known as a subaperture. Receive dynamic focusing is a common practice used to achieve better image quality by focusing at different depths as the echoes arrive in time (Focal point #1 & #2). If the same subaperture was used to focus at focal point #3 (a point off the center of the subaperture), beamforming involves focusing and steering. 2.5 Beam Patterns Beam profiles in ultrasonic transducers include two types corresponding to depth: near field (or Fresnel zone) and the divergent far field (or Fraunhofer zone). Figure 2.4 shows the magnitude of pressure field of a circular transducer driven by a sinusoidal source. The near field is the region where pressure shows rapid variations (depth < Z R ). In the far field (depth > Z R ), the pressure is more uniform and decreases approximately according to 1/Z. 16 Figure 2.4: The near field and the far field of a circular transducer. In the near field, the magnitude of pressure is oscilating. In the far field, the pressure is more uniform and decreases approximately according to 1/Z (Christensen 1988). The transition point Z R between the near field and the far field for a single element circular transducer is defined as: λ 4 2 D Z R = (Eq. 2.4) where D is the diameter of the transducer and λ is the wavelength. Equation 2.4 can be used as an approximation of Z R for non-circular and array transducers. The beam profile is very complex and not well defined in the near field. In the far field however, it takes the general shape shown in Figure 2.5. |P(Z)| Depth Near field Far field Z R 17 Figure 2.5: The far field angular beam profile of a typical transducer array. An ideal beam profile would have a zero width main lobe only. A wider main lobe makes two targets separated by a distance less than the beam width undistinguishable. Although most of the radiated energy is contained in the main lobe, the unwanted side lobes and grating lobes contain some energy radiated at an angle. When the energy of these lobes is reflected by off-axis structures, the signal produced is artifactual and produce "ghost images" blurring the main image. Side lobes are present in all transducers and arise from additional width, length or radial mode vibration of the piezoelectric material. Grating lobes are specific to transducer arrays and are associated with the constructive interference caused by the periodic nature of regularly spaced array elements. They can be observed at an angle Φ g given by:       = − d n g λ φ 1 sin (Eq. 2.5) Φ g Main lobe Grating lobe Side lobe 90° Intensity -90° 18 where λ is the wavelength, d is the pitch (the distance between the centers of successive array elements) and n=±1, ±2, … The pitch value can be chosen to push the grating lobes out the imaging field (i.e. Φ g > 90°). 2.6 Spatial Resolution The axial and lateral resolutions of a transducer are mainly determined by the transmitted pulse duration and the aperture size of the transducer respectively. The resolvability of the echoes from two targets in the axial or lateral direction depends on the spatial resolution in that direction (Figure 2.6). Figure 2.6: Spatial resolution: (a) The echoes from two targets can be easily resolved with full beam width separation between them (b) Less separation between the targets makes them more difficult to resolve (c) If the separation is much less than the beam width, the two echoes become indistinguishable The lateral resolution LR at the focal point Z of a transducer and the axial resolution AR are given by: 19 D Z LR λ = (Eq. 2.6) 2 λ n AR = (Eq. 2.7) where D is the size of the aperture used, λ is the wavelength of the transducer and n is the number of cycles in the transmitted pulse. LR at depth Z is also equal to the -6dB beam width at that depth. From these relationships it can be noticed that an increase in frequency will decrease the wavelength and thus improve both axial and lateral resolutions. Also, a higher BW will improve axial resolution and a decrease in the beam width will improve lateral resolution (Christensen 1988). 20 CHAPTER 3 3-D ULTRASOUND STRAIN IMAGING 3.1 Introduction In contrast to ultrasound imaging that detects the acoustic impedance differences, strain imaging (also known as elastography and elasticity imaging) detects variation in tissue stiffness associated with different tissues. Since its start in 1991 by Ophir et al., there has been significant progress in strain imaging (Ophir 1991). Although clinical elastography is still in early development, it has shown usefulness in differentiating between benign and malignant lesions in the breast, prostate cancer detection, intravascular applications, and ablation monitoring (Lorenz, Sommerfeld et al. 1999; Schaar, Korte et al. 2000; Hiltawsky, Kruger et al. 2001; Varghese, Zagzebski et al. 2002; Hall, Zhu et al. 2003). Real-time strain imaging provides visual feedback that can guide the patient positioning and compression direction. A real-time strain imaging system was developed and used in a small clinical breast imaging study (Hall, Zhu et al. 2003). Another system was developed with a 30 frames/sec (Zahiri-Azar and Salcudean 2006). 3-D strain imaging is another promising improvement. A freehand 3-D system was developed using a 1-D array and a position sensor that accounts for freehand movement errors in the image reconstruction process (Lindop, Treece et al. 2006). To date, strain imaging has largely been done in two dimensions (Alam and Ophir 1997; Insana, Chaturvedi et al. 1997; Kaisar, Ophir et al. 1998; Konofagou and Ophir 1998). Since tissue movement is inherently three-dimensional, these images can suffer 21 from decorrelation noise caused by out-of-plane tissue movement in elevation if the tissue is not physically confined (Alam and Ophir 1997; Kallel and Ophir 1997; Kallel, Varghese et al. 1997). Some techniques involved 2-D companding in which post- compression RF data are compressed or expanded to improve coherence with pre- compression data and minimize decorrelation errors caused by tissue movement in azimuth (Alam and Ophir 1997; Insana, Chaturvedi et al. 1997; Kaisar, Ophir et al. 1998; Konofagou and Ophir 1998). 3-D companding results were also reported showing more improvement to the strain images quality since it accounts for out-of-plane tissue movement in elevation (Insana, Chaturvedi et al. 1997). 3-D strain imaging has the potential to capture virtually all tissue movements, therefore improving image quality by minimizing out-of-plane decorrelation noise. It can also provide additional data about the tumor that could help in characterizing the tumor as benign or malignant such as the shape and volume of the tumor. Some 3-D ultrasound systems use a 1-D array with mechanical translation in the elevational direction using motorized or freehand methods to acquire multiple B-scans followed by reconstruction to create the 3-D image (Fenster and Downey 1996). A 1-D linear sequential array has transducer elements along one direction only. Consequently, focusing of the ultrasound beam can only be done in this direction. Additionally, the motorized or freehand scanning might produce errors and artifacts that could degrade the 3-D image quality. 2- D arrays, on the other hand, have equal focusing capability in both lateral directions, giving them a better lateral resolution. Scanning with 2-D arrays is done electronically in both lateral directions, eliminating the need for motorized or freehand scanning. Overall, 22 2-D arrays have advantages of higher speed, more reliability, and better elevational resolution through electronic focusing. This chapter describes a 3-D strain imaging system using a prototype rectilinear sparse 2-D array. It also presents 3-D strain images of gelatin/agar phantoms with cylindrical and spherical stiff inclusions of various diameters. In addition to the 3-D strain images generated using 3-D beamforming followed by 3-D strain imaging, other signal processing combinations of 2-D or 3-D beamforming followed by 2-D or 3-D strain imaging were used to generate strain images. The quality of all images was compared to evaluate the performance of the 3-D strain imaging system. 3.2 Strain Imaging Principles In order to illustrate the gradient based strain calculation, it is convenient to start with considering a one-dimensional spring system shown in Figure 3.1 where the spring constants represent the stiffness of tissue regions. The spring S1 is stiffer than S2. Both springs have the same length (L1= L2). After applying a certain amount of compression, S1 compresses less than S2 (ΔL1< ΔL2). The local displacements can be estimated starting with 0 at depth= 0, ΔL1 at depth= L1 and ΔL1+ ΔL2 at depth= L1+L2. Differentiating displacements gives strain in which low and high strain values correspond to hard and soft materials respectively. 23 Figure 3.1: Measurement of strain in a one dimensional spring system of different spring constant by applying compression, calculating displacements and differentiating to find strain values. Strain imaging is based on applying static or dynamic mechanical compression to a tissue sample, collecting pre- and post-compression radio frequency (RF) data and using signal processing methods to evaluate strain by calculating displacements and applying a gradient operation to estimate strain (Ophir 1991; Chaturvedi, Insana et al. 1998; Kaisar, Ophir et al. 1998; Varghese, Ophir et al. 2000). Figure 3.2 illustrates the basic steps applied in 2-D strain imaging. A pre-compression (D1) and post-compression (D2) 24 ultrasound images of the same size are acquired. A window of D1 is selected (W1) and compared to several windows (W2) in D2 within a search region S. After finding a W2 that best matches W1 within S, the time delay between W1 and W2 is calculated. The displacement can be calculated from the time delay (linearly proportional relationship) and the value is assigned to the location of W1. The process is repeated for selected windows in D1 and a map of displacements is created. The displacements are differentiated giving the strain image. Figure 3.2: First step in stain imaging is to acquire a pre-compression image (D1) and a post-compression image (D2). A window W1 is selected from D1 and compared to several windows W2 selected from D2 within a search region S to calculate displacements. 3.2.1 Time Delay Estimation Several methods can be used in finding the best match for W1 and time delay estimation (TDE) (Viola and Walker 2005). One of the most commonly used methods is Ultrasonic Transducer array Ultrasonic Transducer array Post-Compression S (N,M) W2 (U,P) Pre-Compression W1 (U,P) D1 D2 Tissue 25 the normalized cross-correlation. The cross-correlation coefficient ρ can be calculated for two 1-D signals s a (t), s b (t) using the following relationship: ∫ ∫ ∫ + + = T T b a T b a dt t s dt t s dt t s t s 0 0 2 2 0 ) ( ) ( ) ( ) ( ) ( τ τ τ ρ (Eq. 3.1) The denominator of Equation 3.1 is the normalizing term making ρ vary from -1 to 1. If the signals are identical, ρ will have a value of 1 at τ =0. So using cross-correlation, the max value of ρ(τ) correspond to the best match between the two signals. The sum of absolute differences (SAD) is another TDE algorithm that provides a comparable performance to cross-correlation but with faster speed (Chaturvedi, Insana et al. 1998). In Figure 3.3 SAD is applied to the 1-D signal S1(x) and parts of S2(x) of equal size. The SAD values reach a minimum value at x=50 which corresponds to the best match between S1 and S2. 26 Figure 3.3: 1-D TDE using SAD: SAD is applied to the signal S1(x) and parts of S2(x) of equal size. The minimum value of the SAD vector corresponds to the best match and its location corresponds to the time delay. The same idea can be applied in two and three dimensions. Applying SAD to W1 and W2 in Figure 3.2 involves subtracting the all pixel values of W1 from those of W2 (both with the size of U × P), applying absolute value to the difference and summing the outcomes to give a single SAD coefficient: ∑∑ = = + + − = U u P p m p n u W p u W m n 1 1 | ] , [ 2 ] , [ 1 | ] , [ ε (Eq. 3.2) where ε is the matrix containing N × M SAD values. W1[u,p] and W2[u, p] are the [u,p] pixel value in W1 and W2 respectively. Applying SAD to the windows W2 in the search region S of size N × M gives a matrix of N × M SAD values. The minimum value of the 27 matrix ε corresponds to the best match and its location in [n,m] corresponds to the estimated tissue displacement. In three dimensions, 3-D windows W1 and W2 are selected from the 3-D pre- and post-compression images respectively with the size of U × P × Q. W2 is selected within a search S to give a SAD values matrix with the size of N × M × L: ∑∑∑ = = = + + + − = U u P p Q q l q m p n u W q p u W l m n 1 1 1 | ] , , [ 2 ] , , [ 1 | ] , , [ ε (Eq. 3.3) where ε is the matrix containing N × M × L SAD values. W1[u,p,q] and W2[u, p,q] are the [u,p,q] pixel value in W1 and W2 respectively. The minimum value of the matrix ε corresponds to the best match and its location in [n,m,l] corresponds to the estimated tissue displacement. 3.2.2 3-D vs. 2-D Strain Imaging Out of imaging plane decorrelation noise is one of the disadvantages of 2-D strain imaging. The assumption that tissue is incompressible (Poisson’s ratio≈0.5) is well accepted in literature (Kallel and Ophir 1997). This means that after an axial compression ΔZ, expansions in azimuth ΔX=0.5 ΔZ and in elevation ΔY=0.5 ΔZ are expected (Figure 3.4). Axial compression and expansion in azimuth cause the deformation of the post- compression image giving rise to decorrelation noise. If the elevational displacement exceeds ±15% of the elevational beam width it causes out-of-plane decorrelation errors (Insana, Chaturvedi et al. 1997). 3-D companding was proposed to compensate for out- 28 of-plane tissue movement by using slices of a 3-D volume to generate a 2-D strain image with minimized decorrelation noises (Insana, Chaturvedi et al. 1997). Figure 3.4: Axial compression and lateral expansion in strain imaging. In principle, 3-D strain imaging captures all tissue movements and thus eliminating out-of-plane decorrelation noise. 3-D strain imaging can also provide important information about the lesion that could help in characterizing the tumor such as the volume. Shape factors such as height, width and length can be used characterize breast lesions (Sahiner, Chan et al. 2003). 29 3.3 Methods 3.3.1 System Description To test the feasibility of 3-D strain imaging using a 2-D array, an experimental setup was designed and built using a sparse 2-D array created in previous work, Ultrasonix Sonix RP ultrasound system and a receive mode multiplexer circuit. The prototype sparse 2-D array has a 40 × 40 mm aperture, 5 MHz center frequency, 45% -6 dB fractional bandwidth, 1024 receivers, 169 transmitters and scans a 40 × 40 × 60 mm 3-D rectilinear volume (Yen and Smith 2004). The array consists of 13 × 13 = 169 transmit elements, each having dimensions of 2.4 × 2.4 mm. Four receive elements with dimensions of 0.6 × 0.6 mm were embedded in a staggered pattern inside each of the transmit pistons (Figure 3.5). Figure 3.5: Sparse 2-D array. (a) Schematic of the entire array. (b) Schematic of a single cell with 0.6 × 0.6 mm receive elements and surrounding transmitters (Yen and Smith 2004). 30 The Ultrasonix Sonix RP ultrasound system with research capability has a 40 MHz sampling frequency, 128 channels and 32 analog to digital converters. Using Texo, an open source C++ program provided by Ultrasonix, this system allows the user to acquire raw RF data and lets the user control the transmit aperture size, transmitted power, transmitted frequency, receive aperture size, receive gain and acquisition length. To acquire the RF data from the 1,024 receive elements, an additional multiplexing circuit was used to interface the receive elements to the 128 channels of the system. In previous work, a 4:1 multiplexer board which uses 32 MAX4052/A dual 4:1 multiplexers was designed (Maxim Integrated Products, Sunnyvale, CA, USA) (Yen and Smith 2004). Multiplexing was done in two stages. Stages 1 and 2 use the 4:1 multiplexer boards to multiplex the 1024 receive elements down to 256 and 64 outputs respectively (Figure 3.6). The data was collected sequentially, keeping the sampling frequency of all RF data at 40 MHz. The parallel port of the Ultrasonix system was used along with C++ commands embedded in Texo to control the addressing of the multiplexer boards. All RF signals were acquired 32 times and averaged to minimize random noise. Figure 3.6: A diagram that shows the interconnection between the receive elements and the Ultrasonix system. 256 outputs 64 outputs Stage 2: 4:1 external multiplexing Stage 1: 4:1 external multiplexing 1024 receive elements The Ultrasonix system 31 3.3.2 Phantom Fabrication 70 × 70 × 70 mm gel based phantoms with stiff agar inclusions were used to test the 3- D strain imaging system. The following recipes were used: Background material: 400 g DI water, 36.79 g N-Propanol, 0.238 g Formaldehyde, 24.02 g Gelatin (275 Bloom) and 3.89 g Graphite. Inclusion material: 50 g DI water, 4.6 g N-Propanol, 0.103 g Formaldehyde, 1.1 g Agar and 1.15 g Graphite (Hall 1997). Using these recipes, inclusions were estimated to be roughly 10 times stiffer than the background. 3.3.3 Data Collection and Signal Processing 3-D beamforming: Using a synthetic aperture approach in 3-D beamforming, one transmit element was excited and signals from all 1,024 receive elements were acquired. This process was repeated for the 169 transmit elements of the array. Off-line beamforming was applied to the RF element data using Matlab (Mathworks Inc., Natick, MA) to create pre- and post- compression 3-D RF echo fields. Azimuthal and elevational receive dynamic focusing every 1mmwas used along with an expanding aperture to keep the F-number at 2. The resulting B-mode image was 38.4 × 38.4 × 41.6 mm with 129 × 129 = 16,641 RF lines and 0.3 mm line spacing. The RF data were band pass filtered in Matlab to minimize noise. 3-D strain imaging: The first step in 3-D strain imaging is to estimate time delays between the pre- and post- compression RF data that are linearly proportional to axial tissue displacement. Since 3-D time delay estimation (TDE) involves heavy processing of 32 large data sizes, a fast TDE method was needed. The sum of absolute differences (SAD) algorithm was used since it provides similar performance to cross correlation but with eight times fewer arithmetic operations (Chaturvedi et al. 1998). This method involves taking a 3-D kernel of pre-compression data and applying a 3-D search for the best matching kernel of the same size in the post-compression data. The output of this search is a 3-D matrix of coefficients. The minimum value of this matrix corresponds to the best match and its location in 3-D corresponds to the estimated tissue displacement. In this approach, TDE was done in two steps. In the first step, coarser time delays were calculated with non-overlapping kernels with the size of 0.6 mm axially and 1 mm in the azimuth and elevation direction and 1 mm axial and 1 mm lateral separation between kernels. A search window size of 1 × 1.1 × 1.1 mm (54 × 7 × 7 samples) was chosen by visual inspection. Sometimes, a false minimum value in the coefficients matrix becomes smaller than the true minimum. Median filtering was used to minimize the number of false minima in this step (Righetti et al. 2002). In the second step, finer time delays were calculated using overlapping kernels with a size of 0.6 mm axially and 1 mm in both lateral directions and 0.1mm axial and 0.3mm lateral separation between kernels (Chaturvedi et al. 1998). This SAD coefficients matrix was interpolated by a factor of five axially using cubic spline interpolation. The number of false minima was reduced in this step by using the coarse time delays as the center for the search window with the size of 0.1 × 1.1 × 1.1 mm (7 × 7 × 7 samples) (Chaturvedi et al. 1998, Zahiri-Azar and Salcudean 2006). The resulting axial displacements were then averaged over a 1.7 mm window (Pellot-Barakat et al. 33 2004). Axial strain was estimated by calculating the gradient of two adjacent values of the averaged displacements in the axial direction. 2-D beamforming: 2-D beamforming was applied to subsets of the same 3-D data to simulate a 1-D array. In this approach, separated azimuthal and elevational beamforming were separated (Fernandez et al. 2003). In the azimuth direction, receive dynamic focusing was applied every 1 mm along with an expanding aperture to keep the F-number at 2. In the elevation direction, a single focus at 30 mm was used to simulate the lens used in 1-D arrays with an aperture size of 9.6 mm. The resulting B-mode image was 38.4 mm in the azimuth direction and 41.6 mm axially with 129 RF lines and 0.3 mm line spacing. 2-D strain imaging: The same kernel size, search window size (in the axial and azimuthal directions) and TDE techniques used in the 3-D strain image creation were used in 2-D strain image creation. The coarser time delays were calculated using non- overlapping kernels with the size of 0.6 mm axially and 1 mm azimuthally. The separation between kernels is 1 mm axially and 1 mm azimuthally. A 1-D 7-sample median filtering was applied to the coarser time delays in both directions. A 0.6 mm axial by 1 mm azimuthal kernel size was used in calculating the finer time delays with 0.1 mm axial and 0.3 mm azimuthal separation between kernels. The SAD coefficients matrix was interpolated by a factor of 5 in both directions using cubic spline interpolation. The size of the search window was 0.1 × 1.1 mm (7 × 7 samples). The resulting axial displacements were then averaged over a 1.7 mm window and differentiated to create the axial strain image. To examine the benefits of 3-D TDE over 2-D TDE and the benefits of 3-D 34 beamforming over 2-D beamforming for strain imaging, the following signal processing combinations were investigated: • 2-D beamforming followed by 2-D TDE (or 2-D strain imaging) (2-D BF, 2-D TDE). • 3-D beamforming followed by 2-D TDE (2-D BF, 3-D TDE). • 2-D beamforming followed by 3-D TDE (2-D BF, 3-D TDE). • 3-D beamforming followed by 3-D TDE (3-D BF, 3-D TDE). This combination utilizes both of the elevational focusing and the 3-D TDE capabilities of the system. 3.3.4 Performance Evaluation Elastographic contrast to noise ratio (CNRe): the following equation was used to evaluate CNR (Chaturvedi et al. 1998). b t 2 b t s var s var ) s mean s mean 2( CNRe ) ( ) ( + − = (Eq. 3.4) In Equation 3.4, var(s) denote the variance. The subscripts t and b represent the target and background, respectively. CNR values were estimated using 3-D kernels chosen from inside the inclusion (s t ) and from the background (s b ). Elastographic signal to noise ratio (SNRe): the following equation was used to evaluate SNR (Varghese and Ophir 1997). ) ( ) ( b b s std s mean SNRe = (Eq. 3.5) 35 In Equation 3.5, std(s) denote the standard deviation. The subscript b represents the background. SNR values were estimated using 3-D kernels chosen from the background. Strain Contrast (SC): the following equation was used to evaluate SC. ) ( ) ( t b s mean s mean SC = (Eq. 3.6) In equation 3.6, the subscripts t and b represent the target and background, respectively. 3.4 Results 3-D pre-compression and post-compression RF data sets of the 70 × 70 × 70 mm gel based phantoms with 12 mm, 9 mm and 6 mm diameter stiff agar cylindrical and spherical inclusions were obtained using the sparse 2-D array. The phantoms were compressed using a 125 × 125 mm aluminum plate with a 1.5% axial compression. Figure 3.7 presents the 3-D axial strain image (generated using 3-D beamforming and 3- D TDE) of the cylindrical inclusion phantoms in a collage of slices in 3 orthogonal planes: azimuth, elevation and axial. Azimuth is the same plane as a 1-D array, elevation is perpendicular to azimuth and the transducer face and axial is analogous to a C-scan which is parallel to the transducer face. The arrows in the top row indicate where the elevational and axial slices were taken and the ones in the middle row indicate where the azimuthal slices were taken. 36 Figure 3.7: Collage of slices in azimuthal, elevational and axial slices of the 3-D axial strain image of the 12 mm, 9 mm and 6 mm (left to right) cylindrical inclusion phantoms generated using 3-D beamforming and 3-D TDE. Top row: azimuthal slices. The arrows indicate where the elevational and axial slices were taken. Middle row: elevational slices. The arrows indicate where the azimuthal slices were taken. Bottom row: Axial slices. The size of the images is 27 mm axially and 30 mm in azimuth and in elevation. The transducer is located on top of the azimuthal and elevational slices and parallel to the axial slices. In Figure 3.8, three perpendicular slices of the 9 mm cylindrical phantom 3-D strain image is shown. 37 Figure 3.8: Perpendicular slices of the 9 mm cylindrical inclusion phantom 3-D strain image. Figure 3.9 is another collage of elevational, azimuthal and axial slices of the 3-D axial strain images (generated using 3-D beamforming and 3-D TDE) of the spherical inclusion phantoms. 38 Figure 3.9: Collage of slices in azimuthal, elevational and axial slices of the 3-D axial strain image of the 12 mm, 9 mm and 6 mm (left to right) spherical inclusion phantoms generated using 3-D beamforming and 3-D TDE. Top row: azimuthal slices. The arrows indicate where the elevational and axial slices were taken. Middle row: elevational slices. The arrows indicate where the azimuthal slices were taken. Bottom row: Axial slices. The size of the images is 27 mm axially and 30 mm in azimuth and in elevation. The transducer is located on top of the azimuthal and elevational slices and parallel to the axial slices. Figure 3.10 shows 3 perpendicular slices of the 9 mm spherical phantom 3-D strain image. The size of the images in Figures 3.7 to 3.10 is 30 mm in azimuth and elevation and 27 mm axially. In all cases, the inclusion is visible in all slices. 39 Figure 3.10: Perpendicular slices of the 9 mm spherical inclusion phantom 3-D strain image. Figure 3.11 compares the CNR values of the axial strain images of the 6 phantoms using the four signal processing combinations described in the Methods section. A CNR kernel size of 26x13x9=3042 voxels (axially, in elevation and in azimuth respectively) was used in all cases. 40 Figure 3.11: A bar graph of the CNR values of the axial strain images of the 6 mm, 9 mm and 12 mm cylindrical and spherical inclusion phantoms. The strain images were generated by applying: 2-D beamforming followed by 2-D TDE (2-D BF, 2-D TDE), 3-D beamforming followed by 2-D strain imaging (3-D BF, 2-D TDE), 2-D beamforming followed by 3-D strain imaging (2-D BF, 3-D TDE) and 3-D beamforming followed by 3-D strain imaging (3-D BF, 3-D TDE). As an example, Figure 3.12 shows the same elevational slice of the strain image volumes of the 9 mm spherical phantom generated using the four signal processing combinations. The white rectangles are the cross section of the 3-D kernels used to calculated CNR values. The kernel representing the background was chosen from a region in the upper left corner of the images, which is relatively free from stress concentration artifacts (Chaturvedi et al. 1998). The size of the images is 23.7 mm axially and 26.1 mm in azimuthally. 41 Figure 3.12: Collage of the same elevational slice of the axial strain images of the 9 mm spherical inclusion phantom generated using the four signal processing combinations: 2- D BF, 2-D TDE, 3-D BF, 2-D TDE, 2-D BF, 3-D TDE and 3-D BF, 3-D TDE. The white rectangles inside the inclusion and on the top left corner of the images are the cross section of the 3-D kernels used to calculate CNR. The size of the images is 23.7 mm axially and 26.1 mm in azimuth. Figure 3.13 compares the SNR values of the axial strain images of the 6 phantoms using the four signal processing combinations described in the Methods section. A SNR kernel size of 250x20x70=350,000 voxels (axially, in elevation and in azimuth respectively) was used in all cases. 42 Figure 3.13: A bar graph of the SNR values of the axial strain images of the 6 mm, 9 mm and 12 mm cylindrical and spherical inclusion phantoms. The strain images were generated by applying: 2-D beamforming followed by 2-D TDE (2-D BF, 2-D TDE), 3-D beamforming followed by 2-D strain imaging (3-D BF, 2-D TDE), 2-D beamforming followed by 3-D strain imaging (2-D BF, 3-D TDE) and 3-D beamforming followed by 3-D strain imaging (3-D BF, 3-D TDE). Figure 3.14 shows the strain contrast values of the axial strain images of the 6 phantoms using the four signal processing combinations described in the Methods section. The average strain contrast for all six phantoms using 3-D BF, 3-D TDE is 4.92 ± 0.83. The ratio between the modulus contrast of the phantoms and the strain contrast estimated is fairly close to two which is consistent with literature (Ponnekanti et al. 1995). 43 Figure 3.14: A bar graph of the strain contrast values of the axial strain images of the 6 mm, 9 mm and 12 mm cylindrical and spherical inclusion phantoms. The strain images were generated by applying: 2-D beamforming followed by 2-D TDE (2-D BF, 2-D TDE), 3-D beamforming followed by 2-D strain imaging (3-D BF, 2-D TDE), 2-D beamforming followed by 3-D strain imaging (2-D BF, 3-D TDE) and 3-D beamforming followed by 3-D strain imaging (3-D BF, 3-D TDE). 3.5 Discussion I have presented initial 3-D strain images using a sparse 2-D array prototype that has a 40 × 40 mm aperture, 5 MHz center frequency, 1024 receivers, 169 transmitters. Using this system, I carried out several signal processing combinations to isolate and identify the benefits of 3-D beamforming with elevational focusing and 3-D TDE for strain imaging. Although more experiments are needed to achieve statistical significance, this initial data suggests that a large CNR improvement can be achieved by switching from 2- 44 D TDE to 3-D TDE. Since strain imaging using SAD is based upon finding the best match in the post-compression data for a certain kernel in the pre-compression data and calculating the displacement between the two, the improvement in the quality of the 3-D results are mainly due to the fact that finding the match and calculating the displacement for a 3-D block gives more accurate results than doing so for a 2-D rectangle. Similar improvements between 1-D and 2-D strain imaging were noticed in early investigations. Comparing the CNR and SNR values of 2-D BF, 3-D TDE to those of 3-D BF, 3-D TDE, the latter shows an improvement in CNR and SNR for all six phantoms. This indicates that 3-D beamforming with elevational focusing can improve the performance of strain imaging with 3-D TDE by increasing elevational resolution. 3-D BF, 2-D TDE is expected to perform worse than 2-D BF, 2-D TDE since 3-D BF accentuates the speckle movement in the elevational direction, something that 2-D TDE cannot detect very well. This resulted in a higher CNR and SNR values for the most part, otherwise the results were comparable. Qualitatively, it is easier to identify the inclusion in 3-D BF, 3-D TDE image than in the other 3 cases (Figure 3.12). In some parts of the strain images, vertical stripes of noise can be noticed. This is mainly caused by the non-uniformity of the transmitted energy from the array elements. 45 CHAPTER 4 A DUAL-LAYER TRANSDUCER ARRAY FOR 3-D RECTILINEAR IMAGING 4.1 Chapter Overview Previous objectives of this research included applying 3-D companding and 3-D spatial compounding using a sparse 2-D array to improve 3-D strain image quality. Due to limitations in the prototype 2-D array used, attempts to implement these ideas were not successful. In an effort to overcome these limitations using a different array design, I interfaced and tested the dual-layer array introduced in this chapter. Additionally, I acquired and beamformed RF data of tissue mimicking phantoms to generate 3-D B- mode images. When compared to the sparse array, this array has the advantages of finer axial resolution and lateral resolution through the enhanced bandwidth and transmit focusing capabilities of the transducer. This gives promise of keeping TDE errors at an acceptable level when higher compression rates are applied. Additionally, a transmit beam with the size of 38.4 mm can be generated using this array. This enables implementing spatial compounding for strain imaging with a wide range of steering angels with the maximum of ±26° at a depth of 40 mm. Unfortunately, the array degraded before I was able to use it to acquire strain images. This chapter discusses the concept and engineering of the dual-layer array and presents the testing results and 3-D ultrasound images acquired using it (Yen et al. 2009). Section 4.2 is an introduction about 3-D rectilinear ultrasound imaging. Section 4.3 describes the array design and the methods used to test it. Section 4.4 presents the testing 46 and imaging results. Finally, in section 4.5 I discuss the results used to demonstrate the feasibility of 3-D rectilinear imaging using the dual-layer array. 4.2 Introduction One of the recent advances of diagnostic ultrasound is 3-D ultrasound imaging. The advantages of 3-D ultrasound over 2-D ultrasound include viewing planes that are usually inaccessible using 2-D ultrasound, volume-rendered images, more accurate volume measurements and improved detection of cystic or cancerous masses (Fenster and Downey 1996). Clinically, 3-D imaging system can improve the detection of breast lesions, reducing the need for biopsy (Moskalik et al. 1995). When a biopsy becomes necessary, 3-D ultrasound-guided breast biopsy procedures may help doctors by giving 3- D spatial location of the applied needle since the needle no longer has to be aligned within a single scan plane (Smith et al. 2001). Additionally, 3-D ultrasound has applications in the assessment of disease in the carotid artery and the evaluation of musculoskeletal injuries (Palombo et al. 1998, Leotta and Martin 2000). 3-D ultrasound imaging systems employ either a 1-D array with mechanical translation or a 2-D array. Using 1-D arrays, 3-D images are created by the physical movement of the array to acquire several 2-D scans followed by offline reconstruction. 1- D arrays have transducer elements along the azimuthal direction, making it capable of focusing the ultrasound beam in that direction only. In 2-D arrays, focusing can be done in azimuth and elevation which provides a better spatial resolution. Additionally, scanning in 2-D arrays is done electronically which makes the imaging process faster and free from 47 reconstruction errors (Fenster and Downey 1996). Several researchers have implemented 3-D ultrasound imaging using 2-D arrays (Light et al. 1998, Yen et al. 2000, Yen and Smith 2004, Austeng and Holm 2002, Oralkan et al. 2003, Savord and Solomon 2003, Seo and Yen 2007). 2-D phased arrays are useful for imaging far-field targets such as cardiac imaging as they offer a wider field of view far away from the transducer. Rectilinear 2-D arrays have a wider field of view closer to the transducer than phased 2-D arrays. Hence, they are more desirable for imaging near-field targets such as breast, carotid artery and abdomen. A fully sampled array would be ideal for 3-D rectilinear imaging. However, an array with 128 2 = 16,384 or 256 2 = 65,536 elements is not feasible at this time. One reason for that is the difficulty in fabricating an array with such a large number of elements. Also, the low capacitance and high electric impedance of the small array elements would result in signal loss. The biggest challenge however, is to connect thousands of elements to the imaging system which cannot be done using coaxial cables. To control a 256 × 256 element array, as many as 65,536 channels and tens of thousands of coaxial cables are needed. One well established technique to reduce the number of interconnections is to design a sparse array with carefully chosen geometries where a small portion of the elements is used (Yen et al. 2000, Yen and Smith 2004). However, due to the extreme sparseness of these arrays where the number of elements greatly exceeds the number of system channels, some clutter is unavoidable. Recent commercial systems use integrated electronics within the transducer handle to connect a 3000-element 2-D phased array to a standard ultrasound system (Savord and Solomon 2003). Application specific integrated 48 circuits (ASICs) within the handle apply preliminary beamforming to funnel down the acoustic signals to 128 system channels. The second stage of beamformation is done by a traditional 128 channel system digital beamformer providing dynamic focusing. Capacitive micromachined ultrasonic transducers (cMUTs) offer a promise to overcome the difficulties in 2-D array fabrication. They are fabricated using standard silicon IC technology allowing through-wafer interconnections. Oralkan et al. (2003) have demonstrated the feasibility of cMUTs 2-D arrays using a 32 × 64-element prototype. Row-column addressing is another technique used to reduce the number of interconnections between the 2-D array and the imaging system. With this technique, an effective N × N 2-D array can be developed using only N transmitters and N receivers. It adopts the Mills cross design in which a row of elements in the 2-D array is used for transmitting an ultrasound beam and a column of elements is used for receiving. The transmit beam is narrow in elevation and the receive beam is narrow in azimuth. The result is a narrow pulse-echo beam in both dimensions because of the multiplicative process of transmit and receive (Yen et al. 2000, Morton and Lockwood 2003, Seo and Yen 2008). Dual-layer or multilayer transducers have been proposed for other diagnostic applications. Merks et al. (2006) investigated a multilayer approach to develop a piston- like transducer for acoustic bladder volume assessment using nonlinear wave propagation. Operating in the 2 MHz range, this transducer uses a 29 mm diameter PZT piston for transmit and a PVDF film for receive. Saitoh et al. (1995) developed a dual frequency 49 probe multilayer ceramic for simultaneous B-mode and Doppler duplex imaging. Using two wafers of PZT with polarities pointing in opposite directions, a dual frequency response probe was developed. The lower frequency range would be used for Doppler and the high frequency range would be used for B-mode imaging. Similar to Saitoh’s work, Hossack proposed using a dual-layer design for harmonic imaging (Hossack et al. 2000). This design used two piezoceramic layers of equal thickness. Both layers of piezoceramic were used together as a single transducer to transmit a pulse at the fundamental frequency. Only the top layer was used for receiving the second harmonic giving increased sensitivity at the second harmonic. As a new realization of the aforementioned row-column method, a dual-layer array for rectilinear 3-D imaging was designed (Yen et al. 2009). This dual-layer design uses one piezoelectric layer for transmit and another separate piezoelectric layer for receive. The P[VDF-TrFE] receive layer is closer to the target, and the PZT transmit layer is underneath the receive layer. Each layer is an elongated 1-D array with the transmit and receive elements oriented perpendicular to each other. The 4 × 4 cm prototype array has a center frequency of 4.8 MHz, 80% -6 dB fractional bandwidth and 256 PZT elements and 256 P[VDF-TrFE] elements. In this chapter, the testing results and 3-D ultrasound images acquired using a dual- layer array are presented (Yen et al. 2009). 3-D images of an 8 mm diameter anechoic cyst phantom and five pairs of nylon wires embedded in a clear gelatin phantom were acquired. I display the azimuth and elevation B-scans as well as the C-scan. The pair of 50 wires with 0.5 mm axial separation is discernible in the azimuthal and elevational B- scans. The theoretical lateral beamwidth was 0.52 mm compared to measured beamwidths of 0.65 mm and 0.67 mm in azimuth and elevation respectively. 4.3 Methods 4.3.1 3-D Rectilinear Scanning For illustrative purposes, Figure 4.1 is a simplified schematic of the rectilinear 3-D scanning process using a dual-layer design with only 8 elements in each layer. The transmit layer contains a 1-D linear array with elements along the azimuth direction. This transmit array performs beamforming, or focusing, in the azimuth direction using the gray subaperture elements (Figure 4.1A). In receive, a second layer contains a 1-D linear array with elements oriented perpendicular with respect to the transmit array. This receive layer is located directly in front of the transmit layer. This allows the receive layer to perform beamforming in the elevation direction using the elements shaded in gray (Figure 4.1B). By moving the locations of transmit and receive subapertures in azimuth and elevation respectively, a rectilinear volume can be scanned. This design can be viewed as an array with multiple Mills cross arrays for 3-D imaging (Yen and Smith 2002, Seo and Yen 2008). 51 Figure 4.1: 3-D scanning process of the dual-layer transducer array in (a) transmit and (b) in receive. Shaded elements indicate the active subaperture (Yen et al. 2009). To evaluate the theoretical imaging performance, simulated on-axis beamplots were acquired using Field II (Jensen and Svendsen 1992). The transmit aperture is a 1-D array with an azimuthal element pitch of one wavelength, or 0.15 mm, and an elevational height of 128 wavelengths, or 38.4 mm. The receive aperture has an elevational element pitch of 0.15 mm and an azimuthal length of 38.4 mm. A Gaussian pulse with a center frequency of 5 MHz and 50% -6 dB fractional bandwidth was used. For the beamplot, a 128-element subaperture was used in both transmit and receive and focused on-axis to (x,y,z) = (0,0,30) mm (Figure 4.2). The -6 dB and -20 dB beamwidths are 0.55 mm and 2.39 mm respectively. The highest clutter levels, around -30 to -40 dB, are seen along the azimuth and elevation axes. The clutter levels drop off dramatically in regions away from the principal azimuth and elevation axes. 52 Figure 4.2: Simulated on-axis beamplots of the dual-layer transducer with focus (x,y,z) = (0,0,30) mm: a) 3-D beamplot, b) Contour plot with lines at -10, -20, -30, -40, and -50 dB, c) azimuthal beamplot, and d) elevational beamplot (Yen et al. 2009). Figure 4.3 shows simulated off-axis beamplots when the focus is located at (x,y,z) = (15,15,30) mm. The -6 and -20 dB beamwidths are 0.97 and 4.01 mm respectively. Similar to the on-axis case, the main sources of clutter lie parallel to the azimuthal and elevational axes. 53 Figure 4.3: Simulated off-axis beamplots of the dual-layer transducer with focus (x,y,z) = (15,15,30) mm: a) 3-D beamplot, b) Contour plot with lines at -10, -20, -30, -40, and - 50 dB, c) azimuthal beamplot, and d) elevational beamplot (Yen et al. 2009). 4.3.2 Data Acquisition The dual-layer transducer array was interfaced with the Sonix RP ultrasound system (Ultrasonix, Vancouver, Canada) using a custom printed circuit board to acquire data. Appendix A describes the setup used to interface the array with ultrasound system. This 54 ultrasound system allows the researcher to control imaging parameters such as transmit aperture size, transmit frequency, receive aperture, filtering, and time-gain compensation. In these experiments, one PZT element was connected to one channel of the Sonix system. One transmit element was excited at a time using a two-cycle, 5 MHz transmit pulse. Sixty-four copolymer elements were each connected to individual system channels configured to operate in receive mode only. With a 40 MHz sampling frequency, data from each receive channel was collected 100 times and averaged to minimize effects of random noise. A different set of 64 receive elements was used until data from all 256 receive elements were collected. This process is repeated until all transmit and receive element combinations were acquired. 4.3.3 Beamforming, Signal Processing, and Display The acquired data was then imported into Matlab (Mathworks, Natick, MA) for offline 3-D delay-and-sum beamforming, signal processing, and image display. After averaging, dynamic transmit (azimuth) and receive (elevation) focusing was done with 0.5 mm increments with a constant subaperture size of 128 elements, or 18.56 mm. A 3- D volume was generated by selecting the appropriate transmit subapertures in azimuth and receive subapertures in elevation to focus a beam directly ahead. Receive beamforming was applied first generating 2-D slices of receive beamformed data. After that appropriate delays were applied to these slices to beamform the data in transmit. The rectilinear volume contained 255 × 255 = 65,025 image lines with a line spacing of 145 μm in both lateral directions. The dimensions of the acquired volume were 37 × 37 × 45 55 mm (azimuth × elevation × axial). Beamformed RF data was filtered with a 64-tap bandpass filter with frequency range 3.75 - 6.25 MHz. After 3-D beamforming, envelope detection was done using the Hilbert transform. Images were then log-compressed and displayed with a dynamic range of 20 to 30 dB. Azimuth and elevation B-scans are displayed along with C-scans which are parallel to the transducer face. 3-D volumes of home-made 70 × 70 × 70 mm tissue mimicking gel-based phantoms were acquired. The first phantom was made of clear gel and contained five pairs of nylon wire targets with axial separation of 0.5, 1, 2, 3, and 4 mm. The bottom wire in each pair was laterally shifted by 1 mm with respect to the top wire. The diameter of the nylon wire was 400 μm. The second phantom was made of gel with added graphite powder to provide scattering. It had an 8 mm diameter cylindrical anechoic cyst located at a depth of 27 mm from the transducer face. The ingredients and quantities of the phantoms materials shown in Table 4.1 are based on recipes given in the literature for evaluating strain imaging techniques (Hall et al. 1997). For each phantom, two rectilinear volumes were acquired: one with the short axis of the target in the azimuth direction and one with the short axis of the target in the elevation direction. 56 Ingredients Background Inclusion DI water 400 g 50.00 g N-propanol 36.70 g 4.60 g Formaldehyde 0.24 g 0.030 g Gelatin 275 Bloom 24.00 g 3.00 g Graphite 3.89 g (cyst phantom) 0 g (wires phantom) 0 g (cyst phantom) Table 4.1: Recipes for the background and inclusion materials in the gel phantoms used in the experiments. 4.4 Results Figure 4.4A-C show the azimuth B-scan, elevation B-scan, and C-scan respectively when the short axis of the wires is in the azimuth direction. All images are log- compressed and shown on a 20 dB dynamic range. The elevation B-scan (Figure 4.4B) shows the pair of wires with 0.5 mm axial separation. The two wires are discernible. The C-scan, taken at a depth of 35 mm, is parallel to the transducer face. Here, one can also see the presence of sidelobes alongside the wires. Figure 4.4D-F show the axial wire target phantom with the short axis of the wires in the elevation direction. The pair of wires with 0.5 mm axial separation is discernible in the azimuth B-scan while the short- axis view is shown in Figure 4.4E. Figure 4.4F shows the C-scan where sidelobes are again present. 57 Figure 4.4: Experimental axial wire target images with short-axis in azimuth (A-C) and short axis in elevation (D-F). All images are log-compressed and shown with 20 dB dynamic range. Figure 4.5 shows the lateral wire target responses in azimuth (Figure 4.5A) and elevation (Figure 4.5B). In both cases, the wire closest to the transducer was used. The - 6 dB beamwidth in azimuth was 0.65 mm and 0.67 mm in elevation compared to a theoretical beamwidth of 0.52 mm in both directions. In both cases, there is a sidelobe above -15 dB and some clutter below -20 dB. 58 Figure 4.5: Azimuthal and elevational lateral wire target responses. Figure 4.6 contains images of the 8 mm diameter cyst phantom. Figure 4.6A-C show two perpendicular B-scans and a C-scan with the short axis of the cyst in azimuth and Figure 4.6D-F show two perpendicular B-scans and a C-scan with the short axis of the cyst in elevation. All images in Figure 4.6 are log-compressed and are shown with 30 dB dynamic range. Figure 4.6A shows the cyst in cross-section. The cyst is not perfectly circular because of mechanical compression of the phantom to prevent motion during the data acquisition process. In the elevational B-scan and C-scan, the cylindrical cyst appears as a rectangle. Figure 4.6D-F show the cyst with short axis in elevation. Although some clutter is present, the cyst is visible in all images. 59 Figure 4.6: Experimental cyst images with the cyst short-axis in azimuth (A-C) and the cyst short axis in elevation (D-F). All images are log-compressed and shown with 30 dB dynamic range. 4.5 Discussion I have presented the 3-D images and impedance measurements of a prototype dual- layer transducer array using PZT/P[VDF-TrFE] for transmit and receive respectively. The experimental results indicate the feasibility of 3-D imaging using a dual-layer transducer array with reduced fabrication complexity and a decreased number of channels compared to a fully sampled 2-D array of comparable size. The cyst was visible in all section slices in the 3-D images. The pair of wires with 0.5 mm axial separation is discernible in azimuth and elevation. Sidelobes from the wire targets and clutter in the anechoic cyst regions are present, which may be due to the variability of element-to-element 60 performance in terms of sensitivity and bandwidth. Using a single wire from the wire targets phantom, the measured lateral beamwidths were 0.65 mm and 0.67 mm in azimuth and elevation respectively compared to the theoretical value of 0.52 mm. This discrepancy may also be due to element-to-element non-uniformity. 61 CHAPTER 5 3-D SPATIAL COMPOUNDING USING A ROW-COLUMN ARRAY 5.1 Chapter Overview In my continuing effort to improve lesion detection using 3-D strain imaging I investigated the use of the row-column array introduced in this chapter to acquire 3-D strain images. The row-column array is conceptually similar to the dual-layer array. After acquiring several 3-D strain images, a high level of noise was noticed in parts of the 3-D strain images while other parts were virtually noise free. Section 5.2 investigates the primary sources of this noise. This high level of noise prevented employing this array for investigations of decorrelation noise reduction in strain images as it would obscure performance evaluations. Consistent with the focus of lesion enhancement using 3-D ultrasound, 3-D spatial compounding applied to 3-D B-mode imaging using the row-column array was investigated next. Section 5.3 is an introduction about 2-D and 3-D spatial compounding. Section 5.4 describes the methods used to demonstrate the feasibility of 3-D spatial compounding using the row-column array including: simulations, tissue mimicking phantom fabrication, and data acquisition and beamforming techniques. In section 5.5 I present the results of 3-D spatial compounding and in section 5.6 I discuss the benefits and drawbacks of this method when compared to conventional 3-D ultrasound. 62 5.2 Strain Imaging Using the Row-column 2-D Array After noticing similar noise patters in all 3-D strain images acquired using this array, I decided to investigate the main sources of this noise. It is known that coherence functions indicate the accuracy of TDE (Lacefield and Waag 2002). Generalized coherence factor (GCF) is a coherence function originally developed to minimize the effects of phase aberration (Li and Li 2000). Figure 5.1 shows a GCF map and a slice of a 3-D strain image of a 10 mm stiffer cylindrical inclusion phantom that were generated at the same location. The big jump from high GCF values on the left to the low GCF on the right is mainly due to element non-uniformity. It can be noticed that the high GCF values region corresponds to the virtually noise free region in the strain image. Also, the low GCF values region corresponds to the strain image region with high level of noise. This suggests that the low coherence between adjacent elements signals is the main reason for the noise in the strain images acquired using the row-column array. (a) 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 (b) 0.01 0.02 0.03 0.04 Figure 5.1: A GCF map and a slice of a 3-D strain image of a 10 mm cylindrical inclusion phantom that were generated at the same location. 63 5.3 Introduction Speckle is a common artifact in coherent imaging modalities (including ultrasound) that degrades target contrast against the surrounding background. It rises from constructive and destructive interference of echoes backscattered from scatterers within the same resolution volume, giving ultrasound images a granular appearance. Several methods have been proposed to reduce speckle noise including filtering of B-mode images, frequency compounding, and spatial compounding (Aysal and Barner 2007, Burckhardt 1978, Trahey and Smith et al. 1986, Trahey and Allison et al. 1986, Li and O’Donnell 1994, Claudon et al. 2002). Compounding involves averaging images of the same target that have uncorrelated speckle patterns. In spatial compounding, this is achieved by viewing the target from different angles using steering or aperture translation techniques (Claudon et al. 2002, Li and O’Donnell 1994, Trahey and Smith et al. 1986, Dahl et al. 2005). Besides speckle noise reduction, spatial compounding offers additional advantages such as reduced acoustic shadowing and suppressed refraction artifacts (Claudon et al. 2002). Clinically, real-time spatial compounding has shown usefulness in the evaluation of breast lesions, peripheral blood vessels, and musculoskeletal injuries (Kwak and You 2004 et al. 2005, Seo et al. 2002, Lin et al. 2002, Entrekin et al. 2001). Figure 5.2 compares a conventional ultrasound image (Figure 5.2a) and a compound image (Figure 5.2b) of Fibroadenoma in a 44-year-old woman (Kwak and You 2004). These images were acquired using an HDI 5000 SonoCT system and a broad-bandwidth L12-5 linear scan head (Philips Medical Systems, Bothell, WA). The compound image 64 shows an improved delineation of the lesion and better appreciation of the internal architecture. (a) (b) Figure 5.2: Ultrasound images of Fibroadenoma in a 44-year-old woman using (a) conventional ultrasound and (b) spatial compounding (Kwak and You 2004). A number of studies investigated the decorrelation of speckle with lateral translation of the transducer to find the optimum displacement for 2-D compounding (Burckhardt 1978, Wagner et al. 1988, Trahey and Smith et al. 1986). The degree of statistical independence between the averaged images governs the success of spatial compounding. Speckle SNR, defined as the ratio of the mean to standard deviation of the detected data, increases by the square root of the number of independent compounded images. If the compounded images are partially correlated, the number of independent images N can be estimated using the following equation (Trahey and Smith et al. 1986, Dahl et al. 2005): 65 ∑ ∑ = + = + = n X n X Y Y X n n N 1 1 , 2 2 ρ (Eq. 5.1) where n is the number of compounded images, ρ X,Y is the correlation coefficient between images X and Y. The correlation coefficient of two m × n pixel image regions, X and Y can be calculated using (Trahey and Smith et al. 1986): ∑∑ ∑∑ ∑∑ = = = = = = − − − − = m o n p p o m o n p p o m o n p p o p o Y X Y Y X X Y Y X X 1 1 2 , 1 1 2 , 1 1 , , , ) ( ) ( ) )( ( ρ (Eq. 5.2) where X o,p is the mean envelope-detected echo magnitude value of image X at location o,p, and X and Y are the mean pixel values of image region X and Y, respectively. 3-D spatial compounding has the advantages of providing a greater number of independent images and reducing out-of-plane refraction artifacts (Krucker et al. 2000). To date, most 3-D spatial compounding research reported in the literature has used a 1-D linear or phased array with mechanical translation and position tracking (Moskalik et al. 1994, Rohling et al. 1996, Rohling et al. 1998, Leotta and Martin 2000, Lacefield et al. 2004). 1-D arrays lack the capability of electronic focusing in elevation which degrades the image quality away from the fixed focus. Additionally, mechanical translation with position tracking can introduce registration errors (Krucker et al. 2000). Krucker et al. used a 1.5-D transducer array and compounded 5 partially correlated volumes. They used mechanical steering in elevation and applied nonrigid registration to align the steered 66 volumes. The reported CNR improvement for -12dB spherical cysts was 1.97 (Krucker et al. 2000). In this chapter, I describe a 3-D ultrasound imaging system using a prototype 2-D array with row column addressing used for 3-D spatial compounding. In this work, I use subaperture translation to compound nine uncorrelated volumes. Using this technique eliminates the need for registration. I present results of several simulations to demonstrate the feasibility of this idea. Finally, I present images of gelatin phantoms with a cylindrical -11 dB cyst and a cylindrical +11 dB tumor and quantify the improvement in CNR and speckle SNR for the compounded images over standard B-mode images. CNR improvements of 2.38 and 2.51 were achieved for the cyst and tumor 3-D images respectively. The speckle SNR improvements were 2.65 for the cyst image and 2.70 for the tumor image. Using five pairs of nylon wires embedded in a clear gelatin phantom, the decrease in lateral resolution caused by this spatial compounding technique was found to be a factor of 2.04. 5.4 Methods 5.4.1 Simulations To test the feasibility of 3-D spatial compounding using a row-column 2-D array, I carried out several simulations using Field (Jensen and Svendsen 1992). In these simulations, a 1-D array with plane wave transmission was used with the imaging parameters listed in Table 5.1. These parameters were chosen to model one way 67 beamforming in the row-column 2-D array used in the experiments to provide a “gold standard” for comparison. Dynamic receive beamforming every 1 mm was applied in all cases with an expanding aperture to keep the F-number ≥ 2. Parameter Value Total number of elements 128 Number of elements in subaperture 32 Sampling frequency 200 MHz Center frequency 5 MHz Bandwidth 50% Lateral element width 283 μm Kerf 25 μm Pitch 308 μm Elevation element height 5 mm System sound speed 1540 m/s Transmission Plane wave Table 5.1: Simulations parameters. The simulations that were carried out are as follows: A. Speckle correlation Because of limited computing resources for processing 3-D experimental data, the effect of aperture translation on speckle correlation using plane wave excitation was investigated with the following simulation. Phantoms with the size of 10 mm × 40 mm × 50 mm (elevation × azimuth × axial) filled with speckle generating scatterers were simulated. Scatterers were generated with Gaussian distributed amplitudes at random 68 locations with 10 scatterers per resolution cell (Oosterveld et al. 1985, Tuthill et al. 1988). After generating element RF data using Field II, several pairs of B-mode images were beamformed independently. Scan lines in these image pairs were generated using two walking apertures that have 0.25 - 2 fractional aperture separation between the centers of the apertures at all time. Eight experiments with a 0.25 fractional aperture separation step were carried out. As an example, Figure 5.3 shows the apertures used to create the first scan line with 0.5 and 1.5 fractional aperture separation. The apertures were walked to create on-axis scan lines whenever it is possible. Figure 5.3: Apertures used to create the first scan line that have (a) 0.5 and (b) 1.5 fractional aperture separation. The apertures were walked to create on-axis scan lines whenever it is possible. Both apertures are indicated by braces and shading. 69 B. Speckle reduction In this simulation, speckle reduction was investigated using three adjacent 32-element apertures to generate three partially correlated images. Figure 5.4 shows the three walking apertures. Using these apertures, element RF data for pure speckle simulated phantoms was beamformed independently. Next, the data was filtered using a 64-tap finite impulse response (FIR) bandpass filter with a frequency range of 3.75 - 6.25 MHz and 200 MHz sampling frequency. The resulting three sets of beamformed RF data were then envelope detected and averaged. The speckle SNR of the resulting compounded image was compared to that of an image generated using standard beamforming using a 64 element aperture applied to the same RF data. The following equation was used to evaluate speckle SNR (Trahey and Smith et al. 1986, Wagner et al. 1983): σ μ = SNR (Eq. 5.3) where μ is the mean and σ is the standard deviation of the detected data for a speckle region that contains no resolvable structures. Figure 5.4: The three apertures used to create the first scan line for the speckle reduction simulation. The apertures were walked to create on-axis scan lines whenever it is possible. All apertures are indicated by braces and shading. 70 C. Contrast Enhancement After creating -11 dB cyst and +11 dB tumor gel based phantoms, I carried out the following simulations to provide a “gold standard” for the experimental results. The simulated phantoms included -11 dB and +11 dB cylindrical lesions with 10 mm diameter. The lesions were located at a depth of 35 mm. The same steps of beamforming, filtering and envelope detection as described in the speckle reduction simulation were used here. The CNR of the image compounded from three images was compared to that of an image generated using standard beamforming. The following equation was used to evaluate CNR (Li and O’Donnell 1994): b t 2 b t s var s var ) s s ( CNR + − = (Eq. 5.1) where s and var s denote mean and variance, respectively. The subscripts t and b represent the target and background, respectively. 5.4.2 System Description A prototype 2-D array with row-column addressing was used in the 3-D imaging experiments. This array has a 5.3 MHz center frequency, 53% -6 dB fractional bandwidth, 256 × 256 = 65,536 elements with λ/2 = 150 μm pitch and row-column addressing (Seo and Yen 2008). The design of this 2-D array utilizes a two-layer electrode pattern where the bottom layer consists of a series of vertical electrodes (Figure 5.5A) and the top layer consists of a series of horizontal electrodes (Figure 5.5B). In 71 transmit, the vertical electrodes serve as the “ground” and the top electrodes serve as the “transmitters”. Figure 5.5: Row-column addressing scheme. The transmit side consists of 8 horizontal electrodes. The receive side consists of 8 vertical electrodes which serve as ground in transmit. Transmit beamforming is done in the vertical or elevational direction while receive beamforming is done horizontally or azimuthally (Seo and Yen 2008). In this example, transmit channel D indicated by the arrow to the right of channel D in Figure 5.5B is excited. This row of elements, shown in the gray shading in Figure 5.5C, then emits a cylindrical wavefront into the field. In elevation, the wavefront appears omnidirectional since the aperture behaves like a single small element. In the azimuth direction, the emitted beam is a planar wavefront because all elements fire simultaneously, and the aperture behaves as a single long element. For receive mode, receive channels A-H are active and the desired receive column is selected (Figure 5.5D). 72 In receive mode, the individual elements along one column (gray shading) is used to record the echoes (Figure 5.5F). With this design, transmit beamforming can be done in the vertical or elevational direction while receive beamforming can be done horizontally or azimuthally. A schematic illustrating this beamforming method is shown in Figure 5.6. Multiple rows can be used for elevational beamforming in transmit and multiple columns can be used for azimuth beamforming in receive. By stepping transmit subapertures across the array with multiple receive beams within the transmit beam, a 3-D rectilinear volume can be acquired (Seo and Yen 2008). Figure 5.6: Transmit and Receive Beamforming for 3-D rectilinear imaging (Seo and Yen 2008). The 2-D array was interfaced with the Ultrasonix (Richmond, BC, Canada) Sonix RP ultrasound system using a custom printed circuit board. Appendix A describes the setup used to interface the array with ultrasound system. This system has a 40 MHz sampling frequency, 128 channels, and 32 analog to digital converters. It allows the user to acquire 73 raw radio frequency (RF) data and gives the user control over transmit aperture size, transmitted power, transmitted frequency, receive aperture size, filtering, and time-gain compensation. Using a synthetic aperture approach, two rows were excited at a time and signals from all 256 columns were acquired. Exciting two λ/2 pitch elements simultaneously makes the effective transmit pitch equal to λ. Excitation was done in this manner because transmit elements were accessed manually, which makes exciting the 256 transmit elements individually a labor intensive process. A two-cycle, 5 MHz transmit pulse was used. Two channels from the Sonix machine were used for transmission and manually multiplexed to access all 128 pairs of transmitters. Sixty-four system channels were used for receiving. A different set of 64 receive elements was connected to the ultrasound system until RF data from all 256 columns were collected. Data was sampled at 40 MHz, collected 100 times, and averaged to suppress random noise. Averaged elements RF data were filtered using a 64-tap FIR bandpass filter with frequency range of 3.75 - 6.25 MHz. Off-line 3-D beamforming was applied to the RF element data using Matlab (Mathworks Inc., Natick, MA). Three 32-element adjacent transmit apertures with 1λ pitch and three 64-element adjacent receive apertures with λ/2 pitch were used to generate the spatially compounded 3-D image (Figure 5.7). Using delay-and-sum beamforming, the nine combinations of these apertures were used to create nine decorrelated volumes. For the contrast phantoms images, dynamic focusing every 1 mm in transmit and receive was used along with an expanding aperture to keep the F number 74 at 2. The resulting B-mode images were 38.4 × 38.4 × 41.4 mm with 129 × 129 = 16,641 scan lines and 0.3 mm line spacing. The wires phantom image was beamformed using dynamic focusing every 0.5 mm in transmit and receive with 257 × 257 scan lines and 0.15 mm line spacing. This was done in order to get a more accurate estimation of the system resolution with and without spatial compounding, Figure 5.7: The three apertures used to create the first scan line for (a) the transmit side and (b) the receive side of the row-column 2-D array. The apertures were walked to create on-axis scan lines whenever it is possible. All apertures are indicated by braces and shading. The resulting nine volumes were then envelope detected and averaged. To quantify the improvements of 3-D spatial compounding, standard beamforming was applied to all data sets using a 64-element transmit aperture with 1λ pitch and 128-element receive aperture with λ/2 pitch. Tx aperture #1 Elevation Azimuth (a) Transmit (b) Receive Tx aperture #2 Tx aperture #3 Rx aperture #1 Rx aperture #2 Rx aperture #3 75 5.4.3 Phantom fabrication Three 70 × 70 × 70 mm gel-based phantoms were used to test this 3-D spatial compounding technique. Two contrast phantoms were made to quantify the contrast enhancement in the compounded images. One phantom included a 10 mm cylindrical -11 dB cyst and the other phantom included a +11 dB tumor. Both inclusions were located approximately at the center. The recipes for the inclusions were determined empirically where I used the background material recipe while varying the amount of graphite. To quantify the loss in resolution, a clear gel phantom with embedded wires was made. The wires phantom included pairs of nylon wire targets with axial separations of 0.5, 1, 2, 3 and 4 mm located at the center of the phantom. The bottom wire in each pair was shifted laterally by 1 mm with respect to the top wire. The diameter of the nylon wire was 400 μm. Table 5.2 show the recipes used (Hall et al. 1997). Material Background Inclusion DI water 400 g 50.00 g N-propanol 36.70 g 4.60 g Formaldehyde 0.24 g 0.030 g Gelatin 275 Bloom 24.00 g 3.00 g Graphite 3.89 g (contrast phantoms) 0 g (wires phantom) 0.10 g (-11 dB cyst) 3.43 g (+11 dB tumor) Table 5.2: Recipes for the background and inclusion materials in the gel phantoms used in the experiments. 76 Figure 5.8 shows images of the phantoms acquired using the Ultrasonix L14-5/38 linear array with synthetic transmit and receive focusing every 1mm. The contrast ratio calculated using these images are -10.50 dB for the cyst phantom and +10.94 dB for the tumor phantom. The cyst is not perfectly at the center in Figure 5.8 because of variable mechanical compression of the phantom during the data acquisition process. Figure 5.8: B-mode images of (a) the -11 dB cyst phantom and (b) the +11 dB tumor phantom using Ultrasonix L14-5/38 linear array. Both images are log-compressed and shown with 30 dB dynamic range. 5.5 Results 5.5.1 Simulations A- Speckle correlation Figure 5.9 shows the mean speckle correlation simulation results versus fractional aperture translation using 25 realizations. It can be noticed that decorrelation rate falls 77 considerably for fractional aperture translations greater than one. Two fractional aperture translations were used in the experimental and simulation beamforming: 1 and 2. Referring to Figure 5.4, Apertures #1 and #2 and apertures #2 and #3 have a fractional aperture translation of 1. Apertures #1 and #3 have a fractional aperture translation of 2. The fractional aperture translation of 1 has a mean correlation of 0.062 with standard deviation = 0.014. The fractional aperture translation of 2 has a mean correlation of 0.037 with standard deviation = 0.016. Figure 5.9: Speckle decorrelation curve for 0 to 2 fractional aperture translation with step size = 0.25 using Field II simulation. The error bars indicate ± 2 standard deviations. B- Speckle reduction Twenty five realizations of simulated element RF data for pure speckle were beamformed using the three apertures described above (Figure 5.4) and using standard beamforming with a 64 element aperture. The mean speckle SNR improvement of the 78 compounded image over the speckle SNR of the three images used for compounding is 1.63, closely matching the mean theoretical speckle SNR improvement value of 1.65 predicted by equation 1. The mean speckle SNR improvement of the compounded image over the speckle SNR of the standard beamforming image is 1.66. C- Contrast enhancement Figure 5.10 shows one realization of a simulated -11 dB cyst phantom. Figure 5.10a was created using standard beamforming with a 64 element aperture. Figure 5.10b was created using spatial compounding of three images obtained using the apertures described above (Figure 5.4). Both images are log-compressed and shown with 30 dB dynamic range. Using 25 realizations, the mean CNR improvement is 1.62 with a standard deviation of 0.11. Figure 5.10: Images of one of the realizations of the -11 dB cyst phantom simulated in Field using: (a) Standard beamforming. (b) Spatial compounding of three images. Both images are log-compressed and shown with 30 dB dynamic range. 79 Figure 5.11 shows one realization of a simulated +11 dB tumor phantom. Figure 5.11a was created using standard beamforming with a 64 element aperture. Figure 5.11b was created using spatial compounding of three images obtained using the apertures described above (Figure 5.4). Both images are log-compressed and shown with 30 dB dynamic range. Using 25 realizations, the mean CNR improvement is 1.66 with a standard deviation of 0.07. The CNR values were calculated by choosing equal sized regions of the interior of the lesion and the background located at the same depth. A CNR kernel size of 17 × 1400 pixels (lateral× axial) was used. Figure 5.11: Images of one of the realizations of the +11 dB tumor phantom simulated in Field using: (a) Standard beamforming. (b) Spatial compounding of three images. Both images are log-compressed and shown with 30 dB dynamic range. 80 5.5.2 Experiments Two sets of 3-D RF data of the cylindrical -11 dB cyst and +11 dB tumor phantoms were acquired using the 2-D array. 3-D CNR values were calculated by choosing equal sized volumes of the interior of the lesion and the background located at the same depth. A CNR kernel size of 110 × 13 × 250 voxels (elevation × azimuth × axial) was used. The CNR of the uncompounded image (standard beamforming) and the image compounded from nine volumes for the cyst phantom are 1.16 and 2.76 respectively. For the tumor phantom, the CNR of the uncompounded image is 1.03 and the CNR of the image compounded from nine volumes is 2.59. This gives a CNR improvement of 2.38 for the cyst image and 2.51 for the tumor image. Figure 5.12 shows the monotonic increase of mean CNR values as the number of compounded volumes increases. These values were obtained experimentally using Equation 3 and theoretically using Equation 1. For number of compounded volumes N= 9, CNR was calculated for the only one possible combination. For number of compounded volumes N= 1 and 8, CNR values were calculated for the only nine combinations possible. For number of compounded volumes N= 2 to 7, CNR values were calculated for nine randomly selected combinations. 81 Figure 5.12: Mean of CNR values for 1 to 9 compounded volumes for: (a) -11 dB cyst phantom (b) +11 dB tumor phantom. Solid lines show experimental CNR values and dashed lines show theoretical CNR values. To evaluate speckle reduction, 25 independent 3-D blocks with a size of 100 resolution cells were chosen from the background of the nine volumes at a depth of around 35 mm. Table 5.3 shows the mean cross-correlation and mean speckle SNR values along with the theoretical and experimental speckle SNR improvement values for the cyst and tumor 3-D images. 82 -11 dB cyst +11 dB tumor Mean cross-correlation values for the nine volumes 0.024 0.022 Mean SNR value for the compounded image 4.99 5.03 Theoretical SNR improvement 2.92 2.81 SNR improvement over the nine volumes 2.67 2.69 SNR improvement over the uncompounded image 2.65 2.70 Table 5.3: Mean cross-correlation, mean speckle SNR values and the theoretical and experimental speckle SNR improvement values for the cyst and tumor 3-D images. Figure 5.13 shows the monotonic increase of mean speckle SNR values as the number of compounded volumes increases. These values were obtained experimentally using Equation 3 and theoretically using Equation 1. For number of compounded volumes N= 9, CNR was calculated for the only one possible combination. For number of compounded volumes N= 1 and 8, SNR values were calculated for the only nine combinations possible. For number of compounded volumes N= 2 to 7, SNR values were calculated for nine randomly selected combinations. 83 Figure 5.13: Mean of speckle SNR values for 1 to 9 compounded volumes for: (a) -11 dB cyst phantom (b) +11 dB tumor phantom. Solid lines show experimental speckle SNR values and dashed lines show theoretical speckle SNR values. Figure 5.14 A-C show the azimuth B-scan, elevation B-scan, and C-scan respectively of the cyst phantom generated using standard beamforming. Figure 5.14 D-F show the image compounded from nine volumes for the cyst phantom. The C-scans and the elevation B-scans are taken approximately at the widest cross section of the cyst and the azimuth B-scans are taken approximately at the center of the phantom. All images are log-compressed and shown on a 30 dB dynamic range. 84 Figure 5.14: Experimental cyst 3-D images with the -11 dB cyst short-axis in azimuth using: (A-C) Standard 3-D beamforming. (D-F) Compounding of nine volumes. All images are log-compressed and shown with 30 dB dynamic range. Figure 5.15 shows isosurface renderings of the cyst standard beamforming 3-D image and the 3-D image compounded from nine volumes. In each case, the isosurface level was adjusted for optimal display of the cyst. 85 Figure 5.15: Isosurface renderings of the -11 dB cyst 3-D image generated using: (a) standard beamforming and (b) compounding of nine volumes. In each case, the isosurface level was adjusted for optimal display of the cyst. Figure 5.16 A-C show the azimuth B-scan, elevation B-scan, and C-scan respectively of the tumor phantom generated using standard beamforming. Figure 5.16 D-F show the compounded image of the tumor phantom. The C-scans and the elevation B-scans are taken approximately at the widest cross section of the tumor and the azimuth B-scans are taken approximately at the center of the phantom. All images are log-compressed and shown on a 30 dB dynamic range. 86 Figure 5.16: Experimental tumor 3-D images with the +11 dB tumor short-axis in azimuth using: (A-C) standard 3-D beamforming and (D-F) compounding of nine volumes. All images are log-compressed and shown with 30 dB dynamic range. Figure 5.17 shows isosurface renderings of the tumor standard beamforming 3-D image and the 3-D image compounded from nine volumes. In each case, the isosurface level was adjusted for optimal display of the tumor. 87 Figure 5.17: Isosurface rendering of the +11 dB tumor 3-D image using: (a) Standard 3- D beamforming. (b) Compounding of nine volumes. In each case, the isosurface level was adjusted for optimal display of the tumor. Figure 5.18 shows the azimuth B-scan of the wires phantom 3-D image with the short axis of the wires in the azimuth direction. Figure 5.18a was generated using standard beamforming, and Figure 5.18b was generated by compounding nine volumes. Using detected data and the wire nearest the transducer, the -6 dB beamwidths of the uncompounded and compounded images were 0.70 mm and 1.43 mm respectively. This gives a reduction in the lateral resolution by a factor of 2.04. 88 Figure 5.18: A short axis slice of the 3-D wires phantom image generated using: (a) standard beamforming and (b) Compounding of nine volumes. Both images are log- compressed and shown with 20 dB dynamic range. 5.6 Discussion In this chapter, I presented initial 3-D spatial compounding results using a prototype 256 × 256 2-D array with row-column addressing, 40 × 40 mm aperture, and 5 MHz center frequency. The 2-D simulations that model one way beamforming in the 2-D array indicate the feasibility of 3-D spatial compounding using a row-column array. Using the experimental setup, I acquired several 3-D images to identify the benefits and the drawbacks of 3-D spatial compounding with aperture translation. Using 2-D simulations, cross-correlation decrease with increased aperture separation was evaluated. It was found that decorrelation rate drops significantly for aperture separations greater than one aperture size. Accordingly, a fractional aperture separation of one between adjacent subapertures was used in simulation and experimental 89 beamforming. Contrast enhancement and speckle SNR improvement 2-D simulations results show good agreement with the theoretical expectations. Using different subsets of uncorrelated volumes for compounding shows the expected monotonic increase of mean CNR and speckle SNR values as the number of compounded volumes increases. The results show good agreement with values predicted by theory (Figure 5.12, 5.13). Compounding nine volumes shows a significant improvement in CNR and speckle SNR for the contrast phantoms 3-D images. Qualitatively, it is easier to identify the inclusions in the compounded images than in the uncompounded images (Figure 5.14, 5.16). Also, the isosurface renderings of the -11 dB cyst and the + 11 dB tumor show significant clutter reduction in the compounded images compared to the uncompounded ones (Figure 5.15, 5.17). In Figure 5.16E, an inclined “ghost” artifact at the lateral edges of the image above the tumor can be seen. This is mainly caused by grating lobes that show up in the transmit direction of the 2-D array (elevational direction in Figure 5.16E). Since the transmit side of the 2-D array has a pitch of 1λ, grating lobes for this side are theoretically located at 90°. Aperture number 3 (Figure 5.4) creates the far left scan lines at large steering angles (maximum of 45° for a depth of 35 mm). Hence, a “ghost” of the tumor appears at the left edge of the elevational slices. This also applies to the far right scan lines created using aperture number 1 (Figure 5.4). This effect was confirmed by finding the presence of this artifact in the individual compounded volumes. Using λ/2 pitch on the transmit side would minimize this problem. However, this requires manual excitation of the 256 90 transmit elements individually which would make data acquisition a lengthy process. The loss in lateral resolution was found to closely match the expected factor of two. This is caused by the fact that the aperture size in the compounded image is half of that in the uncompounded image. 91 CHAPTER 6 SUMMARY AND FUTURE WORK 6.1 Summary The feasibility of 3-D strain imaging has been demonstrated. A 3-D strain imaging system using a 2-D ultrasonic sparse array was designed and built. The system was used to acquire 3-D strain images of six tissue mimicking phantoms with various size and shape stiffer inclusions. Several signal processing combinations were carried out to isolate and identify the benefits of 3-D beamforming with elevational focusing and 3-D TDE for strain imaging. The initial results suggest that a large CNR improvement can be achieved by switching from 2-D TDE to 3-D TDE. Additionally, an improvement in CNR and SNR was noticed in all cases when switching from 2-D BF, 3-D TDE to 3-D BF, 3-D TDE. This indicates that that 3-D beamforming with elevational focusing can improve the performance of strain imaging with 3-D TDE by increasing elevational resolution. 3-D strain imaging with full 3-D data processing have shown an average CNR improvement of around 63% over 2-D strain imaging. The 3-D images and impedance measurements of a prototype dual-layer transducer array were presented. The transducer uses two layers of PZT/P[VDF-TrFE] for transmit and receive respectively. The experimental results indicate the feasibility of 3-D imaging using a dual-layer transducer array with reduced fabrication complexity and a decreased number of channels compared to a fully sampled 2-D array of comparable size. The impedance measurements showed good agreement between simulation and experiment. 92 3-D images of an anechoic cyst tissue mimicking phantom and multi wires phantom were acquired. The cyst was visible in all section slices. The pair of wires with 0.5 mm axial separation is discernible in azimuth and elevation. Sidelobes from the wire targets and clutter in the anechoic cyst regions are present, which may be due to the variability of element-to-element performance in terms of sensitivity and bandwidth. Lastly, I presented 3-D spatial compounding results using a prototype row-column 2- D array. 2-D simulations and 3-D experiments indicate the feasibility of 3-D spatial compounding with aperture translation using a row-column array. 3-D images of two contrast phantoms (-11 dB cyst and the + 11 dB tumor) and a wires phantom were acquired. 3-D compounding results show the expected monotonic increase of mean CNR and speckle SNR values as the number of compounded volumes increases. Compounding nine volumes shows a significant improvement in CNR and speckle SNR for the contrast phantoms 3-D images. The average improvement of the lesion CNR and speckle SNR were 2.44 and 2.42 respectively. All SNR and CNR values calculated show good agreement with values predicted by theory. Qualitatively, it is easier to identify the inclusions in the compounded images than in the uncompounded images. Also, the isosurface renderings of the -11 dB cyst and the + 11 dB tumor show significant clutter reduction in the compounded images compared to the uncompounded ones. The loss in lateral resolution was found to closely match the expected factor of two. This is caused by the fact that the aperture size in the compounded image is half of that in the uncompounded image. 93 6.2 Previous Research Objectives Previous objectives of this research included applying 3-D companding and 3-D spatial compounding using a sparse 2-D array to improve 3-D strain image quality. Due to limitations in the prototype 2-D array used, attempts to implement these ideas were not successful. 2-D companding was implemented successfully using a commercial 1-D array. The 2-D companding experimental results agree with the literature that states that the benefits of companding can mostly be seen for compression rates in the range of about 3-5% (Chaturvedi 1998). Pre- and post- compression images acquired using the prototype 2-D arrays had low correlation for compression rates higher than 2%. This resulted in more errors in time delay estimations (TDE) that companding could not fix. The sparseness of the array which gives rise to clutter is the main reason for the TDE errors when relatively low compression rates were applied (2-3%). It is well established that a key factor in the success of spatial compounding is low correlation between the compounded images (Trahey and Smith et al. 1986). Decorrelation between images increases with increasing the steering angle. The synthetic aperture approach that was adapted for acquiring data using the sparse array involved using an unfocused transmit beam with the size of 4.8 × 4.8 mm. This allows for a maximum steering angle of about ±2° at the depth of 40 mm. This resulted in relatively high correlation between the strain images generated using different steering angles from -2° to +2° (Figure 6.1). This in turn resulted in an insignificant improvement in the quality of strain imaging. This limitation can be overcome by employing a fully sampled 94 2-D array (or a more populated sparse array) with transmit focusing and steering capabilities. Another type of array that can help overcome this limitation is a 2-D array that uses a wider planar transmit beam where a wider range of receive beam steering angles can be achieved. Figure 6.1: The planar 4.8 × 4.8 mm transmit beam used in the sparse 2-D array allows for a narrow range of steering angles for the focused receive beam. In an attempt to employ a better performing transducer array for 3-D strain imaging, I investigated the dual-layer array presented in Chapter 4. I interfaced and tested the array as well as acquired and beamformed RF data to generate 3-D B-mode images. When compared to the sparse array, this array has the advantages of higher axial resolution and lateral resolution through the enhance bandwidth and transmit focusing capabilities of the transducer respectively. This gives promise of keeping TDE errors at an acceptable level when higher compression rates are applied. Additionally, a transmit beam with the size of 38.4 mm can be generated using this array. This enables implementing spatial Focused Rx beam θ=±2 ° 40 mm 4.8 mm Unfocused Tx beam 95 compounding with a wide range of steering angels with the maximum of ±26° at a depth of 40 mm. Unfortunately, the array degraded before I was able to use it to acquire strain images. Another attempt was made to employ the row-column array introduced in Chapter 5 to acquire 3-D strain images. This array has similar advantages as the dual-layer array over the sparse array. After acquiring several 3-D strain images, a high level of noise was noticed in parts of the 3-D strain images while other parts were virtually noise free. Investigating the GCF values applied to B-mode images acquired using this 2-D array indicate that element non-uniformity is the main reason for the noise in these strain images (Figure 5.1). This high level of noise prevented employing this array for investigations of decorrelation noise reduction in strain images as it would obscure performance evaluations. Table 6.1 compares the three transducers used in this research. Using P[VDF-TrFE] as a receive layer in the dual-layer array gives it a substantially higher bandwidth when compared to the two other arrays. The receive element impedance indicate that the SNR of RF data acquired using the row-column array can be significantly lower than the SNR for the other two arrays. The λ/2 pitch of the dual-layer and the row-column arrays theoretically eliminates grating lobes at any steering angle. The transmit focusing and steering capabilities of the dual-layer and the row-column arrays can improve lateral resolution when compared to the sparse array. 96 Sparse Array Dual-Layer Array Row-Column Array Aperture size 40 × 40 mm 40 × 40 mm 40 × 40 mm Center frequency 5 MHz 5 MHz 5 MHz -6 dB fractional bandwidth 45% 80% 53% Rx impedance (at 5MHz) 1.6 kΩ 1.3 kΩ 120 Ω Tx element pitch 2.4 mm 0.15 mm 0.15 mm Rx element pitch 0.6 mm 0.15 mm 0.15 mm Tx focusing and steering capabilities No Yes Yes Table 6.1 A comparison between the three transducer arrays used in this research. Although the prototypes for the dual-layer array and the row-column array had some fabrication flaws that made them not suitable for 3-D strain imaging, their designs hold promise for improving 3-D strain imaging through implementing decorrelation noise reduction techniques. Fabrication methods can be improved to enhance element uniformity in terms of both sensitivity and bandwidth. Possible fabrication issues to be addressed include achieving highly uniform bonding pressures over the entire array surface and developing fixtures to ensure good planarity as the layers are bonded together. Section 6.3 describes in more details how the row-column or the dual-layer arrays can be successfully employed for 3-D spatial compounding for 3-D strain imaging. Section 6.4 illustrates how 3-D companding can be implemented to reduce decorrelation noise in 3-D strain images. 97 6.3 3-D Spatial Compounding for 3-D Strain Imaging Strain imaging, like all imaging modalities, suffers from artifacts that degrade image quality. Random noise can give rise to such artifacts as it reduces correlation between pre- and post-compression signal causing TDE errors. The effect of random noise can be minimized using several methods including filtering and frame averaging. Another main source of artifacts is tissue compression that distorts the post-compression echo signal. After applying compression to a tissue sample in the process of creating a strain image, the tissue shrinks in the axial direction and expands in both lateral directions. This warping causes the local kernels of the post-compression image to be slightly different from those of the pre-compression image. This reduces the accuracy of TDE and introduces decorrelation noise. Several techniques have been developed to reduce decorrelation noise to improve elastographic signal-to-noise ratio (SNRe) such as temporal stretching, multicompression averaging, wavelet denoising and Spatial-angular compounding for strain imaging (Alam and Ophir 1997, Alam et al. 1998, Varghese et al. 1996, Li and Chen 2002, Techavipoo et al. 2004, Rao et al. 2006, Rao and Varghese 2008). Most of these techniques can be implemented conjunctly to suppress strain imaging artifacts. In conventional ultrasound, speckle noise may be reduced using spatial compounding. It involves averaging images of the same target that have uncorrelated speckle patterns. These images are obtained by viewing the target from different angles using steering or aperture translation techniques (Claudon et al. 2002, Li and O’Donnell 1994, Trahey and 98 Smith et al. 1986, Dahl et al. 2005). Speckle SNR increases by the square root of the number of independent compounded images. Spatial compounding for strain imaging utilizes the same principle utilized for speckle noise reduction in conventional ultrasound with spatial compounding. Techavipoo et al. (2004) have successfully implemented this technique for 2-D strain imaging by mechanically translating a 1-D phased array ultrasound transducer in azimuth. However, this technique is subject to registration errors due to transducer positioning. More recently, Rao et al. (2006) used 1-D linear array with beam steering to generate several decorrelated 2-D strain images. They reported significant improvement in CNRe and SNRe (Eq. 3.4 and Eq. 3.5 respectively). In a different work, Rao et al. (2008) have investigated the effects of aperture size, imaging frequency, and applied strain on correlation between pre- and post- compression 3-D signals in theory and simulation. Their results indicate that using 3-D motion tracking, smaller aperture size, lower applied strain, and lower imaging frequencies can minimize decorrelation artifacts. The aforementioned researches indicate the feasibility of implementing 3-D spatial compounding for 3-D strain imaging using the methods described in Chapter 5. The dual-layer array and row-column array introduced in Chapter 4 and Chapter 5 respectively have an unfocused transmit beam with a size of 38.4 mm. This enables implementing spatial compounding with beam steering at a maximum steering angle of ±26°. Another alternative is to implement spatial compounding with aperture translation as described in Chapter 5. Pre- and post compression 3-D volumes of gel/agar phantoms can be acquired synthetically. Next, three 9.6 mm adjacent transmit apertures and three 9.6 99 mm adjacent receive apertures can be used in 3-D beamforming as described in section 5.3.2. Nine pairs of pre- and post-compression volumes with 129 × 129 = 16,641 scan lines and 0.3 mm line spacing can be generated. By applying 3-D TDE on each of the nine pairs, nine strain images can be generated. To quantify the improvements, standard beamforming can be applied to all data sets using 19.2 mm transmit and receive apertures followed by 3-D TDE and 3-D strain image formation. Using a smaller aperture for 3-D spatial compounding applied to conventional ultrasound results in resolution loss. However, as Rao et al. (2008) have demonstrated, using a smaller aperture for strain imaging is expected to result in better performance in terms of SNRe and CNRe. This is due to the fact that using a smaller aperture results in a wider beam width which decreases decorrelation pre- and post-compression RF data. 6.4 3-D Companding Companding (also known as temporal stretching) is another method used to reduce decorrelation noise in strain images. In companding, the normal displacements calculation is followed by axial expansion and lateral compression of the post- compression image to restore coherence with the pre-compression image. TDE is applied to the pre- and the companded post-compression images to calculate a set of “residual” displacements. The two sets of displacements are added together and differentiated to estimate strain. A more detailed version of this method (illustrated in Figure 6.2) involves doing companding in two steps (Chaturvedi et al. 1998). In the first step, 4 × 4 × 4 = 64 global 100 displacements matrix that represents the deformations of the whole imaging volume is calculated using 3-D SAD. These displacements values can be used to estimate one strain value for each direction using regression analysis. The three strain values are next used to warp or scale the post compression to create a post global companding image that restores coherence with the pre-compression image only partially. In the second step, the local displacements between the pre-compression and the post global companding images are calculated by applying SAD to smaller overlapping 3-D kernels. The local displacements are used to shift the local 3-D kernels of the post global companding image. Since the kernels in local companding are overlapping and the displacements are spatially varying, the net effect of shifting is to locally deform the post global companding image and create the post local companding image. Residual displacements are calculated next using 1-D cross-correlation in the axial direction between pre- compression image the post local companding image. The local displacements and the residual displacements are added together and differentiated to give strain values. Lastly, the strain calculated by cross-correlation and local SAD is added to global strains to give the final strain image. 101 Figure 6.2: A flow chart summarizing strain image generation using companding Several previous studies have used 2-D companding and reported CNR improvement and increased target visibility in 2-D strain imaging (Alam and Ophir 1997; Chaturvedi et al. 1998). 3-D companding was used to compensate for out-of-plane decorrelation in 2-D strain images ( Insana et al. 1997). According to that, a considerable improvement in the CNR of 3-D strain images is anticipated by using 3-D companding. 102 6.5 In vitro Experiments 6.5.1 In vitro Murine Breast Tumor Strain Imaging The biochemistry and molecular biology of a rodent is similar to that of a human (Cardiff and Wellings 1999). However, murine breast tumors and mammary glands generally do not look structurally similar to human breast tumors (Cardiff and Wellings 1999). Nevertheless, In vitro murine breast tumor imaging experiments can offer a more rigorous test of the 3-D strain imaging system compared to the gelatin phantoms. They can demonstrate whether the developed system is capable of imaging a mammal breast tumor with acceptable contrast. An excised fresh tumor can be embedded in gelatin with graphite scatterers to mimic soft biological tissue and then imaged (Bilgen, Srinivasa et al. 2003). The time between tumor removal and imaging should be be less than 12 hours. 6.5.2 In vitro Ultrasound Imaging of Excised Tissue Specimens In vitro animal experiments can be conducted to test the dual-layer array and 3-D spatial compounding more rigorously. 103 BIBLIOGRAPHY Alam SK and Ophir J. “Reduction of signal decorrelation from mechanical compression of tissues by temporal stretching: applications to elastography,” Ultrasound Med. Biol. 23, 95-105 (1997). Alam SK, Ophir J, Konofagou EE. “An adaptive strain estimator for elastography,” IEEE Trans. Ultrason., Ferroelect., Freq. Contr. 45, 461-472 (1998). Austeng A and Holm S. “Sparse 2-D arrays for 3-D phased array imaging – experimental validation,” IEEE Trans. Ultras., Ferro., and Freq. Control, 49, 1087-1093 (2002). Awad SI, Yen JT. “3-D Spatial Compounding Using a Row-Column Array,” Ultrasonic Imaging, (submitted January 2009) Awad SI, Yen JT. “3D strain imaging using a rectilinear 2D array,” Ultrasonic Imaging 29, 220-230 (2007) Aysal TC, Barner KE. “Rayleigh-maximum-likelihood filtering for speckle reduction of ultrasound images,” IEEE Trans On Med Imag 26, 712-727 (2007). Belohlavek M, Foley DA, Gerber TC, Kinter TM, Greenleaf JF, Seward JB. “Three- and four-dimensional cardiovascular ultrasound imaging: a new era for echocardiography,” Mayo Clin Proc. 68, 221-240 (1993) Bilgen M, Srinivasa S, Lachman LB, Ophir J. “Elastography imaging of small animal oncology models: a feasibility study,” Ultrasound in Medicine and Biology 29, 1291- 1296 (2003). Burckhardt CB. “Speckle in ultrasound B-mode scans,” IEEE Trans. Sonics Ultrason. 25, 1-6 (1978). Cardiff RD and Wellings SR. “The comparative pathology of human and mouse mammary glands,” Journal of Mammary Gland Biology 4, 105-122 (1999). Chaturvedi P, Insana MF, Hall TJ. "2-D companding for noise reduction in strain imaging," IEEE Trans. Ultrason., Ferroelect., Freq. Contr. 45, 179-191 (1998). Christensen D. Ultrasonic Bioinstrumentation, John Wiley & Sons, Inc (1988). Claudon M, Tranquart F, Evans DH, Lefèvre F, Correas JM. “Advances in ultrasound,” European Radiology 12, 7-18 (2002). 104 Dahl JJ, Guenther DA, Trahey GE. “Adaptive Imaging and Spatial Compounding in the Presence of Aberration,” IEEE Trans Ultrason Ferroelec Freq Contr 52, 1131-1144 (2005). Entrekin RR, Porter BA, Sillesen HH, et al. “Real-Time Spatial Compound Imaging: Application to Breast, Vascular, and Musculoskeletal Ultrasound,” Seminars in Ultrasound, CT, and MRI 22, 50-64 (2001). Fenster A and Downey DB. “3-D ultrasound imaging: a review,” IEEE Engineering in Medicine and Biology Magazine 15, 41-51(1996). Fernandez AT, Gammelmark KL, Dahl JJ, Keen CG, Gauss RC, Trahey GE. “Synthetic elevation beamforming and image acquisition capabilitites using an 8 x 128 1.75D array,” IEEE Trans. Ultrason., Ferroelect., Freq. Contr 50, 40-57 (2003). Hall TJ, Bilgen M, Insana MF, Krouskop TA. “Phantom materials for elastography,” IEEE Trans. Ultrason., Ferroelect., Freq. Contr. 44, 1355-1365 (1997). Hall TJ, Zhu Y, Spalding CS. “In vivo real-time freehand palpation imaging,” Ultrasound Med. Biol. 29, 427-435 (2003). Hiltawsky KM, Krüger M, Starke C, Heuser L, Ermert H, Jensen A. “Freehand ultrasound elastography of breast lesions: clinical results,” Ultrasound Med. Biol. 27, 1461-1469 (2001). Hossack JA, Mauchamp P, Ratsimandresy L. “A high bandwidth transducer optimized for harmonic imaging,” 2000 IEEE Ultrasonics Symposium, pp. 1021-1024. Insana MF, Chaturvedi P, Hall TJ, Bilgen M. “3-D Companding using linear arrays for improved strain imaging,” 1997 IEEE Ultrasonics Symposium, 1435-1438. Jensen JA, Svendsen JB. “Calculation of pressure fields from arbitrarily shaped, apodized, and excited ultrasound transducers,” IEEE Trans Ultrason Ferroelec Freq Contr 39, 262-267 (1992). Jespersen SK, Wilhjelm JE, Sillesen H. “Multi-Angle compound Imaging,” Ultrason. Imag. 20, 81-102 (1998). Kallel F and Ophir J. “Three-dimensional tissue motion and its effect on image noise in elastography,” IEEE Trans. Ultrason., Ferroelect., Freq. Contr. 44, 1286-1296 (1997). Kallel F, Varghese T, Ophir J, Bilgen M. “The nonstationary strain filter in elastography: part II. Lateral and elevational decorrelation,” Ultrasound Med. Biol. 23, 1357-1369 (1997). 105 Kinsler LE, Frey AR, Coppens AB, Sanders JV, Fundamentals of acoustics, John Wiley & Sons, Inc (1982). Konofagou E and Ophir J. “A new elastographic method for estimation and imaging of lateral displacements, lateral strains, corrected axial strains and Poisson’s ratios in tissues,” Ultrasound Med. Biol. 24, 1183-1199 (1998). Krucker JF, Meyer CR, LeCarpentier GL, Brian Fowlkes J, Carson PL. “3D spatial compounding of ultrasound images using image-based nonrigid registration,” Ultrasound Med Biol 26, 1475-1488 (2000). Kwak JY, Kim EK, You JK, Oh KK. “Variable breast conditions: comparison of conventional and real-time compound ultrasonography,” J Ultrasound Med 23, 85–96 (2004). Lacefield JC, Pilkington WC, Waag RC. “Comparisons of lesion detectability in ultrasound images acquired using time-shift compensation and spatial compounding,” IEEE Trans Ultrason Ferroelec Freq Contr 51, 1649-1659 (2004). Lacefield JC, Waag RC. “Spatial coherence analysis applied to aberration correction using a two-dimensional array system,” J. Acoust. Soc. Am. 112, 2558-2566 (2002). Lanfranchi M, “Breast Ultrasound,” New York: Marban Books, pp. 273–275 (2000). Leotta DF, Martin RW. “Three-dimensional imaging of the rotator cuff: spatial compounding and tendon thickness,” Ultrasound Med Biol 26, 509-525 (2000). Li PC, Chen MJ. “Noise reduction in elastograms using temporal stretching with multicompression averaging,” IEEE Trans Ultrason Ferroelec Freq Contr 49, 39-46 (2002). Li PC, Li ML. “Adaptive Imaging Using the Generalized Coherence Factor,” IEEE Trans Ultrason Ferroelec Freq Contr 50, 93-110 (2000). Li PC, O’Donnell M. “Elevational spatial compounding,” Ultrason Imag 16, 176–189 (1994). Light ED, Davidsen RE, Fiering JO, Hruschka TA, Smith SW. “Progress in two- dimensional arrays for real-time volumetric imaging,” Ultrasonic Imaging 20, 1-15 (1998). 106 Lin DC, Nazarian LN, O’Kane PL, et al. “Advantages of real-time spatial compound sonography of the musculoskeletal system versus conventional sonography,” American Journal Of Roentgenology 179, 1629-1631 (2002). Lindop JE, Treece GM, Gee AH, Prager RW. “3D elastography using freehand ultrasound,” Ultrasound Med. Biol. 32, 529-545 (2006). Lorenz A, Sommerfeld HJ, Garcia-Schurmann M, Philippou S, Senge T, Ermert H. “A new system for the acquisition of ultrasonic multicompression strain images of the human prostate in vivo,” IEEE Trans. Ultrason., Ferroelect., Freq. Contr. 46, 1147-1154 (1999). Merks EJ, Bouakaz A, Bom N, Lancée CT, van der Steen AF, de Jong N. “Design of a multilayer transducer for acoustic bladder volume assessment,” IEEE Trans. Ultras., Ferro., And Freq. Control, 53, 1739–1738 (2006). Moskalik A, Carson PL, Meyer CR, et al. “Registration of three-dimensional compound ultrasound scans of the breast for refraction and motion correction,” Ultrasound Med Biol 21, 769-778 (1995). Oosterveld BJ, Thijssen JM, Verhoef WA. “Texture of B-mode echograms: 3-D simulations and experiments of the effects of diffraction and scatterer density,” Ultrason Imag 7, 142-160 (1985). Ophir J. “Elastography, a quantitative method for imaging the elasticity of biological tissue,” Ultrason. Imag. 13, 111–134 (1991). Oralkan O, Ergun AS, Cheng CH, Johnson JA, Karaman M, Lee TH, Khuri-Yakub BT. “Volumetric Ultrasound Imaging Using 2-D CMUT Arrays,” IEEE Trans. Ultras., Ferro., and Freq. Control, 50, 1581-1594 (2003). Palombo C, Kozakova M, Morizzo C, Andreuccetti F, Tondini A, Palchetti P, Mirra G, Parenti G, Pandian NG. “Ultrafast Three-Dimensional Ultrasound: Application to Carotid Artery Imaging,” Stroke 29, 1631-1637 (1998). Pellot-Barakat C, Frouin F, Insana MF, Herment A. “Ultrasound elastography based on multiscale estimations of regularized displacement fields,” IEEE Trans on Medical Imaging 23, 153-163 (2004). Ponnekanti H, Ophir J, Huang Y, Cespedes I. “Fundamental mechanical limitations on the visualization of elasticity contrast in elastography,” Ultrasound Med Biol 21, 533-543 (1995). Rao M, Chen Q, Shi H, Varghese T. “Spatial-angular compounding for elastography using beam steering on linear array transducers,” Medical Physics 33, 618-626 (2006). 107 Rao M, Varghese T. “Correlation analysis of three-dimensional strain imaging using ultrasound two-dimensional array transducers,” J Acoust Soc Am. 124, 1858-1865 (2008). Righetti R, Srinivasan S, Ophir J. “Axial resolution in elastography,” Ultrasound Med. Biol. 28, 101-113 (2002). Rohling RN, Gee AH, Berman L. Automatic registration of 3-D ultrasound images, Ultrasound Med Biol 24, 841-854 (1998). Rohling RN, Gee AH, Berman L. “Spatial compounding of 3-D ultrasound images,” Information Processing In Medical Imaging 1230, 519-524 (1996). Sahiner B, Chan HP, Roubidoux MA, Helvie MA, Hadjiiski LM, Ramachandran A, Paramagul C, LeCarpentier GL, Nees A, Blane C. “Computerized characterization of breast masses on three-dimensional ultrasound volumes,” Medical Physics 31, 744-754 (2003). Saitoh S, Izumi M, Mine Y. “A dual frequency ultrasonic probe for medical applications,” Trans. Ultras., Ferro., And Freq. Control,42, 294-300 (1995). Savord B and Solomon R. “Fully sampled matrix transducer for real time 3D ultrasonic imaging,” 2003 IEEE Ultrasonics Symposium, pp. 945-953. Schaar JA, De Korte CL, Mastik F, Strijder C, Pasterkamp G, Boersma E, Serruys PW, Van Der Steen AF. “Characterizing vulnerable plaque features with intravascular elastography,” Phys. Med. Biol. 45, 1465-1475 (2000). Schmidt MA, Ohazama CJ, Agyeman KO, Freidlin RZ, Jones M, Laurienzo JM, Brenneman CL, Arai AE, Von Ramm OT, Panza JA. “Real-time three-dimensional echocardiography for measurement of left ventricular volumes,” Amer. J. Card. 84, 1434–1439 (1999). Seo BK, Oh YW, Kim HR, et al. “Sonographic evaluation of breast nodules: comparison of conventional, real-time compound, and pulse-inversion harmonic images,” Korean J Radiol 3, 38–44 (2002). Seo CH, Yen JT. “A 256 × 256 2-D array transducer with row-column addressing for 3-d rectilinear imaging,” IEEE Trans Ultrason Ferroelec Freq Contr (accepted 2008). Smith JA and Andreopoulou E. “An overview of the status of imaging screening technology for breast cancer,” Annals of Oncology 15, i18–i26 (2004). 108 Smith WL, Surry KJM., Mills GR, Downey DB, Fenster A. “Three dimensional ultrasound-guided core needle breast biopsy,” Ultrasound in Medicine and Biology 27, 1025-1034 (2001). Souchon R, Rouvière O, Gelet A, Detti V, Srinivasan S, Ophir J, Chapelon JY. “Visualisation of HIFU lesions using elastography of the human prostate in vivo,” Ultrasound Med. Biol. 29, 1007-1015 (2003). Szabo TL. “Diagnostic ultrasound imaging,” Elsevier Academic Press, Inc (2004). Techavipoo U, Chen Q, Varghese T, Zagzebski JA, Madsen EL. “Noise reduction using spatial-angular compounding for elastography,” IEEE Trans Ultrason Ferroelec Freq Contr 51, 510-520 (2004). Thomas A, Kümmel S, Fritzsche F, Warm M, Ebert B, Hamm B, Fischer T. “Real-time sonoelastography performed in addition to B-mode ultrasound and mammography : Improved differentiation of breast lesions?,” Acad Radiol 13, 1496-1504 (2006). Trahey GE, Allison JW, Smith SW, Von Ramm OT. “A quantitative approach to speckle reduction via frequency compounding,” Ultrason Imag 8, 151–164 (1986). Trahey GE, Smith SW, Von Ramm OT. “Speckle pattern correlation with lateral aperture translation: experimental results and implications for spatial compounding,” IEEE Trans Ultrason Ferroelec Freq Contr 33, 257-264 (1986). Tuthill TA, Sperry RH, Parker KJ. “Deviations from Rayleigh statistics in ultrasonic speckle,” Ultrason Imag 10, 81-89 (1988). Varghese T, Konofagou EE, Ophir J, Alam SK, Bilgen M. “Direct strain estimation in elastography using spectral cross-correlation,” Ultrasound Med. Biol. 26, 1525-1537 (2000). Varghese T, Zagzebski JA, Lee FT Jr. “Elastography imaging of thermal lesion in the liver in vivo following radio frequency ablation: Preliminary results,” Ultrasound Med. Biol. 28, 1467-1473 (2002). Viola F and Walker WF. “A spline-based algorithm for continuous time-delay estimation using sampled data,” IEEE Trans. Ultrason., Ferroelect., Freq. Contr. 52, 80-93 (2005). Wagner RF, Smith SW, Sandrik JM, Lopez H. “Statistics of speckle in ultrasound B- scans,” IEEE Trans Sonics Ultrason 25, 156-163 (1983). 109 Walker WF and Trahey GE. “A fundamental limit on delay estimation using partially correlated speckle signals,” IEEE Trans. Ultrason., Ferroelect., Freq. Contr. 42, 301-308 (1995). Yen JT, Seo C, Awad SI, Jeong J. “A Dual-Layer Transducer Array for 3-D Rectilinear Imaging,” IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control 56, 204-212 (2009) Yen JT and Smith SW. “Real-time rectilinear volumetric imaging,” IEEE Trans. Ultras., Ferro., and Freq. Control, 49, 114-124 (2002). Yen JT and Smith SW. “Real-time rectilinear volumetric imaging using receive mode multiplexing,” IEEE Trans. Ultrason., Ferroelect., Freq. Contr. 51, 216-226 (2004). Yen JT and Smith SW, “Real-time curvilinear and improved rectilinear volumetric imaging,” 2001 IEEE Ultrasonics Symposium, 1117-1122. Yen JT, Steinberg JP, Smith SW. “Sparse 2-D Array Design for Real-Time Rectilinear Volumetric Imaging,” IEEE Trans. Ultras., Ferro., And Freq. Control, 47, 93-110 (2000). Zahiri-Azar R and Salcudean SE. “Motion estimation in ultrasound images using time domain cross correlation with prior estimates,” IEEE Trans. on Biomedical Eng. 53, 2640-2650 (2006). 110 APPENDIX A: INTERFACING THE TRANSDUCER ARRAYS WITH THE ULTRASOUND SYSTEM To acquire data using the dual-layer and the row-column transducer arrays, the arrays were interfaced with the Sonix RP ultrasound system (Ultrasonix, Vancouver, Canada). Table A.1 lists the parts used to interface the arrays to the system. Part name Catalog number Description Manufacturer ZIF connector ITT Cannon DL1- 156PW4 A zero insertion force (ZIF) connector that mates with the channels connector on the Ultrasonix system. ITT Industries Socket strip Samtec SSW-113- 07-G-D A double row, 13 contacts per row, through hole, right angle, gold plated, socket strip that mates with the headers on the ZIF connector. Samtec Inc. Cable Digikey MC34R- 100-ND Round, shielded, flat cable that has 34 wires with 28 AWG wire gauge. 3M Corp. Header strip Samtec TSW-117- 07-G-D A double row, 17 contacts per row, through hole, right angle, gold plated, header strip that mates with the socket strip and wire-mount socket strip. Samtec Inc. Wire- mount socket strip Digikey MSC34K- ND A double row, 17 contacts per row, gold plated, wire-mount socket strip that mates with header strip. 3M Corp. Table A.1: A list of the parts used to interface the dual-layer and the row-column transducer arrays to the ultrasound system. Six socket strips were used to access a total of 156 contacts on the ZIF connector. Two 8-feet pieces of the cable were cut. Each cable contains 34 wire, two of which were used as ground and the rest were used to access 32 receive elements. The wires in the two 8-feet cable were soldered to the socket strips connecting the cables to 64 receive 111 channels of the system. The other ends of the two 8-feet cable were soldered to two header strips. Eight 2-feet pieces of the cable were cut and wire-mount sockets were mounted on both ends of the cables. Eight header strips were soldered on the transmit and receive custom printed circuit boards to access the 256 transmit elements and the 256 receive elements in each array. The 2-feet cables were used as extension cables to access the receive header strips on the printed circuit boards. Two 8-feet 24 AWG gauge solid wire were soldered to the socket strip to be used for transmit. The other end of these wires were soldered to a mini test clips. To acquire data, the test clips were connected to two transmit elements at a time and the two 8-feet cables were connected to two 2-feet cables. A different set of two 2-feet cables were used until data from all 256 receive elements were collected. This process is repeated until all transmit and receive element combinations were acquired. Figure A.1 illustrates interface between the ultrasound machine and the transducer array. 112 Figure A.1: A diagram that illustrates the interface between the ultrasound machine and the transducer array. Transducer array 2-feet cables 8-feet cables Ultrasound system ZIF connect 8-feet wires Receive PCB Transmit PCB Header strips 113 APPENDIX B: C++ CODE FOR DATA ACQUISITION WITH ULTRASONIX The following is the C++ code used to acquire RF data from the dual-layer and row- column arrays using the setup described in Appendix A. It uses Texo version 1.2 software development kit (SDK) provided by Ultrasonix. In the data collection process, pairs of transmit elements were manually multiplexed while collecting data from 64 receive elements. Then a different set of 64 receive elements was used while pairs of transmit elements were manually multiplexed. This process is repeated until all transmit and receive element combinations were acquired. This code saves the different combinations of transmit and receive with 32 receive signals and 100 frames per signal in separate raw data files. It also displays instructions which transmit elements the user should connect the transmit wires to as well as receive cables. Table B.1 lists the imaging parameters used in this code. Parameter Value Transmit power 100% Number of cycles 2 Center frequency 5 MHz Acquisition depth 120 mm Number of frames 100 Receive gain 70% Table B.1: Imaging parameters used in the C++ code. 114 //Texo version 1.2 #include <stdio.h> #include <string.h> #include "stdafx.h" #include <texo.h> #include <texo_def.h> #include <math.h> bool newImage(void *, unsigned char *, int); bool createSequence1(texo &, int, int, int, int); bool createSequence2(texo &, int, int, int, int); bool saveData(texo &, char[3], int); #ifndef DATAPATH #define DATAPATH "../../dat/" #endif struct entry { char str[5]; }; int main() { texo tex; int sel, numFrames; int rxID, txID, num_cycle, depth, power_num, sleep_num, frame_num; //TGC_CURVE tgc; struct entry txname1[128]={ {"a000"},{"a001"},{"a002"},{"a003"},{"a004"},{"a005"},{"a006"}, {"a007"},{"a008"},{"a009"},{"a010"},{"a011"},{"a012"},{"a013"}, {"a014"},{"a015"},{"a016"},{"a017"},{"a018"},{"a019"},{“a020"}, {"a021"},{"a022"},{"a023"},{"a024"},{"a025"},{"a026"},{"a027"}, {"a028"},{"a029"},{“a030"},{"a031"}, {"b000"},{"b001"},{"b002"},{"b003"},{"b004"},{"b005"},{"b006"}, {"b007"},{"b008"},{"b009"},{"b010"},{"b011"},{"b012"},{"b013"}, {"b014"},{"b015"},{"b016"},{"b017"},{"b018"},{"b019"},{"b020"}, {"b021"},{"b022"},{"b023"},{"b024"},{"b025"},{"b026"},{"b027"}, {"b028"},{"b029"},{"b030"},{"b031"}, {"c000"},{"c001"},{"c002"},{"c003"},{"c004"},{"c005"},{"c006"}, 115 {"c007"},{"c008"},{"c009"},{"c010"},{"c011"},{"c012"},{"c013"}, {"c014"},{"c015"},{"c016"},{"c017"},{"c018"},{"c019"},{"c020"}, {"c021"},{"c022"},{"c023"},{"c024"},{"c025"},{"c026"},{"c027"}, {"c028"},{"c029"},{"c030"},{"c031"}, {"d000"},{"d001"},{"d002"},{"d003"},{"d004"},{"d005"},{"d006"}, {"d007"},{"d008"},{"d009"},{"d010"},{"d011"},{"d012"},{"d013"}, {"d014"},{"d015"},{"d016"},{"d017"},{"d018"},{"d019"},{"d020"}, {"d021"},{"d022"},{"d023"},{"d024"},{"d025"},{"d026"},{"d027"}, {"d028"},{"d029"},{"d030"},{"d031"}}; struct entry txname2[128]={ {"w000"},{"w001"},{"w002"},{"w003"},{"w004"},{"w005"},{"w006"}, {"w007"},{"w008"},{"w009"},{"w010"},{"w011"},{"w012"},{"w013"}, {"w014"},{"w015"},{"w016"},{"w017"},{"w018"},{"w019"},{"w020"}, {"w021"},{"w022"},{"w023"},{"w024"},{"w025"},{"w026"},{"w027"}, {"w028"},{"w029"},{"w030"},{"w031"}, {"x000"},{"x001"},{"x002"},{"x003"},{"x004"},{"x005"},{"x006"}, {"x007"},{"x008"},{"x009"},{"x010"},{"x011"},{"x012"},{"x013"}, {"x014"},{"x015"},{"x016"},{"x017"},{"x018"},{"x019"},{"x020"}, {"x021"},{"x022"},{"x023"},{"x024"},{"x025"},{"x026"},{"x027"}, {"x028"},{"x029"},{"x030"},{"x031"}, {"y000"},{"y001"},{"y002"},{"y003"},{"y004"},{"y005"},{"y006"}, {"y007"},{"y008"},{"y009"},{"y010"},{"y011"},{"y012"},{"y013"}, {"y014"},{"y015"},{"y016"},{"y017"},{"y018"},{"y019"},{"y020"}, {"y021"},{"y022"},{"y023"},{"y024"},{"y025"},{"y026"},{"y027"}, {"y028"},{"y029"},{"y030"},{"y031"}, {"z000"},{"z001"},{"z002"},{"z003"},{"z004"},{"z005"},{"z006"}, {"z007"},{"z008"},{"z009"},{"z010"},{"z011"},{"z012"},{"z013"}, {"z014"},{"z015"},{"z016"},{"z017"},{"z018"},{"z019"},{"z020"}, {"z021"},{"z022"},{"z023"},{"z024"},{"z025"},{"z026"},{"z027"}, {"z028"},{"z029"},{"z030"},{"z031"}}; struct entry rxname[4]={{"1&2 "},{"3&4 "},{"5&6 "},{"7&8 "}}; // initialize and set the data file path if(!tex.init(DATAPATH)) return 0; // always load config 0 for now if(!tex.load(0)) return 0; // choose a probe 116 if(!tex.activateProbeConnector(0)) return 0; // set the new frame callback tex.setCallback(newImage, 0); // initialize global parameters power_num=15; depth=120000; sleep_number=630; frame_number=100; tex.setPower(power_num, power_num, power_num); tex.addTGC(0.7); // tell program to initialize for new sequence if(!tex.beginSequence()) return 0; // build sequence if(!createSequence1(tex,0,0,1,100000)) return 0; if(!createSequence2(tex,0,0,1,100000)) return 0; // tell program to finish sequence if(tex.endSequence() == -1) return 0; for (rxID=1;rxID<5;rxID++) { printf("\n\nConnect to RX ");printf(rxname[rxID-1].str); printf(" (enter any number to continue): "); scanf("%d", &sel); for (txID=1;txID<129;txID++) { rpt: printf("\n\nConnect to TX ");printf(txname1[txID-1].str); printf(" (enter 1 to acquire, 9 to exit): "); scanf("%d", &sel); if (sel==1) { tex.beginSequence(); createSequence1(tex,16,0,depth); 117 tex.endSequence(); numFrames = 0; while (numFrames<frame_number) { tex.runImage(); ::Sleep(sleep_number); tex.stopImage(); numFrames = tex.getCollectedFrameCount(); } fprintf(stdout, "Acquired (%d) frames\n", numFrames); saveData(tex,txname1[txID-1].str, rxID);//*/ tex.beginSequence(); createSequence2(tex,16,0,depth); tex.endSequence(); numFrames = 0; while (numFrames<frame_number) { tex.runImage(); ::Sleep(sleep_number); tex.stopImage(); numFrames = tex.getCollectedFrameCount(); } fprintf(stdout, "Acquired (%d) frames\n", numFrames); saveData(tex,txname2[txID-1].str, rxID);//*/ } else if (sel==9) { tex.shutdown(); return 0; } else goto rpt; } } // exit program tex.shutdown(); return 0; } 118 bool createSequence1(texo &tex, int txcenter, int rxcenter, int depth) { int j,lineSize,rx_id; texoTransmitParams tx; texoReceiveParams rx; int rx_map[32]={31,30,29,28,27,26,25,24,7,6,5,4,3,2,1,0, 63,62,61,60,59,58,57,56,39,38,37,36,35,34,33,32}; tx.aperture = 2; tx.focusDistance = 400000; ///um tx.angle = 0; tx.frequency =5000000; strcpy(tx.pulseShape, "+-+-"); tx.tableIndex = -1; rx.aperture = 0; rx.angle = 0; rx.maxApertureDepth = 20000; rx.acquisitionDepth = depth; rx.speedOfSound = 1540; rx.channelMask[0] = 0; rx.channelMask[1] = 1; rx.applyFocus = false; rx.useManualDelays = false; rx.decimation = 0; rx.customLineDuration = 0; rx.lgcValue = 4000; rx.tgcSel = 0; rx.tableIndex = -1; tx.centerElement = (txcenter)*10; for (rx_id=0; rx_id<32; rx_id++) 119 { rx.centerElement = (rx_map[rx_id])*10; for(j = (rx_map[rx_id]%32); j < ((rx_map[rx_id]%32)+1); j++) { rx.channelMask[0] = ~(1 << j); lineSize = tex.addLine(rfData, tx, rx); if(lineSize == -1) return false; } } return true; } bool createSequence2(texo &tex, int txcenter, int rxcenter, int depth) { int j,lineSize,rx_id; texoTransmitParams tx; texoReceiveParams rx; int rx_map[32]={95,94,93,92,91,90,89,88,71,70,69,68,67,66, 65,64,127,126,125,124,123,122,121,120,103,102,101,100,99,98,97,96}; tx.aperture = 2; tx.focusDistance = 400000; ///um tx.angle = 0; tx.frequency =5000000; strcpy(tx.pulseShape, "+-+-"); tx.tableIndex = -1; rx.aperture = 0; rx.angle = 0; rx.maxApertureDepth = 20000; rx.acquisitionDepth = depth; rx.speedOfSound = 1540; rx.channelMask[0] = 0; rx.channelMask[1] = 1; rx.applyFocus = false; rx.useManualDelays = false; rx.decimation = 0; rx.customLineDuration = 0; 120 rx.lgcValue = 4000; rx.tgcSel = 0; rx.tableIndex = -1; tx.centerElement = (txcenter)*10; for (rx_id=0; rx_id<32; rx_id++) { rx.centerElement = (rx_map[rx_id])*10; for(j = (rx_map[rx_id]%32); j < ((rx_map[rx_id]%32)+1); j++) { rx.channelMask[0] = ~(1 << j); lineSize = tex.addLine(rfData, tx, rx); if(lineSize == -1) return false; } } return true; } // store data to disk bool saveData(texo &tex, char cr[4], int rxID) { int numFrames, frameSize; FILE * fp; struct entry { char str[14]; }; struct entry s1[5]={ {"D:\\newrf\\RX1\\"},{"D:\\newrf\\RX2\\"},{"D:\\newrf\\RX3\\"}, {"D:\\newrf\\RX4\\"},{"D:\\newrf\\RX5\\"} }; char * s2 = strcat(s1[rxID-1].str,"ele_"); char * s3 = strcat(s2,cr); char * fullpath= s3; 121 numFrames = tex.getCollectedFrameCount(); frameSize = tex.getFrameSize(); if(numFrames < 1) { fprintf(stderr, "No frames have been acquired\n"); return false; } if( !(fp = fopen(fullpath, "wb+")) ) { fprintf(stderr, "Could not store data to specified path\n"); return false; } fwrite(tex.getCineStart(), frameSize, numFrames, fp); fclose(fp); printf(fullpath); printf(" was saved successfully\n"); return true; } // print out stats for new frame bool newImage(void *, unsigned char * data, int frameID) { //fprintf(stdout, "new frame recieved addr=(0x%08x) id=(%d)\n", data, frameID); return true; }

Abstract (if available)

Abstract 3-D ultrasound has been found useful for diagnostic imaging. It has several advantages over 2-D ultrasound including viewing planes that are usually inaccessible using 2-D ultrasound, more accurate volume measurements and volume-rendered images. These capabilities can improve the detectability of cancerous tumors in soft tissue such as the breast and prostate. Additionally, real-time 3-D ultrasound can be utilized in guiding procedures such as breast needle biopsy.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

High-frequency ultrasound imaging system with Doppler features for biomedical applications using 30~35 mHz linear arrays and an analog beamformer

PDF

High frequency ultrasound array for ultrasound-guided breast biopsy

PDF

Improved contrast in ultrasound imaging using dual apodization with cross-correlation

PDF

Array transducers for high frequency ultrasound imaging

PDF

High frequency ultrasonic imaging for the development of breast biopsy needle with miniature ultrasonic transducer array

PDF

Ultrasound imaging and coherence estimation using spatial filters

PDF

Development of back-end processing system for high frequency ultrasound b-mode imaging

PDF

Synthetic aperture imaging platform based on CMOS high voltage 1 to 64 multiplexer / de-multiplexer for ultrasound guided breast biopsy needle

PDF

Novel beamforming with dual-layer array transducers for 3-D ultrasound imaging

PDF

A high frequency array- based photoacoustic microscopy imaging system

PDF

Transducers and signal processing techniques for simultaneous ultrasonic imaging and therapy

PDF

Development of high-frequency (~100mhz) PZT thick-film ultrasound transducers and arrays

PDF

Configurable imaging platform for super-harmonic contrast-enhanced ultrasound imaging

PDF

Fresnel beamforming for low-cost, portable ultrasound systems

PDF

An ultrasound localization system for endovascular aortic aneurysm repair

PDF

High-frequency ultrasound array-based imaging system for biomedical applications

PDF

Single-cell analysis with high frequency ultrasound

PDF

High frequency ultrasonic phased array system and its applications

PDF

Microfluidic cell sorting with a high frequency ultrasound beam

PDF

Design and development of ultrasonic array transducers for specialized applications

Asset Metadata

Creator Awad, Samer (author)

Core Title Lesion enhancement for three dimensional rectilinear ultrasound imaging

Contributor Electronically uploaded by the author (provenance)

School Andrew and Erna Viterbi School of Engineering

Degree Doctor of Philosophy

Degree Program Biomedical Engineering

Publication Date 05/12/2009

Defense Date 03/26/2009

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag 2-D array,3-D imaging,OAI-PMH Harvest,spatial compounding,strain imaging,ultrasound

Language English

Advisor Yen, Jesse T. (committee chair), Brun, Todd A. (committee member), Cannata, Jonathan Matthew (committee member), Liu, Brent (committee member), Shung, K. Kirk (committee member)

Creator Email samer.awad@gmail.com,sawad@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-m2245

Unique identifier UC1202163

Identifier etd-Awad-2881 (filename),usctheses-m40 (legacy collection record id),usctheses-c127-241907 (legacy record id),usctheses-m2245 (legacy record id)

Legacy Identifier etd-Awad-2881.pdf

Dmrecord 241907

Document Type Dissertation

Rights Awad, Samer

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