Characterization of Ultrasound Elevation Beamwidth Artefacts for Brachytherapy Needle Insertion

Transcription

1 Characterization of Ultrasound Elevation Beamwidth Artefacts for Brachytherapy Needle Insertion by Mohammad Peikari A thesis submitted to the School of Computing in conformity with the requirements for the degree of Master of Science Queen s University Kingston, Ontario, Canada August 2011 Copyright c Mohammad Peikari, 2011

2 Abstract Ultrasound elevation beamwidth is the out of plane thickness causing image artefacts normally appearing around anechoic areas in the medium. These artefacts could also cause uncertainties in localizing objects (such as a surgical needle) in the ultrasound image slices. This thesis studies the clinical significance of elevation beamwidth artefacts in needle insertion procedures. A new measurement device was constructed to measure the transrectal ultrasound elevation beamwidth. The beam profiles of various lateral and axial distances to the transducer were generated. It is shown that the ultrasound elevation beamwidth converges to a point within its focal zone close to the transducer. This means that the transrectal ultrasound images have the best resolution within the focal zone of the ultrasound close to the transducer. It is also shown that the ultrasound device settings have a considerable impact on the amount of beamwidth artefacts. Needle tip localization error was examined for a curvilinear transrectal ultrasound transducer. Beveled prostate brachytherapy needles were inserted through all holes of a grid template orthogonal to the axial beam axis. The effects of device imaging parameters were also investigated on the amount of localization error. Based on the developed results, it was found that the imaging parameters of an ultrasound device have direct impact on the amount of object localization error from 0.5 mm to 4 mm. The smallest localization error occurs laterally close to the i

3 center of the grid template, and axially within the beam s focal zone. Similarly, the largest localization error occurs laterally around both sides of the grid template, and axially within the beam s far field. Using the ultrasound device with appropriate imaging settings could minimize the effects of these artefacts. We suggest to reduce the gain setting of the ultrasound device. This will reduce the energies assigned to the off-axis beams and as a result, the elevation beamwidth artefacts are minimized. ii

4 Acknowledgements This work would not have been completed with out the consistent support of my parents, Ahmad Reza Peikari and Shekoufeh Fathi. I can not put down in words how much their caring love have always been a source of motivation for me with out which, I would never have had the fortune of being where I stand today. I would like to thank my supervisor, Dr. Gabor Fichtinger, for providing me with the freedom yet the guidance and support to pursue this research to this point. I would also thank the members of the Percutaneous Surgery (Perk) Lab of Queen s University, Thomas. K. Chen, Andras Lasso, and Tamas Heffter, for providing a collaborative and encouraging environment to discuss, solve and proceed on the issues of this work. I would also thank all my friends and colleagues here at Queen s for bringing joy and friendship to the lab s working environment: Hamed Peikari, Farhad Imani, Mostafa Mostafavi, Mohsen Yazdi, Andrew Dickinson, Laura Bartha, Ehsan Dehghan, Yashar Madjidi, Hossein Sadjadi, Alexandra Pompeu Robinson, Reza Seifabadi and Eric Moult. It has been a great pleasure knowing everyone here at Quneen s University. iii

5 Statement of Co-authorship The work presented in this thesis was done under the supervision of Dr. Gabor Fichtinger, who provided feedback, general direction and corrections to the manuscript. Otherwise the material presented in this document is the original work of the author. Any published (or unpublished) ideas and/or techniques from the work of others are fully acknowledged in accordance with the standard referencing practices. This thesis presents three papers that have already been accepted or in the process within international and top quality journals and conferences. The structure is as follows: Chapter 2: Section-Thickness Profiling for Brachytherapy Ultrasound Guidance. A version of this chapter was presented in the Society of Photo-Optical Instrumentation Engineers (SPIE) 2011 Medical Imaging and was placed among the seven best student paper award finalists. Mohammad Peikari (main author) contributed the coding, data analysis, designed the three-dimensional ultrasound beam profiling device, acquired the ultrasound images and designed the experiments and writing of the manuscript. Thomas K. Chen, Dr. Everette. C. Burdette, and Dr. Gabor Fichtinger supported the work by providing feedbacks and helpful discussions through out the course of project and contributed by editing the manuscript. Chapter 3: Effects of Ultrasound Section-Thickness on Brachytherapy Needle iv

6 Tip Localization Error. A version of this chapter is accepted for publication in the Medical Image Computing and Computer Assisted Intervention (MICCAI) Mohammad Peikari (main author) contributed the coding, data analysis, acquired the ultrasound images and designed the experiments and writing of the manuscript. Tamas Heffter contributed to the coding of the image acquisition software and helped in ultrasound image acquisition. Thomas K. Chen, Andras Lasso, and Dr. Gabor Fichtinger supported the work by providing feedbacks and helpful discussions through out the course of project and contributed by editing the manuscript. Chapter 4: Characterization of Ultrasound Elevation Beamwidth Artefacts for Brachytherapy Needle Insertion. A version of this chapter is submitted to the Journal of Medical Physics. Mohammad Peikari (main author) contributed the data analysis, designed the three-dimensional ultrasound beam profiling device, acquired the ultrasound images and designed the experiments and writing of the manuscript. Tamas Heffter contributed to the coding of the image acquisition software. Thomas K. Chen, Andras Lasso, Dr. Everette C. Burdette and Dr. Gabor Fichtinger supported the work by providing feedbacks and helpful discussions through out the course of project and contributed by editing the manuscript. v

7 Contents Abstract Acknowledgements Statement of Co-authorship Contents List of Tables List of Figures i iii iv vi viii ix 1 Introduction Motivation Background Thesis Objectives Contributions Organization of Thesis Section-Thickness Profiling for Brachytherapy Ultrasound Guidance Introduction The US Physics Prior Art The Clinical Brachytherapy Setup Methodology The Concept The Experimental Setup Results Discussion Future work Conclusions vi

8 3 Effects of Ultrasound Section-Thickness on Brachytherapy Needle Tip Localization Error Introduction Methodology Experimental Setup US Beam Profile Estimation Needle Offset Measurement on US Images Results and Discussion Conclusion Characterization of Ultrasound Elevation Beamwidth Artefacts for Brachytherapy Needle Insertion Introduction Clinical Significance and Background Elevation Beamwidth Artefacts Prior Art in Elevation Beamwidth Characterization Materials and Methods US Beamwidth Calculation Principles Wire-Bridge Phantom Needle Localization Error Measurements Hardware Configurations Experimental Setup Experiments and Results US Beamwidth Measurement Needle Tip Localization Analysis Comparison Between Beam Profile and Needle Tip Profile Side Lobe Extraction Discussion Conclusions Conclusion Summary Future Work Bibliography 71 vii

9 List of Tables 2.1 Comparison summary of our results with previous experiments Summary of the needle insertion offsets for different US settings Combinations of different TRUS imaging settings used in our experiments viii

10 List of Figures 2.1 (a) US beam pattern in Axial, Lateral, and Elevation axes. (b) Axial resolution, (c) Lateral resolution, and (d) Elevation resolution (a)-(b) Beam pattern of the B-mode and A-mode linear array US with corresponding point reflectors[8]. (c) Axial, lateral and elevational axes convention with respect to the TRUS beam pattern. (d) Sectionthickness estimation principle (a) Location of an object tissue (prostate) could be wrongly detected in presence of the US section-thickness[8]. (b) Different parts of the clinical TRUS device. (c) A schematic showing prostate brachytherapy procedure (a) CAD design of the TRUS-bridge phantom. (b) stepper, (c) transrectal probe, (d) TRUS-bridge phantom mounted on an industrystandard stepper, (e) rubber membrane on the inclined surface, (f) metal clamps, and (g)-(h) the experimental setup (a) A subset of elevation beamwidth measurements. (b) Examples of section-thickness artifacts over different depths of the TRUS images, frequency= 6MHz ix

11 2.6 TRUS elevation beam profiles at 5MHz and 6MHz central operating frequencies (a) US beam pattern in axial, lateral, and elevation axes. (b) Axial resolution, (c) Lateral resolution, and (d) Elevation resolution (a)-(b) Beam pattern of an unfocused B-mode and A-mode US linear array with corresponding point reflectors [8]. (c) Axial, lateral and elevational axes convention with respect to the TRUS beam pattern. (d) Section-thickness estimation principle (a) The beam profiling phantom mounted on a Burdette Medical Systems stepper, (b) Needle insertion setup, (c) Needle offset measurement principle (a) A subset of elevation main beam thickness measurements from 28mm to 90mm axial distance, at gain= 55%, dynamic range= 78dB, and power= 0. (b) TRUS elevation beam profile at 6MHz central operating frequency from 10mm to 65mm axial distance, at gain= 55%, dynamic range= 78dB, and power= 0. (c) Needle tip appearance in a TRUS image at gain= 0%, dynamic range= 15dB, and power= (a) US beam profile for gain= 100%, dynamic range= 50dB, and power= 0. (b) Average and std. of N(j) N(1) for gain= 100%. (c) US beam profile for gain= 50%, dynamic range= 50dB, and power= 0. (d) Average and standard deviation of N(j) N(1) for gain= 50%. (e) US beam profile for gain= 0%, dynamic range= 50dB, and power= 0. (f) Average and standard deviation of N(j) N(1) for gain= 0% x

12 4.1 (a) A schematic showing prostate brachytherapy procedure. (b) Implanting needles inserted through grid template holes (a) US beam pattern in axial, lateral, and elevation axes. (b)-(c) Beam pattern of a B-mode and A-mode TRUS with corresponding point reflectors (a) Two-dimensional representation of the main and side lobe beams. (b) Diagram showing the off-axis side lobe energies encountering two objects. (c) The US device assumes that the returning echoes from the off-axis side lobes came from the main beam and misplaces the structure (a) Ultrasound beamwidth measurement principle in presence of the main lobe only. (b) Ultrasound beamwidth measurement principle in presence of the main lobe and side lobe energies (a) Different parts of the clinical TRUS device. (b) The Wire-Bridge phantom on a clinical TRUS stepper (a) Needle offset measurement principle in presence of main lobe only. (b) Needle offset measurement principle in presence of main lobe and side lobe energies (a) The needle insertion experimental setup. US probe and needles are deep into a glycerol-water bath. (b) Standard grid template used to guide needle insertion in brachytherapy procedures. (c) Needle tip appearance in a TRUS image at gain= 0%, dynamic range= 15 db, and power= xi

13 4.8 (a) The beam profiling experimental setup. The US probe and Wire- Bridge phantom are submerged in a glycerol-water bath. (b) Examples of beamwidth artefacts over different depths of the TRUS images, frequency= 6 MHz, gain= 50%, dynamic range= 50 db, and power= (a)(c)(e) US beam pattern at various lateral positions labeled with A, a, and B on the grid template for different gain values at dynamic range=50 db, and power=-4. (b)(d)(f) Mean and Std. of N(j) N(1) at various lateral positions for different gain, dynamic range, and power settings (a)(c)(e) US beam pattern at various lateral positions labeled with b, C, and c on the grid template for different gain values at dynamic range=50 db, and power=-4. (b)(d)(f) Mean and Std. of N(j) N(1) at various lateral positions for different gain, dynamic range, and power settings (a)(c)(e) US beam pattern at various lateral positions labeled with D, d, and E on the grid template for different gain values at dynamic range=50 db, and power=-4. (b)(d)(f) Mean and Std. of N(j) N(1) at various lateral positions for different gain, dynamic range, and power settings xii

14 4.12 (a)(c)(e) US beam pattern at various lateral positions labeled with e, F, and f on the grid template for different gain values at dynamic range=50 db, and power=-4. (b)(d)(f) Mean and Std. of N(j) N(1) at various lateral positions for different gain, dynamic range, and power settings (a) US beam pattern at the lateral position labeled with G on the grid template for different gain values at dynamic range=50 db, and power=-4. (b) Mean and Std. of N(j) N(1) at various lateral positions for different gain, dynamic range, and power settings (a) Wire reflection corresponding to the lateral position labeled b on the grid template. Gain=0%, dynamic range=50 db, and power=-4. (b) Intensity profile of the wire reflection with out side lobe energies. (c) Wire reflection corresponding to the lateral position labeled b on the grid template. Gain=100%, dynamic range=50 db, and power=- 4. (d) Intensity profile of the wire reflection with main lobe energy (highest intensity peak) and side lobe energies (low intensity peaks).. 62 xiii

15 1 Chapter 1 Introduction 1.1 Motivation Prostate cancer is the second leading cause of cancer related death in men and one of the most common cancers in North America. There have been 217,730 estimated new cases (28% of all new cancers in men) and 32,050 estimated deaths in 2010 in the United States alone [1]. There may not be a need to treat this cancer by invasive treatments in case when the cancer is detected at its early stages. For this group of patients, less invasive treatment options may be more adequate with lesser side effects and complications. Prostate brachytherapy is one of the treatment options available for the prostate cancer when it is detected at its early stage. Brachytherapy procedure is performed much faster with least possible side effects when compared to other treatment options such as radical prostatectomy or external beam radiation. This procedure entails permanent implantation of small radioactive seeds under live ultrasound image visualisation to eradicate the cancer.

16 1.1. MOTIVATION 2 Ultrasound image artefacts commonly occur in medical ultrasound (US) images and is one of the major sources of confusion to the physician. Medical US has attracted the attention of a wide range clinicians and researchers developing different types of clinical applications for various types of tasks. It has been discussed in the literature that the US may include certain types of artefacts caused by the beam thickness elevation beamwidth and the unwanted parts of the beam emitted off-axis (side lobes). The artefacts appear mainly around the anechoic areas of the tissue and may conceal the true tissue structure which leads to incorrect medical diagnosis. The non-uniform US beam thickness may also cause uncertainties in tool tip localization along the elevation axis of the beam if the tools are inserted perpendicular to the long axis of the beam. The motivating application of this research is the transrectal ultrasound (TRUS) guided prostate brachytherapy. The first stage before the brachytherapy procedure is treatment planning. A prostate volume study is performed few weeks prior to the operation by acquiring cross-sectional images of the prostate at typically 5 mm intervals. The aim of the plan is to calculate the seed positions in three dimensions through-out the prostate which will deliver an envelope dose covering the whole prostate gland. A subset of seeds specified in the plan are loaded into implant needles and a stylet is placed behind every seed train. Each needle contains a train of seeds and is destined to go through one hole of the grid template to the patient s body according to the preoperative plan. The needle tip position is tracked in the ultrasound images and when it appears at the right position within the prostate, the needle is pulled back, depositing the seeds. In a brachytherapy procedure needles are inserted perpendicular to the US beam.

17 1.2. BACKGROUND 3 Hence, the non-uniform US main beam thickness (elevation beamwidth) causes inaccurate needle placements within the prostate and thus insufficient dose delivery to the target organ. In addition, the off-axis side lobe energies emitted apart from the US main energy would worsen the amount of localization error by showing the objects on the wrong spot in the images. The objective of this work is to study the clinical relevance of the needle tip placement error during the brachytherapy procedure in presence of the two main sources of errors, the US main and side lobe artefacts. 1.2 Background Goldstein et al. [8] first identified the problem of ultrasound elevation beamwidth artefacts. He observed its effects on the ultrasound images by using a diffusive flat surface placed into a water bath with the ultrasound beams intersecting to the diffusive surface. The inclination of the diffusive surface, scattering properties of an interface, and the transducer beamwidth are the three variables defining the amount of elevation beamwidth in his work. The elevation beamwidth artefacts were examined by several transducer types to observe the effects of elevation beamwidth along the lateral and elevational axes of the beam. He observed that the artefactual bands in the ultrasound images taken from within the ultrasound focal zones have higher intensities when compared to the ones from the near or far fields of the beam. Skolnick [9] compared both the scan plane and elevation beamwidth artefacts by sweeping a linear array US probe oriented 90 and 45 to his experimental phantoms. He used two phantoms for his experiments, one with multiple filaments located 1 cm apart from each other in a vertical row, and the other one with an inclined surface to the US beam. The ultrasound probes used in these experiments were

18 1.2. BACKGROUND 4 fixed by a wooden jig to ensure the beam makes proper inclination angle to the phantoms. He obtained close measurements of the elevation beamwidth artefacts when experimenting with both the mentioned phantoms. Richard [10] tried to measure the ultrasound elevation beamwidth profile using a test object with multiple inclined diffusive surfaces 15 mm apart from each other making 45 to the ultrasound beam. He observed the effects of electrical focusing on the amount of elevation beamwidth artefacts. Chen et al. [11] developed a technique to incorporate the ultrasound elevation beamwidth in improving the free hand ultrasound calibration precisions. He made use of a diffusive flat surface inclined 45 to the ultrasound beam to generate a beam profile over various depths of the beam. Ultrasound images of the profiling tool were acquired when it was deep in a water bath. Rubber membrane was the chosen diffusive material for his experiments since it is a good sound reflector in water. Laing et al. [13] reported the effects and importance of the side lobe artefacts on the ultrasound images. He illustrated the genesis of these artefacts and provided clinical examples of the commonly visualized side lobe artefacts. He employed a round plastic container filled with de-gassed water and scanned using different settings of the ultrasound device. This experiment repeated with a sponge included in the outside wall of the container. Diffuse echoes of moderate to high intensity were seen between the back wall of the container and specular side lobe artefact (low intensity echoes generated after intersecting side lobe energies with the objects in the medium). Only specular artefacts remained when the sponge was removed indicating that visible side lobe echoes could be generated from non-specular surfaces. In order to differentiate the effects of ultrasound main and side lobes, he scanned a phantom

19 1.3. THESIS OBJECTIVES 5 with the same settings of the transducer and observed difference in the amount of elevation beamwidth and side lobe effects. In vivo experiments carried out over highly curved, specularly reflecting surfaces, or highly reflective interfaces such as gas (urinary bladder, diaphragm, and gallbladder). Barthez et al. [15] reproduced the side lobe artefacts in vitro using different ultrasound transducers and tried to recognize these artefacts in vivo. He observed the side lobe effects when imaging a phantom composed of a water bath, a metallic wire, and a wooden tongue depressor. He imaged the phantom using all sorts of US probes (linear array, curved linear array, vector array, and sector mechanical transducer). The shape and intensity of the artefact varied with the transducer type. In vivo situations similar to phantom experiments were also investigated using clinical patients and compared to the observed artefacts in phantom experiments. In vitro artefacts were recognized in vivo when highly reflective objects were images adjacent to an anechoic region (e.g: urinary bladder wall and urine). There also have been several works to improve the US image quality and eliminate the off-axis US energies by implementing different adoptive beam forming methods on radio frequency data [17, 18, 19, 20, 21]. 1.3 Thesis Objectives This thesis studies the main and side lobe artefacts of ultrasound in prostate brachytherapy needle insertion under the following assumptions: The brachytherapy stepper and grid template guide are manufactured mechanically accurately.

20 1.4. CONTRIBUTIONS 6 The brachytherapy stepper does not diverge from its axis of displacement during the experiments. The imaging medium of the experiments is normal water at room temperature or water and glycerol solution. The measurements are based on observing the elevation beamwidth artefacts. The objective of this thesis is to measure the amount of localization error caused by the ultrasound elevation beamwidth effects and give relevant suggestions to reduce or diminish its effects in spatial target localizations. The following steps are set to achieve the objectives: Finding the best possible phantom design to measure the ultrasound elevation beamwidth making most of the accurately manufactured brachytherapy stepper. Generating ultrasound beam profile over various depths of the effective imaging region. Comparing the effects of main beam thickness and side lobe energies with each other. Providing suggestions through experiments to reduce or diminish the amount of uncertainties in ultrasound images. 1.4 Contributions The following contributions have been made towards achieving the objectives:

21 1.4. CONTRIBUTIONS 7 Designed two ultrasound beam profiling phantoms to measure the transrectal ultrasound elevation beamwidth over various depths of the imaging region. Both the phantoms are compatible with commercial brachytherapy steppers. One of the phantoms has a 45 inclined surface with respect to the ultrasound beam covered with rubber as a diffusive surface and the other with multiple wires. Generated ultrasound beam profiles for all positions along the axial and lateral axes of the ultrasound beam. Examined brachytherapy needle tip localization error for all the possible lateral and axial distances to the ultrasound transducer in presence of the main and side lobe energies. The measured error values are also compared to the beam profiles. Examined the effects of the ultrasound imaging parameters (gain, power, and dynamic range) on needle tip localization error in presence of the main and side lobe energies. Made feasible recommendations to minimize the effects of ultrasound elevation beamwidth. Measured the maximum amount of errors in needle insertion for a brachytherapy procedure. Identified the best region within the ultrasound image slices which are comparatively accurate in object localizations.

22 1.5. ORGANIZATION OF THESIS Organization of Thesis This thesis has been written in the manuscript-based style, as permitted by the Faculty of School of Graduate Studies at Queen s University. In the manuscript-based thesis, each chapter represents an individual research work that has been submitted, prepared for submission, or published in a peer reviewed publications or conferences. Each chapter represents a separate and complete research work in the sense that it includes an introduction to the work presented in that chapter, the methodology, results and discussion. Chapter 2 introduces the TRUS-Bridge phantom to measure the elevation beamwidth of a transrectal ultrasound probe. An ultrasound beam profile was presented representing the main beam thickness over all depths of the imaging region. A set of ultrasound images with elevation beamwidth effects also presented in this chapter. Chapter 3 presents a comparison between the artefacts generated due to the main beam thickness and the off-axis side lobe energies. The ultrasound beam profile used in this chapter was generated using the device presented in Chapter 2. Presented experimental results to indicate the possibility of needle tip localization error during an ultrasound-guided brachytherapy procedure due to the elevation beamwidth artefacts. A variety of ultrasound device settings have been used to observe their effects on the amount of artefacts. The experiments were carried out for laterally close and all possible axial depths to the ultrasound transducer. In Chapter 4 we present needle tip localization error measurement results for all possible lateral and axial positions of the ultrasound image slices (bounded to the brachytherapy grid template guide). These results were compared with beam profiles generated from a new version of the profiling tool (Wire-Bridge phantom).

23 1.5. ORGANIZATION OF THESIS 9 We show that the device imaging parameters have direct impacts on the amount of image artefacts and needle tip localization uncertainties. In Chapter 5 the theses concludes with a brief summary and outline of future work.

24 10 Chapter 2 Section-Thickness Profiling for Brachytherapy Ultrasound Guidance Introduction Ultrasound (US) is an attractive imaging modality because it is widely available, nonionizing, and can visualize in real-time almost any body tissue. However, due to the physics of the US beam propagation through the medium and image formation, the acquired US images may contain a certain artefact along the beam s elevational axis. This artefact commonly appears in anechoic areas [8] that hiders the identification of tissue structures and the ability to make appropriate medical diagnoses. 1 A version of this chapter has been presented: Mohammad Peikari, T. K. Chen, E. C. Burdette, and G. Fichtinger, Section-Thickness Profiling for Brachytherapy Ultrasound Guidance, the Society of Photo-Optical Instrumentation Engineers (SPIE) Medical Imaging, 2011, Orlando, USA (student paper award finalist)

25 2.1. INTRODUCTION The US Physics The transducer in an US machine sends and receives the sound waves. US sound waves propagate through the medium when excitation pulses are applied to each element (crystal) of the array transducer. In A-mode scanning, an ultrasound beam is propagated from one (or a group of) crystal(s) on the transducer. The transducer is then silent for a small fraction of time, waiting for return echoes from the medium before it generates the next pulse; this process continues by exciting each set of crystals on the US transducer. The information from each line of sight (a line of sight represents activation of one or a group of crystals in the transducer array) is then put together to form an US B-mode image. Focusing is a way to improve the resolution of an US B-mode image; the narrower the beam, the better the image quality. Focusing can be achieved in two ways, mechanically and electronically [4]. In mechanical focusing, US beams are passed through an acoustic lens which works in the same manner as an optical lens (external focusing), or by curving crystals which bends the sound waves toward a point in the tissue (internal focusing). Mechanical focusing is typically employed to improve imaging resolution in the elevation direction [4]. Electrical focusing however, is a more sophisticated technique which is based on using timing circuits and clocks, and alternative sequences of firing various crystals in a group by a small time delay. Electrical focusing is typically used to narrow the US beam in the lateral direction [4]. US image quality may be explained using three different factors: axial resolution (ability to distinguish between two reflecting points along the long axis of the beam), lateral resolution (ability to distinguish between two reflecting points perpendicular

26 2.1. INTRODUCTION 12 Figure 2.1: (a) US beam pattern in Axial, Lateral, and Elevation axes. (b) Axial resolution, (c) Lateral resolution, and (d) Elevation resolution. to the long axis of the beam), and elevation resolution (out-of-plane direction) [4] as depicted in Fig The effect of the section-thickness artefacts in the US elevation plane can be explained in detail using Fig Figure 2.2(a) shows the cross section of an unfocused beam pattern when it leaves a linear array transducer crystal. Figure 2.2(b) shows the A-mode echoes received by the transducer in presence of a reflecting material (human tissue). The first three reflected beams correspond to the three objects in the near-field located at the same depth (time) to the front end of the transducer. Since the strength of an US beam is highest at its center and decreases gradually toward the side edges[8], the maximum reflection intensity corresponds to point A which is located along the axial path of the beam. Reflection levels corresponding to points B and C are lower because they are close to the side edge of the beam pattern. Points D-F are located farther from the US beam and their reflection intensities follow the same concept, with the exception that, the overall intensity of the US beam decreases as it propagates through the medium. Hence point D has the highest

27 2.1. INTRODUCTION 13 Figure 2.2: (a)-(b) Beam pattern of the B-mode and A-mode linear array US with corresponding point reflectors[8]. (c) Axial, lateral and elevational axes convention with respect to the TRUS beam pattern. (d) Section-thickness estimation principle. reflection intensity and point F the least. The echoes reflected from objects at a known depth along the axial path and lateral position of the US beam will be absorbed simultaneously. The transducer sums all the echo reflections at a time and interprets its value as a single reflecting object located on the beam s central line[8]. This means, even if there is no reflecting object located on the central beam ray, the device assumes that the reflected echo corresponds to an object along the central portion of the beam pattern. This is depicted as point G in Fig. 2.2(b) and as an object tissue in Fig. 2.3 (a). The closer the reflecting object is to the central beam ray the less the section-thickness artefacts will be and vice versa[8]. Inspired by Goldstein s original principle and its variants[8, 9, 10], we have previously designed a Bridge phantom to measure the section-thickness for US freehand imaging and used the results to improve the calibration precision[11]. In this paper, we propose a practical and more precisely designed beamwidth profiling technology that is specifically tailored for the industry-standard brachytherapy transrectal US

28 2.1. INTRODUCTION 14 imaging Prior Art Goldstein[8] observed the effects of the section-thickness on the US images by changing the angle of the inclined surface for various experiments. In his work, he was not concerned about measuring the thickness of the artefacts on the US images. Richard[10], tried to measure the US section-thickness by using several inclined surfaces located successively below each other in a phantom to capture many section-thickness artefacts in a single image for low frequency probes. One challenge in his work was holding and maintaining the probe at a 45 angle during the experiments. Skolnick[9] used both an inclined surface and a phantom with multiple filaments located 1cm apart in a vertical row. He compared both the scan plane and elevation plane section-thicknesses by sweeping the linear array US probe oriented 90 and 45 to the filaments respectively. He obtained very close results when comparing the elevational section-thickness at specified depth using both the phantoms. He overcame the difficulty of holding the probe at the right pose by using a wooden jig The Clinical Brachytherapy Setup In a brachytherapy procedure (Fig. 2.3(c)), the TRUS device is inserted into the patient s rectum for imaging the prostate. Radioactive seeds are placed within sharp hollow needles, which are inserted into the prostate through the perineum one at a time. The position coordinates within the prostate are reported using real-time ultrasound-guided visualisation of the needle as it advances through the prostate. During the procedure, needle tips are visible as a bright spot within the black prostate

29 2.2. METHODOLOGY 15 tissue in the TRUS images. Needles are inserted at a preplanned position through the patient s perineum with the help of the grid mounted on the TRUS stepper. Needle tips are manually tracked by the surgeon as they advance through the prostate in TRUS images and when they appear at a preplanned position, radioactive seeds are deposited in the prostate. There are three important parts of a TRUS system to consider: a transrectal ultrasound probe, a stepper, and a grid template as shown in Fig. 2.3(b). The TRUS probe can move in and out of the patient s rectum allowing the brachytherapist to observe the prostate in live TRUS imagery during the procedure. The grid template located on the TRUS stepper guides the physician during the needle insertion process. The precisely machined stepper ensures that the TRUS probe always moves perpendicular to the grid template, and acquires TRUS images parallel to it. We modelled the US section-thickness, as the first step toward eliminating its effects on the TRUS images. Using the current TRUS stepper design as a frame work, we developed a simple phantom to estimate the ultrasound beam section-thickness. 2.2 Methodology The Concept We measured the section-thickness of the curvilinear transducer of a TRUS probe quantitatively using the approach described by Goldstein [8]. Figure 2.2(d) shows a cross-section of the beam pattern propagated from the TRUS curvilinear transducer intersected with an inclined diffusive material at 45 angle to the beam s central axis. Consider the diffusive material AB located at a known distance from the TRUS

30 2.2. METHODOLOGY 16 Figure 2.3: (a) Location of an object tissue (prostate) could be wrongly detected in presence of the US section-thickness[8]. (b) Different parts of the clinical TRUS device. (c) A schematic showing prostate brachytherapy procedure. transducer. As the sound propagates through the medium, it first intersects point A and the transducer would receive an echo wave sooner than any other point along the beam s path. The last echo would be generated by point B which is located at the farthest distance from the transducer. However, the TRUS machine assumes all the received echoes are from the reflectors located at the central beam ray. Therefore, the reflection of the diffusive line AB would be represented as line CD located on the central beam ray. As a result, the TRUS image would include a thick bright band with its height equal to the length of CD. Furthermore, since the diffusive material makes a 45 angle to the central beam axis, the length of CD is approximately equal to the TRUS elevation beamwidth as shown in Fig. 2.2(d). We designed the TRUS-bridge phantom such that the inclined plane makes a 45 angle to the vertical beam plane (Fig. 2.4(d)). The phantom was specifically designed to adapt on a standard brachytherapy system making use of the precisely machined encoded stepper. The phantom was equipped with stepper hole rods and side walls to prevent unwanted shakes and provide sufficient strength. The inclined surface of

31 2.2. METHODOLOGY 17 the phantom was made sufficiently long to show beamwidth artefacts in the TRUS images from all depths from the diffusive material. The phantom base also ensured that the diffusive surface always maintains its 45 angle to the probe. The supports to the inclined surface of the phantom eliminate the possibility of the diffusive surface bending, as shown in Fig. 2.4(a). The TRUS beamwidth artefacts were segmented manually by choosing a few points from the boundary of the bright band on the TRUS images. The selected points were chosen from the central region of the beamwidth so as to ensure distorted parts of the bright band on the image sides are not influencing the true beamwidth measurements. The images were analysed using a 2.33GHz Intel Core 2 Quad processor personal computer having a 3.25GB of RAM The Experimental Setup To validate our design, we tested our phantom on a commercial grade probe stepper (Fig. 2.4(b)) (Burdette Medical Systems, IL, USA) but the design can be easily adapted to any other stepper. The template grid was replaced by the TRUS-bridge phantom as shown in Fig A rubber membrane was used as a diffusive material (Fig. 2.4(e)) since its acoustic impedance is similar to that of water, making it a suitable sound reflector in water. The rubber membrane was held to the phantom using several rigid metal clamps (Fig. 2.4(f)), however, we aim to eliminate these in a future design.. The TRUS probe (Fig. 2.4(c)) can move back and forth to capture the returning echoes reflected by the diffusive surface. The distance from the TRUS transducer and the diffusive surface defines the depth at which the beamwidth is measured.

32 2.3. RESULTS 18 Figure 2.4: (a) CAD design of the TRUS-bridge phantom. (b) stepper, (c) transrectal probe, (d) TRUS-bridge phantom mounted on an industry-standard stepper, (e) rubber membrane on the inclined surface, (f) metal clamps, and (g)-(h) the experimental setup. The precisely machined TRUS stepper (Fig. 2.4(b)) holds the phantom in the right pose so that the inclined surface remains at a 45 angle to the diffusive surface. TRUS images were acquired at various depths when the whole system was inserted into a clear water bath as shown in Fig. 2.4(g). Images were acquired using the two available TRUS frequencies of 5MHz and 6MHz, as these are the frequencies normally used to observe the prostate during the brachytherapy procedures. 2.3 Results Using the TRUS-bridge phantom, we measured the elevation beamwidth at 29 different axial depths from the TRUS transducer. Fig. 2.5(a-b) illustrates a few examples of these measurements over various depths of the effective imaging region. The overall calculated beam profiles for the two operating frequencies are compared in Fig The measurements cover the axial depths from 27mm to 75mm from the TRUS transducer.

33 2.4. DISCUSSION 19 The beam in the elevation direction was focused closely to the transducer, i.e. the near-field of the US beam was relatively smaller than its far-field, e.g., at 5MHz the section-thickness starts with 3.16mm at 28.23mm axial depth and drops rapidly to 2.28mm at 29.13mm. Similarly at 6MHz the section-thickness starts with 3.04mm at 27.19mm axial distance and decreases to 2.59mm at 29.03mm axial distance from the TRUS transducer. This also indicates that the focal zone (around 29mm) corresponding to both the frequencies is located around the same axial depth. The beamwidth then starts diverging quickly after the focal zone to 5.6mm at 70.44mm and 6.1mm at 72.23mm for 5MHz and 6MHz frequencies respectively. The overall pattern of the beam profile for both the frequencies tends to follow the same shape and remains within a limited boundary since the two operating frequencies were chosen to be close to each other. Table 2.1 compares our results with previous experiments and shows that the focal length of different types of US probes may vary. 2.4 Discussion The TRUS device would provide a better elevational resolution within the focused beam region of the transducer. The beamwidth segmentation procedure was manual and time-consuming however, the beam profile curve could have been smoother if the beamwidth segmentation procedure was performed automatically over several hundred acquired TRUS images. We found that the TRUS device settings have direct impacts on the visibility of the section-thickness artefacts in the TRUS images. The thickness of the rubber membrane is negligible and hence the bright artefacts on the TRUS images are mainly due to the US section-thickness; for that reason, if the TRUS device settings (dynamic

34 2.4. DISCUSSION 20 Figure 2.5: (a) A subset of elevation beamwidth measurements. (b) Examples of section-thickness artifacts over different depths of the TRUS images, frequency= 6MHz. range, gain, and sector) are wrongly set, the correct boundary of the bright artefacts will not be visible in the TRUS images. There is no sharp focal zone narrowing in the beam profile in the elevation plane. This could be due to the size of the curvilinear array (around 38mm) and its aperture (around 8mm). Modifying the number of electronically set focal zones or the focal lengths in the image plane had no influence on the beam profile pattern. Focusing along the lateral axis of the image plane can only improve the quality of image formation (this observation is expected since the focusing along the elevation plane is achieved mechanically). In addition, the estimated beam profile represents a localization error map in the

35 2.4. DISCUSSION 21 acquired images. Incorporating this error map into the brachytherapy applications might give them the ability to consider the likelihood of position errors on the TRUS data which are otherwise treated uniformly in current practice. For example, according to the beam profiles, one should trust image regions near the elevation focal zone more than regions located at the greater depth where the localization uncertainty increases significantly with the growing section-thickness. An advantage of our design was using the TRUS stepper as a support to ensure that the US beam plane displacement maintains exactly a 45 angle through out the experiment. This eliminated all the challenges in maintaining the sound waves at a 45 angle to the inclined diffusive surface. The inclined surface supports also increase the reliability of the inclined surface to maintain the 45 angle during every step of the experiment. The jigged beam profile points (mainly after 40mm axial depth) are probably due to error in manual segmentation, bad image resolution where the correct boundary of the beamwidth artefacts are not clear, or the effects of the mechanical play of the device. Finally, our design is made to work with water as an imaging medium which differs from the speed of sound in the biological tissue (1540 m/s). Although this difference causes small measurement errors, more accurate results could be achieved by using a tissue-mimicking medium (e.g. Gel) in which speed and absorption of sound are similar to those in biological tissues.

36 2.5. FUTURE WORK 22 Figure 2.6: TRUS elevation beam profiles at 5MHz and 6MHz central operating frequencies. 2.5 Future work As future work, we plan to modify our TRUS-bridge phantom to be able to estimate the section-thickness of the TRUS curvilinear and sagittal transducers in successive experiments. Automating the process of beamwidth segmentation from a set of US images would help in finding a smoother beam profile curve from a few hundred US images. We plan to incorporate the estimated TRUS beam profiles (sagittal and curvilinear arrays) to improve the TRUS calibration accuracy. We also suspect that the non-uniform beam pattern in the elevation direction may have negative impacts on the target localization error (TRE) in ultrasound-guided surgery applications. In future work, we plan to investigate the relationship of the TRE and the beam profile.

37 2.6. CONCLUSIONS 23 Author Probe Type Probe Central Frequency Elevation Focal Depth (MHz) (mm) Goldstein[8] Linear array (long internal focusing) 3.5 Qualitative evaluation Linear array (short internal focusing) 3.5 No measurements reported Richard[10] Convex one-dimentional array 3.5 Qualitative evaluation Convex one and one-half-dimentional array 3.5 No measurements reported Phased array sector Skolnick[9] Curved array SP Curved array GP Chen[11] et.al Linear array Linear array Peikari et.al TRUS 5 29 TRUS 6 29 Table 2.1: Comparison summary of our results with previous experiments. 2.6 Conclusions To the best of our knowledge there has been no previous work done to measure the TRUS beam profile. We presented the TRUS-bridge phantom to measure the sectionthickness of a TRUS probe used in brachytherapy procedures. A set of TRUS images with the section-thickness artefacts was acquired using the phantom from various axial depths of the effective imaging region for the two available device frequencies. TRUS beamwidths were segmented manually by selecting a few points from the artefact s central portion to minimize the effects of distortion on the beamwidth measurement. A beam profile of the segmented beamwidths were generated for all possible depths for the two central frequencies of 5MHz and 6MHz. In addition, we took the advantage of the precise design of the TRUS stepper as a frame work to ensure the TRUS beam ray always collides at a 45 angle to the diffusive surface.

38 2.6. CONCLUSIONS 24 Acknowledgements This work was funded by the Natural Sciences and Engineering Research Council of Canada under the Idea to Innovation program. Gabor Fichtinger was supported as Cancer Care Ontario Research Chair.

39 25 Chapter 3 Effects of Ultrasound Section-Thickness on Brachytherapy Needle Tip Localization Error Introduction Ultrasound (US) imaging is ubiquitous in intra-operative surgical guidance. It has been discussed in the literature [8, 9, 10, 11, 12] that the US images may contain certain artefacts caused by the section-thickness (elevation beamwidth) of the beam, orthogonal to both the axial and lateral beam axes. These artifacts may conceal tissue 1 A version of this chapter has been accepted for publication: Mohammad Peikari, T. K. Chen, A. Lasso, and G. Fichtinger, Effects of Ultrasound Section-Thickness on Brachytherapy Needle Tip Localization Error, Medical Image Computing and Computer Assisted Intervention (MICCAI) Sept. 2011, Toronto, Canada

40 3.1. INTRODUCTION 26 Figure 3.1: (a) US beam pattern in axial, lateral, and elevation axes. (b) Axial resolution, (c) Lateral resolution, and (d) Elevation resolution. structures and may lead to incorrect medical diagnosis [8]. The motivating application for the present work is transrectal ultrasound (TRUS) guided prostate cancer brachytherapy, a procedure that entails the permanent implantation of small radioactive capsules through hollow needles into the patient s prostate in order to eradicate the cancer with radiation. During the brachytherapy procedure, the physician uses ultrasound slices to visualize the current position of the needle tip that appears as a bright spot in the image. Section-thickness is a unique problem for TRUS-guided brachytherapy because the needles are perpendicular to the ultrasound image slice. The nonuniform section-thickness causes error in localizing the needle tip and thus lead to inaccurate needle placement and ultimately to suboptimal deposition of the radioactive dose. The objective of this paper is to quantify the needle tip placement error in during the brachytherapy procedure. The quality of an US image primarily depends on three factors. Axial and lateral resolutions, which is the ability of the US device to distinguish between two structures along the axial and lateral direction, respectively. The third factor is the so called

41 3.1. INTRODUCTION 27 elevation resolution, where the US device assumes that all received echoes originate from structures situated precisely on the central line of the US beam [4], as depicted in Fig. 4.2 and in more details in Fig Figure 3.2(a) and (b) show the crosssection of an unfocused beam pattern and its corresponding A-mode echoes generated by reflecting materials in the medium. The first reflected beams, A-C, correspond to the three objects located at the same depth (D) from the transducer within the nearfield of the US beam. Since the strength of the US beam is at its peak at the center and it decreases gradually toward the side edges of the beam [8], the maximum echo amplitude (A) corresponds to point A on the central beam line and the minimum echo corresponds to point C near the side edges of the beam. The same concept applies to reflecting objects D-F, except that the overall US beam intensity decreases farther away from the transducer. Since echoes from the same axial distance and lateral position of the US beam are received by the transducer at the same time, they are absorbed simultaneously. All echoes received at the same time are summed and interpreted as a single object located on the US central beam line [8]. As a consequence, echoes from an object located along the side edge of the US beam appear to be originating from a non-existing virtual reflector on the central region of the US beam, shown as point G in Fig. 3.2(b). Putting it simply, reflectors along the side edge of the beam do not appear at their true position. The US section-thickness has been measured using phantoms incorporating inclined surfaces and multiple filaments [8, 9, 10, 11]. Recently, we have constructed a device to measure the section-thickness of side firing TRUS probes [12]. We used the same replica of this device for measuring needle localization and placement errors

42 3.2. METHODOLOGY 28 Figure 3.2: (a)-(b) Beam pattern of an unfocused B-mode and A-mode US linear array with corresponding point reflectors [8]. (c) Axial, lateral and elevational axes convention with respect to the TRUS beam pattern. (d) Section-thickness estimation principle. caused by the section-thickness of the TRUS beam in a standard clinical brachytherapy setup. To the best of our knowledge there has been no published work on this subject in the open literature. 3.2 Methodology Experimental Setup To generate a beam profile for the TRUS probe at 6MHz central frequency, we moved the probe back and forth (along the elevation axis) to acquire US images of the inclined-plane device (Fig. 3.3(a)). This is performed when the TRUS probe transducer and profiling phantom were inserted into a clear water bath. To observe needle tip effects (appeared as bright spots in Fig. 3.4(c)) on the US images, 6 needles were inserted through holes of a brachytherapy stepper grid template as shown in Fig. 3.3(b). The needles were placed at the grid s central

43 3.2. METHODOLOGY 29 Figure 3.3: (a) The beam profiling phantom mounted on a Burdette Medical Systems stepper, (b) Needle insertion setup, (c) Needle offset measurement principle. holes to ensure the distorted parts of the image (along the sides) does not have any influence in our measurements. For each needle, we moved the TRUS probe back and forth (along the elevation axis) in a water bath until the reflection from the needle tips appeared on the US image as they intersect with the US beam s boundary. The corresponding probe depths were then recorded for further analysis US Beam Profile Estimation To have an estimate of localization error for a TRUS probe, the US main beam thickness for all depths of the US beam (effective imaging region) from the transducer is found. We do this for the curvilinear transducer by using the same principle first explained in [8]. According to this approach (Fig. 3.2(d)), as the US beam propagates through the medium it first intersects with point A (nearest point to the transducer) on the inclined diffusive surface (45 to the beam). Similarly, the last point with which the US beam intersects would be the point B (farthest point to the transducer). The sound echoes return to the transducer. However, since the US device assumes all

44 3.2. METHODOLOGY 30 the echoes received to the transducer along the elevation axis of the beam are from the structures on the beam s central line, the line AB would be assumed as line CD to the device. Hence the TRUS image would include a thick bright band which the thickness approximately represents the US main beam thickness. A set of TRUS images with different imaging parameters (gain, dynamic range, and power) were collected at 6MHz for all depths of the imaging region. The gain setting of the US modifies the amplification of the signals during the signal processing step, dynamic range defines the ratio of the highest to smallest signal that can be accommodated by a system component, and power adjusts the intensity of the beam transmitted into the medium. The artefacts (bright bands) were segmented manually from the images and their distances to the US transducer (position of the band) were taken as the depth measurements. A beam profile of US main beam thickness versus axial distance from the transducer is then plotted as shown in Fig. 3.4(b) and Fig. 3.5(a),(c), and (e). Figure 3.4(a) shows a subset of these measurements Needle Offset Measurement on US Images To calculate the needle insertion offsets for every experiment, we set the closest needle to the transducer (needle No. 1) as the reference needle and subtract the depths of the other observed needle positions from that of the reference needle as shown in Fig. 3.3(c). The subtracted values represent the amount of divergence or convergence of the beam pattern with respect to the reference needle. Hence the section-thickness relationship between every two inserted needles is defined as: B(j) = ((N(i) N(j))) 2 + B(i)

45 3.3. RESULTS AND DISCUSSION 31 Figure 3.4: (a) A subset of elevation main beam thickness measurements from 28mm to 90mm axial distance, at gain= 55%, dynamic range= 78dB, and power= 0. (b) TRUS elevation beam profile at 6MHz central operating frequency from 10mm to 65mm axial distance, at gain= 55%, dynamic range= 78dB, and power= 0. (c) Needle tip appearance in a TRUS image at gain= 0%, dynamic range= 15dB, and power= 7. Where i and j are the two axial depths where the needles are inserted; N(i) and N(j) are the needle insertion depths at axial depths i and j respectively; and B(i) and B(j) are the US main beam thickness at the corresponding i and j axial depths respectively. 3.3 Results and Discussion In order to observe the effects of the US device imaging parameters on the beam pattern we performed a series of needle insertion tests for 27 combinations of US gain (0, 50, and 100%), dynamic range (15, 50, and 100 db) and power (0, -4, and -7). The results are shown in Fig. 3.5 and Table 3.1. According to Fig. 3.5(b)-(d), when gain=50% or gain=100%, the needle offsets are monotonously increasing, which indicates the beam diverges constantly. However, this

46 3.3. RESULTS AND DISCUSSION 32 is not the case when comparing with the beam profile generated at the same settings (Fig. 3.5(a)-(c)). The beam profile pattern shows that the US beam converges up to a focal point and diverges right after that. This contradiction in the beam profile and needle insertion offset plots could be because of the US side lobes artefacts. Side lobes consist of multiple low-intensity off-axis ultrasound beams that produce image artefacts due to the error in localizing the returning echoes within the main US beam [4, 13]. If a highly reflective structure is encountered, it will be wrongly positioned in the image along the main US beam [4]. When the gain is set to high, the energy assigned to the US side lobes increases and hence their effects on the TRUS images increase. Hence, during needle insertion, the needle tips first intersect with the side lobe energies and their echo artefacts are shown as if they are intersected with the main US beam. This clearly shows that the US main lobe thickness and the side lobe artefacts together might have large effects on localizing needle tips and objects within the TRUS images. On the other hand, when gain=0%, the side lobe energies are set to be minimum and the needle offsets are less than zero (Fig. 3.5(f)) which indicates that the beam constantly converges up to a focal zone (around 30 mm from the transducer). The beam pattern starts diverging quickly after the focal zone which matches the main beam thickness measurements using the profiling phantom shown in Fig. 3.5(e). This indicates that the US section-thickness is small and hence the section-thickness must not introduce much error in the images captured in this setting. On the other hand, when gain is set to zero, the amount of false reflections due to the US side lobes are minimized and hence the needle tip reflections on the TRUS images are ensured to be due to intersecting the needle tips and US main beam only.

47 3.4. CONCLUSION 33 Figure 3.5: (a) US beam profile for gain= 100%, dynamic range= 50dB, and power= 0. (b) Average and std. of N(j) N(1) for gain= 100%. (c) US beam profile for gain= 50%, dynamic range= 50dB, and power= 0. (d) Average and standard deviation of N(j) N(1) for gain= 50%. (e) US beam profile for gain= 0%, dynamic range= 50dB, and power= 0. (f) Average and std. of N(j) N(1) for gain= 0%. 3.4 Conclusion To the best of our knowledge there has been no previous work to examine the effects of imaging parameters and the US section-thickness on needle insertion depth estimation error. The US section-thickness is the combination of the both main sound-energy lobe (the main beam) and the side energy lobe. The side lobe artefacts maybe an important issue to be addressed during needle insertion procedures since they may

48 3.4. CONCLUSION 34 Gain (%) Dyn. Range (db) Power N(j)-N(1) (mm) j=1 j=2 j=3 j=4 j=5 j= Table 3.1: Summary of the needle insertion offsets for different US settings. introduce further localization errors beyond the main beam thickness artefacts. Both the beam profile and the needle insertion experiments have provided evidence that the high-gain in US imaging setting would increase the side lobe energy of the US beam. This could result in a large elevation section-thickness profile, which directly leads to larger errors in needle insertion for TRUS-guided brachytherapy. It is highly recommended to reduce the gain of the US imaging device to as low as practically possible to suppress the side lobe-introduced section-thickness, which would effectively minimize the needle insertion depth estimation errors (up to around 5 mm). The US beam profile also could help the surgeon in considering the likelihood of position errors during needle insertions.

49 3.4. CONCLUSION 35 The proposed technology is indeed tailored to brachytherapy, but the underlying principle applies to three-dimensional localization in US imagery, in general, as the US section-thickness is inherent to the modality. Side-lobes artefacts are present in every application where a needle (catheter, etc.) penetrates the US beam in the elevational direction which occurs quite ubiquitously. Acknowledgements This work was funded by the Natural Sciences and Engineering Research Council of Canada under the Idea to Innovation program. Gabor Fichtinger was supported as Cancer Care Ontario Research Chair. Thomas K Chen was supported as a MITACS Accelerate PhD Fellow.

50 36 Chapter 4 Characterization of Ultrasound Elevation Beamwidth Artefacts for Brachytherapy Needle Insertion Introduction Clinical Significance and Background Prostate cancer is the second leading cause of cancer death in men in the United States and one of the most common cancers in North America [1]. Prostate brachytherapy is an effective treatment for early-stage prostate cancer which involves permanent implantation of small radioactive seeds into the prostate gland under real-time transrectal ultrasound (TRUS) guidance to the cancer. The success of the procedure 1 A version of this chapter is submitted: Mohammad Peikari, T. K. Chen, A. Lasso, T. Heffter, E. C. Burdette, and G. Fichtinger, Characterization of Ultrasound Slice Thickness Artifacts for Brachytherapy Seed Implantation, Journal of Medical Physics.

51 4.1. INTRODUCTION 37 depends on an accurate seed implantation. Errors in implant locations could result in dosimetric degradation of the treatment. The patient is placed on the lithotomy position and TRUS probe is placed inside his rectum, as shown in Fig Transperineal implant needles are inserted through the guide holes of the template in planned locations. The needle tip position is tracked by TRUS. When inserted to the correct cancer site in the prostate, the needle is pulled back while depositing the seeds [2, 3] Elevation Beamwidth Artefacts Image artefacts are commonly encountered in medical ultrasound images. It has been discussed in the literature that US may contain certain types of artefacts caused by the beam orthogonal to both the axial and lateral beam axes, the main beam thickness, and the side lobes. The artefacts may lead to a misinterpretation of the true tissue structure [8]. The target application of this research is prostate brachytherapy (Fig. 4.1). The main beam thickness of US is a unique problem of TRUS-guided brachytherapy because the needles are inserted perpendicular to the US beam. Hence, the non-uniform main beam thickness causes error in localizing needle tips which in turn may lead to in-accurate needle placements and dose delivery. In addition, the off-axis side lobe echoes could also worsen the amount of localization error by showing the off-axis objects in incorrect location in the US images. Side lobes result from width and length mode vibrations, and immediate reverberation at crystal-tissue interfaces [4]. The objective of this work is to study the clinical relevance of the needle tip placement error during the brachytherapy procedure in presence of the two main source of

52 4.1. INTRODUCTION 38 Figure 4.1: (a) A schematic showing prostate brachytherapy procedure. (b) Implanting needles inserted through grid template holes. errors, the main and side lobe artefacts of the US. The US image is created by interpreting the intensities of the reflected echoes when intersected with an object tissue located at a known distance to the transducer. It is generally assumed that the US image is of zero thickness which contradicts the fact that the US beam can only be mechanically focused at a depth resulting in a finite, non-uniform elevation beamwidth. The quality of the US image along the beam s elevation axis is defined by the elevation resolution (Fig. 4.3), in which the US device assumes that all the received echoes originate from structures located precisely on the central line of the US beam [4, 8] as depicted in Fig. 4.2(b). Figure 4.2 (b) and (c) show the cross-section of a focused TRUS beam pattern and its corresponding A-mode echoes. Returning echoes are generated after US beam intersects with reflecting materials located within the beam s boundary. The first three reflected echoes correspond to the reflecting objects (A-C) located at the same axial distance. Since the strength of the US beam energy is maximum near the central

53 4.1. INTRODUCTION 39 Figure 4.2: (a) US beam pattern in axial, lateral, and elevation axes. (b)-(c) Beam pattern of a B-mode and A-mode TRUS with corresponding point reflectors. line of the sound wave and decreases elevationally [4, 8], the maximum echo amplitude corresponds to the reflecting point A and the minimum amplitude echo corresponds to point C. The US energy intensity degrades as the beam travels through the medium farther away from its source (transducer). Hence, the overall intensities of the reflected echoes from points D and F decreases as shown in Fig. 4.2(c). Echoes from the same axial distance and lateral position of the US beam arrives to the transducer at the same time. US device sums all the simultaneously received echoes and interpret their values as a single object located on the US central beam line. As a result, echoes receiving from any other objects not located on the US central beam line will be considered to be from the objects located on the central beam line, shown as point F in Fig. 4.2(b). Therefore, reflectors located elevationally farther from the beam central line do not appear at their true position. Transducer side lobes consist of multiple low energy sound beams emitted off-axis

54 4.1. INTRODUCTION 40 from the main US lobe that produce image artefacts due to error in positioning the returning echo as shown in Fig Although the US beam pattern is generally considered to be fairly coherent within the near field, in reality, the beam diverges quickly without focusing [4, 5, 6, 7, 13, 16]. Acoustic lenses are therefore used to improve the lateral and elevational image quality by focusing the sound wave at a specific axial depth [4, 5, 12]. The beam pattern, therefore, consist of a main lobe which looks like an inverted cone with its apex located at a known distance (focal depth) from the transducer face. In addition to the main lobe of sound, unwanted parts of the US beam are also produced which occur outside of the main beam (Fig. 4.3(a)). Theoretical calculations of the side lobe intensities suggest that the energy of the side lobe beams are at most one-hundredth of the main US beam (20 db lower) [13]. This level of intensity is however sufficient to produce enough echoes (side lobe artefacts) from the strong reflectors located off the main beam [4]. Side lobe artefacts appear as a series of parallel low intensity lines at regular intervals [4, 13, 14] near strong, curved, and highly reflective surfaces [13]. The US device assumes that the depth at which an echo is displayed is proportional to the time it takes for the echo to leave the transducer, reflect off the reflective surface, and return to the transducer. Side lobe artefacts occur when the off-axis beams interact with highly reflective acoustic surfaces. Hence, the returning echoes to the transducer are recorded to be along the path of the main US beam as shown in Fig. 4.3(b)-(c).

55 4.1. INTRODUCTION 41 Figure 4.3: (a) Two-dimensional representation of the main and side lobe beams. (b) Diagram showing the off-axis side lobe energies encountering two objects. (c) The US device assumes that the returning echoes from the off-axis side lobes came from the main beam and misplaces the structure Prior Art in Elevation Beamwidth Characterization Goldstein et al. [8] was the first to examine the effects of the US elevation beamwidth on the images by using an inclined plane. He observed the elevation beamwidth artefacts on different human organ geometries using longitudinal and transverse scan of the tissues. Richard [10] later tried to measure the beam thickness using several parallel inclined surfaces at 45 located 15 mm below each other to capture many elevation beamwidth artefacts in one single image for low frequency probes. Skolnick [9] compared both the scan plane and elevation beamwidth artefacts by sweeping a linear array US probe oriented 90 and 45 to his experimental phantoms. He used two phantoms for his experiments, one with multiple filaments located 1 cm apart from each other in a vertical row, and the other one with an inclined surface to the US beam. He obtained close measurements of the elevation beamwidth artefacts when

56 4.2. MATERIALS AND METHODS 42 experimenting with both the mentioned phantoms. Laing et al. [13] first reported the effects and importance of the side lobe artefacts on the US images. He illustrated the genesis of these artefacts and provided clinical examples of the commonly visualized side lobe artefacts. Barthez et al. [15] reproduced the side lobe artefacts in vitro using different US transducers and tried to recognize these artefacts in vivo. He observed the side lobe effects when imaging a phantom composed of a water bath, a metallic wire, and a wooden tongue depressor. He imaged the phantom using all sorts of US probes (linear array, curved linear array, vector array, and sector mechanical transducer). There also have been several works to improve the US image quality and eliminate the off-axis US energies by implementing different adoptive beam forming methods on radio frequency data [17, 18, 19, 20, 21]. 4.2 Materials and Methods US Beamwidth Calculation Principles In this work we refer the US elevation beamwidth artefact as a combination of the main lobe thickness and the low-intensity side lobes. We also refer the US beamwidth artefact depth to be the axial distance from the US transducer face to the middle of the bright artefact on the images. We measure the US beamwidth using the same approach originally proposed by Goldstein [8]. According to his approach, if we image a 45 oriented diffusive surface to the US beam, the US beamwidth is the thickness of the resultant artefact on the US image. Figure 4.4(a) shows the cross section of an US main beam intersected with a

57 4.2. MATERIALS AND METHODS 43 diffusive inclined wire/surface along the elevation axis of the beam. As the US beam propagates through the coupling medium, it first intersects with point A (closest point to the transducer) on the inclined diffusive material. The inclined wire and the US beam has to maintain 45 with respect to each other. The last point with which the US beam intersects with is point B (farthest point to the transducer). The sound echoes from the inclined wire return to the transducer. Where the US device sums all the simultaneously returned echoes assuming they are actually from the reflectors located on the central beam axis. Hence, the line AB would be displayed as line CD by the US device, as shown in Fig. 4.4(a) with its thickness approximately represents the US beamwidth. If there are side lobe energies present around the US main lobe, the corresponding returned echoes to the transducer are considered to be the echoes returned after intersecting the US main lobe to the reflecting materials. Hence, the same concept applies to categorizing the returned echoes by the US device as explained before with the exception that the US device would include a thicker bright artefact in the images because the side lobe energies were the first to intersect with the inclined diffusive material. The side lobe generated echoes will have lesser intensities since the side lobe energy levels are much lesser than the main lobe energies as shown in 4.4(b) Wire-Bridge Phantom We have previously developed a device (the TRUS-Bridge phantom) to measure and quantify the elevation beamwidth of TRUS probes [12]. The TRUS-Bridge phantom had an inclined surface oriented 45 angle to the US main beam emitted from the transducer. A rubber membrane was used to cover the inclined surface since the latex

58 4.2. MATERIALS AND METHODS 44 Figure 4.4: (a) Ultrasound beamwidth measurement principle in presence of the main lobe only. (b) Ultrasound beamwidth measurement principle in presence of the main lobe and side lobe energies. material can make a good reflector in water. In this work, we used a modified version of the TRUS-Bridge phantom, where we replaced the rubber membrane with nylon wires (Wire-Bridge phantom). We used multiple nylon wires in place of the rubber material to ensure the measurements are not influenced by the thickness of the rubber membrane. Thirteen wires were placed at the same lateral position as the template grid holes, as shown in Fig. 4.5(b). The Wire-Bridge phantom was specifically designed to accommodate any standard commercial brachytherapy stepper. The phantom was made of high-strength, industry-grade thermoplastic and equipped with supporting side walls and screw holes to be rigidly affixed onto the stepper on the template holder, which ensures the geometric accuracy of the incline wires.

59 4.2. MATERIALS AND METHODS 45 Figure 4.5: (a) Different parts of the clinical TRUS device. (b) The Wire-Bridge phantom on a clinical TRUS stepper Needle Localization Error Measurements We calculated the needle localization error by first measuring the boundary of the US beam where the first needle tip reflections are starting to appear. The brachytherapy needles are inserted and fixed through the grid template holes. The TRUS probe is then moved back and forth along the elevation axis of the beam until the inserted needle tip reflection starts to appear in the TRUS images. We then recorded the brachytherapy stepper position reading at which the needle tips are first observed in the TRUS images along the axial axis of the US beam for each inserted needle. The observed needle positions are then compared to a reference point. The reference point in our experiments is the point at which the first inserted needle reflection appears in a vertical column (corresponding to the row number 1 on the grid template guide, Fig. 4.7(b)). This gives us an estimate of the needle tip localization offset with respect to the US beam shape. We measure this offset at each template hole where the needle is inserted (both laterally and axially). The depths of other observed inserted needle tips at other axial positions corresponding to the exact position of the

60 4.2. MATERIALS AND METHODS 46 template grid holes are subtracted from the reference point to give an estimate of the needle offset with respect to the reference needle. Hence, the beamwidth relationship between every two inserted needles is defined as follow: B(j) = ((N(i) N(j))) 2 + B(i) Where i and j are the two inserted needle indexes being compared along axial depths; N(i) and N(j) are the needle insertion depths for the two needles, i and j respectively; and B(i) and B(j) are the beam thicknesses at the corresponding i and j axial depths respectively, as shown in Fig When there are side lobes, the needle tips intersect with these energies sooner than they intersect with the main lobe. As mentioned before, the US device assumes that the echoes from side lobe energies are from the reflectors intersected with the main lobe and it positions these echoes within the main beam echoes. Hence, in this case the needle tip reflections appear sooner than when intersecting with the main lobe as depicted in Fig. 4.6(b) Hardware Configurations There are three parts of a TRUS system to consider: a TRUS probe, a stepper, and a grid template guide as shown in Fig. 4.5(a). The TRUS probe can move in and out (along the elevation axis of the beam) of the patient s rectum allowing the physician to observe and track the prostate in live TRUS imagery during the procedure. The grid template on the TRUS stepper guides the physician during the needle insertion process. The stepper allows the TRUS probe always to move perpendicularly to the grid template and the US beam.

61 4.2. MATERIALS AND METHODS 47 Figure 4.6: (a) Needle offset measurement principle in presence of main lobe only. (b) Needle offset measurement principle in presence of main lobe and side lobe energies Experimental Setup In order to examine the effects of the US imaging parameters on the beam pattern and localization error we performed a series of needle insertion tests for 27 combinations of US gain (0, 50, and 100%), dynamic range (15, 50, and 100 db) and power (0, -4, and -7). A total of 169 needles inserted with the guidance of grid template mounted on the brachytherapy stepper for each of the 27 combinations. We examine a wide variety of cases to have an estimate of localization errors for different lateral positions of the beam. Table 4.1 shows the combinations of different TRUS imaging settings used to acquire images in our experiment. Please note that in the chapter 3, we examined 27 different combinations of imaging parameters for one vertical column (corresponding to the label D) of the grid template guide whereas in this chapter, these parameters where examined for all columns of template guide.

62 4.2. MATERIALS AND METHODS 48 Dynamic Range (db) Gain (%) Power Table 4.1: Combinations of different TRUS imaging settings used in our experiments. US Beamwidth Measurement We used Wire-Bridge phantom to measure beam profiles at the lateral locations that exactly correspond to the columns of the template grid holes. The phantom contains 13 parallel inclined wires positioned 5 mm apart according to the template hole lateral positions. Several TRUS images were acquired with different imaging parameters (gain, power, and dynamic range) at the central frequency of 6 MHz (the frequency at which brachytherapy procedures are being operated), as shown in Fig. 4.8 (b). The phantom was replaced by the grid template and fixed rigidly using screws to the template holder. TRUS images were acquired in a bath of distilled water and 7% glycerol by weight (Fig. 4.8(a)). The glycerol-doped medium provides a speed of sound of 1540 m/s which matches the speed of sound in human tissues [8, 22]. Needle Tip Localization Analysis We used the grid template to guide needle insertion in our experiments. The template contains a matrix of dimension of holes. The size of the grid matrix is 60

63 4.2. MATERIALS AND METHODS 49 mm by 60 mm and there is a 5 mm gap between its rows and columns. The distance from the TRUS probe to the first row of guide holes (labeled with 1 in Fig. 4.7(b)) was 10 mm. Using the template grid helped to ensure that the needles are inserted exactly perpendicular to the US beam along its elevation axis. The TRUS probe was moved back and forth along the elevation axis of the beam until the first needle tip reflection appeared in the TRUS images, the stepper position was then recorded for that elevational depth, as shown, in Fig. 4.7(a). Needle tip reflections appeared as bright spots in the US images as shown in Fig. 4.7(c). This needle tip measurement was repeated for all 169 holes of the grid template to estimate the localization error for all possible needle insertions during a brachytherapy procedure. Different imaging parameters (gain, power, and dynamic range) were used for these experiments at central frequency of 6 MHz to see the effect of these parameters on observing the location of the needle tips in the TRUS images. Comparison Between Beam Profile and Needle Tip Profile In order to compare the US beam profile and the needle tip profile, we conducted several experiments with similar imaging settings (gain, power, and dynamic range). For the beam profile measurements, we kept the device s dynamic range and power at a constant value and set the gain to three values to cover the whole range of possible gain values (from low to high gain). We did not conduct the same experiments for different dynamic range and power values since we had observed that different settings of these two parameters did not change the overall measurement of the beamwidth. Furthermore, we experimented with all possible combinations of the imaging parameters, since the needle tip profile measurement was the most important part of

64 4.2. MATERIALS AND METHODS 50 Figure 4.7: (a) The needle insertion experimental setup. US probe and needles are deep into a glycerol-water bath. (b) Standard grid template used to guide needle insertion in brachytherapy procedures. (c) Needle tip appearance in a TRUS image at gain= 0%, dynamic range= 15 db, and power= -7. the experiments. The results of all the experiments were combined by taking the mean and standard deviation of the values at similar settings. Side Lobe Extraction The inclined wires of the phantom cover the entire elevation axis of the beam. Hence, the inclined wires can cover both US main and side lobes. A bright spot with highest intensity appears in the images when the main US beam intersects with the inclined fishing wires. In addition, a group of thin, parallel, and low intensity bands appear on both sides of the main lobe artefacts after side lobes intersect with the wires, as shown in Fig 4.14(c).

65 4.3. EXPERIMENTS AND RESULTS 51 Figure 4.8: (a) The beam profiling experimental setup. The US probe and Wire- Bridge phantom are submerged in a glycerol-water bath. (b) Examples of beamwidth artefacts over different depths of the TRUS images, frequency= 6 MHz, gain= 50%, dynamic range= 50 db, and power= Experiments and Results US Beamwidth Measurement Using the Wire-Bridge phantom, we measured the US elevation beamwidth at different axial depths from the TRUS transducer. TRUS images were acquired with 6 MHz of central frequency, at 90 mm of imaging depth, 50 db of dynamic range, and different gain settings (0, 50, and 100%). Figure 4.8(b) illustrates a few examples of these measurements over several depths of the imaging region. The bright wire reflections in the TRUS images along with their low intensity side lobe artefacts (if present) were manually segmented by selecting points from the boundaries of the artefacts. A plot of linear interpolation of the US

66 4.3. EXPERIMENTS AND RESULTS 52 beamwidth versus axial distance to the probe transducer (beam pattern/profile) were presented for each wire reflections in the images (a total of 13 lateral positions), as shown in Figs The US beam pattern for each lateral locations of the images corresponding to the grid template holes were considered to find the relationship between the needle localization offsets and the beam pattern. The measurements cover the axial depths from 10 mm to 90 mm from the TRUS transducer face. All the TRUS images were acquired at 6 MHz central frequency, the frequency commonly used in the usual brachytherapy procedures. Elevation Beamwidth With Respect To Imaging Settings The following observations are made when comparing the elevation beamwith with respect to TRUS imaging settings: The beam in the elevation direction had much smaller beamwidth when the gain was set to minimum as shown in Figs The beamwidth increased rapidly when the gain was increased to 50% and 100%. This rapid change was due to the increase in side lobe energies when gain was set to a higher value. The beamwidth is less than 4.5 mm when gain=0% between mm axial distance, but more than 14 mm when gain is set to a high value (gain=50% or gain=100%). Hence, the object localization is minimum in the case when TRUS gain is set to 0%, as indicated by blue lines in the beam pattern plots (Figs ). The US beamwidth is approximately mm at a distance of mm to the

67 4.3. EXPERIMENTS AND RESULTS 53 transducer face when gain=0% and remains low upto mm at the distance of mm to the transducer. This value is approximately mm and mm when gain=50% and 100% respectively close to the transducer (13-15 mm) and increases up to mm and mm farther from the transducer (75-78 mm). Please note that there might be manual segmentation errors of a couple of mm due to the mistake in recognizing the true artefact boundaries. Elevation Beamwidth With Respect To The Lateral Positions Of The Template Grid The following observations are made when comparing the elevation beamwith with respect to lateral positions corresponding to the locations of the holes on the template grid guide: The beam for all the lateral positions of the US beam follows almost the same pattern. The beam starts converging first, when gain=0%, up to a focal point where the beam is the narrowest, and then starts diverging as the beam travels axially through the medium at farther distance from the transducer. When gain is set to 50% or 100% the effects of side lobe energies increases and there will be low intensity echoes generated after these energies intersect with wires (shown in Fig. 4.8 (b)). Hence, the US beamwidth increases constantly without converging to a focal point as shown in the beam profile measurement figures

68 4.3. EXPERIMENTS AND RESULTS Needle Tip Localization Analysis Observing the needle tip localization plots the following results can be reported: According to Figs , when gain=50% or 100%, the needle offset graphs are increasing monotonously, which indicates the US beamwidth increases as the axial depth increases. This increase in the beamwidth is due to the increase in the main lobe thickness and the oriented side lobe energies which diverge as we get farther from the transducer. The needle insertion error is minimum close to the transducer since the side lobe energies are not far apart from the main energy beam and the needle tip intersects to the side lobes when they are closer to the actual main beam. The beam profile pattern also agrees with this finding. According to the beam profile graphs, when gain is turned high, the US beamwidth increases at least 2 times more than the needle offset values. This could be explained using the US beamwidth relationship equation introduced in section and Fig. 4.4(b). As explained before, when a highly reflective material is intersected with the off-axis side lobes, the returning echoes are wrongly placed within the main US beam as if the object is intersected with the main beam [13]. Therefore, during needle insertion, the needle tips are first intersect with the side lobe energies and the first echo reflections appear as if the needles are intersected with the main US beam. This clearly indicates that the presence of both the main and side lobe beams have large effects on localizing needle tips in the TRUS images.

69 4.3. EXPERIMENTS AND RESULTS 55 On the other hand, when gain=0%, side lobe energies are minimized or diminished and the needle offsets are less than zero near the transducer. This means that the beam monotonously converges up to a focal zone (around mm from the transducer) where the beam thickness is minimum (around 1.5 mm). The beam thickness diverges quickly right after the focal zone upto 4 mm at 78 mm axial distance to the transducer surface. This indicates that the US beam thickness is much smaller when gain is set to zero than the case where gain is set to be high. Hence, the amount of localization error decreases when gain is turned down. Looking at the needle offset diagrams at gain=0%, when the needles are inserted at a location laterally farthest from the center of grid template (labeled with D on the template), needle offsets are greater than when close to the center. The maximum needle offset starts from approximately 2.3 mm at 70 mm distance from the reference needle and decreases to 1.3 mm as we move laterally towards the central portion of the grid template. The offset values start increasing as we move laterally towards the other side of the grid template (labeled as G ). The over all offset values laterally located away from the central portion of the grid template is also proportionally greater than the offset values captured along the central portion. This result indicates that the localization errors are higher when considering the holes located farther away from the center of the grid template and we will get more accurate localizations along the central portion of the grid.

70 4.3. EXPERIMENTS AND RESULTS 56 In addition, the US beam is narrower along the grid central holes. When gain=0%, the offset values within the first mm from the reference needle are negative indicating that the beam was narrower with respect to the reference needle (as shown in Fig. 4.10(d)(f), Fig. 4.11(b)(d)(f), and Fig. 4.12(b)). Therefore, localization errors are minimal as we move laterally towards the central portion of the grid template and when inserting needles axially within 40 mm distance to the TRUS transducer Comparison Between Beam Profile and Needle Tip Profile As explained before, side lobes consist of multiple low-intensity off-axis beams that produce image artefacts around the main US lobe [4, 13, 16]. When the gain is set to high, the energy assigned to the US beams (main and side lobes) increase and hence their effects on TRUS images increase. These effects include, images with higher intensities, noise levels, and more main and side lobe artefacts. Comparing the elevation beamwidth and needle tip measurement profiles we can make the following points: When gain=0%, the amount of energies assigned to main and side lobes are minimum, hence the artefacts caused by high energy beams are diminished or minimized. When gain=50% or 100%, energies assigned to the US beam increases and there will be more energies assigned to the side lobe beams. Hence, side lobe artefacts may be generated when these energies intersect with any object in the imaging

71 4.3. EXPERIMENTS AND RESULTS 57 Figure 4.9: (a)(c)(e) US beam pattern at various lateral positions labeled with A, a, and B on the grid template for different gain values at dynamic range=50 db, and power=-4. (b)(d)(f) Mean and Std. of N(j) N(1) at various lateral positions for different gain, dynamic range, and power settings.

72 4.3. EXPERIMENTS AND RESULTS 58 Figure 4.10: (a)(c)(e) US beam pattern at various lateral positions labeled with b, C, and c on the grid template for different gain values at dynamic range=50 db, and power=-4. (b)(d)(f) Mean and Std. of N(j) N(1) at various lateral positions for different gain, dynamic range, and power settings.

73 4.3. EXPERIMENTS AND RESULTS 59 Figure 4.11: (a)(c)(e) US beam pattern at various lateral positions labeled with D, d, and E on the grid template for different gain values at dynamic range=50 db, and power=-4. (b)(d)(f) Mean and Std. of N(j) N(1) at various lateral positions for different gain, dynamic range, and power settings.

74 4.3. EXPERIMENTS AND RESULTS 60 Figure 4.12: (a)(c)(e) US beam pattern at various lateral positions labeled with e, F, and f on the grid template for different gain values at dynamic range=50 db, and power=-4. (b)(d)(f) Mean and Std. of N(j) N(1) at various lateral positions for different gain, dynamic range, and power settings.

75 4.3. EXPERIMENTS AND RESULTS 61 Figure 4.13: (a) US beam pattern at the lateral position labeled with G on the grid template for different gain values at dynamic range=50 db, and power=- 4. (b) Mean and Std. of N(j) N(1) at various lateral positions for different gain, dynamic range, and power settings. medium. The higher the gain, the higher the amount of artefacts would be in the US images. Similarly, when gain is set to be high, the main beam energy increases as well and the reflected echoes from the medium would have higher energies. Therefore, almost all the reflected echoes receive to the US transducer and there will be a wider bright artefact (comparing to lower gain setting) appearing on the US slices Side Lobe Extraction Since the amount of side lobe energies are much less than the main lobe energy, there is a sudden drop in the pixel intensities between every parallel side lobe artefacts. Hence, this indicates the presence of multiple oriented side lobe energies around the main lobe energy, as shown in Fig

76 4.4. DISCUSSION 62 Figure 4.14: (a) Wire reflection corresponding to the lateral position labeled b on the grid template. Gain=0%, dynamic range=50 db, and power=-4. (b) Intensity profile of the wire reflection with out side lobe energies. (c) Wire reflection corresponding to the lateral position labeled b on the grid template. Gain=100%, dynamic range=50 db, and power=-4. (d) Intensity profile of the wire reflection with main lobe energy (highest intensity peak) and side lobe energies (low intensity peaks). 4.4 Discussion The fact that the US focal zone is located close to the transducer essentially suggests that the TRUS device would provide the best elevational resolution within the focused beam region. In addition, the measured beam profiles and the needle offsets represent error maps in the TRUS images for different lateral distances of the beam. The calculated error maps could be used to improve brachytherapy applications by considering the likelihood of position errors on the TRUS data which are otherwise ignored or treated uniformly in the current practice. Electronically set focal zones or focal lengths did not have any influence on the beam profile pattern and the thickness of the beam artefacts on the TRUS images.