The Potential for Ultrasonic Image-Guided Therapy Using a Diagnostic System

Home , Nonlinear acoustics

by Kristin Frinkley Bing

Department of Biomedical Engineering Duke University

Date:

Approved:

Kathryn R. Nightingale, Ph.D., Supervisor

Gregg E. Trahey, Ph.D.

Stephen W. Smith, Ph.D.

Pei Zhong, Ph.D.

Cynthia D. Guy, M.D.

Dissertation submitted in partial fulﬁllment of the requirements for the degree of Doctor of Philosophy in the Department of Biomedical Engineering in the Graduate School of Duke University

2008 abstract

(Biomedical Engineering)

The Potential for Ultrasonic Image-Guided Therapy Using a Diagnostic System

Kristin Frinkley Bing

Department of Biomedical Engineering Duke University

Date:

Approved:

Kathryn R. Nightingale, Ph.D., Supervisor

Gregg E. Trahey, Ph.D.

Stephen W. Smith, Ph.D.

Pei Zhong, Ph.D.

Cynthia D. Guy, M.D.

An abstract of a dissertation submitted in partial fulﬁllment of the requirements for the degree of Doctor of Philosophy in the Department of Biomedical Engineering in the Graduate School of Duke University

2008 Copyright c 2008 by Kristin Frinkley Bing

Ultrasound can be used for a variety of therapeutic purposes. High-intensity focused ultrasound (HIFU) has progressed over the past decade to become a viable therapeutic method and is valuable as a non-invasive alternative to many

surgical procedures [62]. Ultrasonic thermal therapies can also be used to release thermally sensitive liposomes encapsulating chemotherapeutic drugs. In the brain, the permeability of the blood-brain barrier to drugs, antibodies, and gene transfer can be increased with a mechanical mechanism using ultrasound and contrast

agent [66,115,116]. The work presented in this dissertation tests the hypothesis that a diagnostic system can be used for combined imaging and therapeutic applications. In order to evaluate the effectiveness of a diagnostic system for use in therapeutic applications, a set of non-destructive tests is developed that can predict the potential for high acoustic output. A rigorous, nondestructive testing regimen for standard, diagnostic transducers to evaluate their potential for therapeutic use is formulated. Based on this work, transducer heating is identified as the largest challenge. The design and evaluation of several custom diagnostic transducers with various modifications to reduce internal heating are described. These transducers are compared with diagnostic controls using image contrast, face heating, hydrophone, and ARFI displacement measurements. From these results, we conclude that the most promising design is a passively and actively cooled, PZT-4 multilayer composite transducer, while the acoustically lossless lens and capactive micro-machined transducers evaluated herein

iv are determined to be ineﬀective. Three therapeutic applications are evaluated for the combined system. Image- guided spot ablations, such as in the treatment of early stage liver cancers, could not be successfully performed; however, the additional acoustic output requirements are determined to be on the order of 2.4 times those that can be currently produced without transducer damage in a clinically relevant amount of time (10-20 seconds per spot). The potential of a diagnostic system for a hyperthermia application is shown by producing temperatures for the duration necessary to release chemotherapeutic agents from thermally-activated liposomes without damage to the transducer. Fi- nally, a mechanically-based therapeutic method for opening the BBB with ultrasonic contrast agent and specialized sonication regimes under ultrasonic B-mode guidance is demonstrated. These studies indicate that a diagnostic system is capable of both moderate thermal and mechanical therapeutic applications under co-registered image-guidance.

v Contents

Abstract iv

List of Tables xi

List of Figures xvi

Acknowledgments xxiv

1 Introduction 1 1.1 Background ...... 4 1.1.1 Transducer Heating Damage Mechanisms ...... 4 1.1.2 Dual-ModeArraysStartingfromHIFU ...... 5

1.1.3 Tissue Heating: Thermal Lesions and Drug Delivery ...... 7 1.1.4 ARFI Visualization of HIFU Lesions ...... 9 1.1.5 Blood-BrainBarrierDisruption ...... 10

2 Therapeutic Potential Metric 12 2.1 Introduction...... 12 2.2 Background ...... 13

2.2.1 Intensity Measurements ...... 13 2.2.2 PowerandThermalIndex ...... 15 2.2.3 Acoustic Radiation Force Measurements with ARFI Imaging . 16 2.2.4 ThermalMeasurements...... 17

vi 2.2.5 NonlinearLosses ...... 18 2.3 Methods...... 19 2.3.1 DiagnosticSystems ...... 19 2.3.2 AutomatedPeaking...... 19

2.3.3 Intensity Measurements ...... 22 2.3.4 ThermalMeasurements...... 23 2.3.5 ARFI Displacement Measurements ...... 24 2.3.6 FieldandFEMSimulations ...... 25 2.3.7 Impedance Measurements ...... 28

2.4 Results...... 29 2.4.1 Experimental TPM Components ...... 29 2.4.2 TPMSummary ...... 32 2.5 Discussion...... 37

2.6 Conclusions ...... 42 2.7 Acknowledgments...... 43

3 Custom Diagnostic Transducers 44 3.1 Introduction...... 44 3.2 Background ...... 46 3.2.1 PZTTransducerHeating...... 46

3.2.2 PassiveandActiveCooling ...... 48 3.2.3 cMUTs and Fresnel Focusing ...... 49 3.3 Methods...... 51 3.3.1 TransducerHeating...... 52

3.3.2 Intensity Measurements ...... 53

vii 3.3.3 AcousticRadiationForce...... 54 3.3.4 Diagnostic Image Quality Assessment ...... 55 3.3.5 GoldenSpike ...... 55 3.4 Results...... 56

3.4.1 PZT-4MultilayerComposite...... 56 3.4.2 LosslessLens ...... 60 3.4.3 PassiveCooling ...... 63 3.4.4 Capacitive Micromachined Ultrasound Transducer ...... 69 3.5 Discussion...... 74

3.6 Conclusions ...... 78 3.7 Acknowledgments...... 79

4 Liver Ablation 80 4.1 Introduction...... 80 4.2 Background ...... 82 4.2.1 TypicalHIFUExposures ...... 82

4.2.2 NonlinearAcoustics...... 82 4.3 Methods...... 85 4.3.1 ParametricAnalysis ...... 85 4.3.2 HIFUTesting ...... 87

4.3.3 Finite Element Thermal Modeling ...... 89 4.3.4 TransducerEvaluation ...... 91 4.4 Results...... 91 4.4.1 ParametricEvaluation ...... 91

4.4.2 HIFUStudies ...... 94

viii 4.4.3 TransducerDamage...... 98 4.5 Discussion...... 98 4.6 Conclusions ...... 104 4.7 Acknowledgments ...... 105

5 Liposomal Drug Delivery 106 5.1 Introduction...... 106

5.2 Background ...... 108 5.3 Methods...... 108 5.3.1 Intensity Measurements ...... 108 5.3.2 ExVivoHeating ...... 109 5.3.3 InVivoHeating...... 110

5.4 Results...... 111 5.4.1 AcousticOutput ...... 111 5.4.2 ExVivoHeating ...... 112 5.4.3 InVivoHeating...... 112

5.5 Discussion ...... 114 5.6 Conclusions ...... 116 5.7 Acknowledgments ...... 116

6 Blood-Brain Barrier Disruption 118 6.1 Introduction...... 118 6.2 Methods...... 121

6.2.1 AnimalSetup ...... 121 6.2.2 UltrasoundApplication...... 121 6.2.3 BBBOpeningProcedure ...... 122

ix 6.2.4 ImageAnalysis ...... 125 6.2.5 Histology ...... 126 6.3 Results...... 126 6.4 Discussion ...... 133

6.5 Conclusion...... 139 6.6 Acknowledgments ...... 139

7 Conclusions and Future Work 140 7.1 Conclusions ...... 140 7.2 FutureWork...... 141 7.2.1 TransducerDesign ...... 141 7.2.2 Hyperthermia Applications ...... 142

7.2.3 Blood-Brain Barrier Applications ...... 142

A Liposomal Release Assay 144 A.1 Methods...... 144 A.1.1 LiposomeEvaluation ...... 144 A.1.2 ExVivoReleaseAssay ...... 146 A.2 Results...... 147

A.2.1 LiposomeEvaluation ...... 147 A.2.2 ExVivoRelease...... 148 A.3 Discussion ...... 150 A.4 Acknowledgments ...... 153

Bibliography 154

Biography 167

x List of Tables

2.1 Physical properties of the diagnostic transducers evaluated in this study. The footprint reflects the full width and height of the array, but the same lateral aperture was used for comparing the transducers in the experiments presented herein...... 20 2.2 All data were acquired at the center frequency of each transducer. The system voltage at which each measurement is acquired is given. All ARFI data were acquired with an F/3 excitation configuration and an F/2 tracking configuration, while temperature measurements used an F/1.5 configuration. A ‘–’ indicates the data were not acquired. The attenuation for the displacement and thermal simulations and experiments was 0.5 dB/cm/MHz, and the phantom and simulation Young’smoduluswas4.5kPa...... 36 2.3 All data were acquired at 2.67 MHz. The system voltage at which each measurement is acquired is given. All ARFI data were acquired using an F/3 excitation configuration and an F/2 tracking configuration, while the temperature measurements and Field II (beamwidth and gradient) simulations used an F/1.5 configuration. A ‘–’ indicates the data was not taken. The attenuation for the displacement and thermal simulations and experiments was 0.5 dB/cm/MHz, and the phantom and simulation Young’s modulus was 4.5 kPa. The Field II simulations had 0.3-mm resolution in all three dimensions...... 38

3.1 Thermal diffusivities (κ), conductivities (K), and coefficients of expansion (CE) for common transducer materials and tissues. The first property affects the rate of heat diffusion (see equation 1.2 without perfusion). The latter two thermal properties relate to the likelihood of thermal shock. (M = matching layer, B = backing layer, L = lens) 48 3.2 Sequence parameters for the ARFI and thermal testing of the Gore PZT5-H single layer and PZT-4 composite triple layer transducers. . 57

xi 3.3 ARFI image lesion contrast and average displacement (1.3 ms after radiation force excitation) in homogeneous phantom material surrounding the lesion at depths from 5-6 cm for the Gore probes. Each radiation force excitation was a 2.5 MHz, unapodized, F/3.5, 160-µs pulse focused at 6 cm. The variations over 4 trials in contrast and displacement were negligible ( 0.02 for contrast and 0.07 µm for displacement). For comparable≤ image metrics, lower system voltages wereusedforthe3layerarray...... 60

3.4 Sequence parameters for the ARFI and thermal testing of the 8L5 (lossless lens) and VF10-5 (control) transducers...... 61 3.5 Comparison of acoustic output at the focus of the 8L5 (lossless lens) transducer with a control using maximum ARFI displacements. Each radiation force excitation was a 5.7 MHz, unapodized, F/1.5, 70-µs pulse focused at 1.8 cm. The standard deviations were over 4 ARFI acquisitions at different phantom locations...... 62 3.6 Face heating comparison between the 8L5 (lossless lens) transducer and VF10-5 (control) using M-mode sequences with 10-cycle, 5.7-MHz pulses transmitting at a PRF of 7.9 kHz for 1 or 5 seconds. The system voltages in each row are such that the focal acoustic output, according to ARFI displacement, was equal between the two arrays. Each temperature value is the average over 4 sequence firings (no repeaking). The precision of face heating measurements without repeaking is 8%. 62 ± 3.7 Sequence parameters for the ARFI and thermal testing of the Duke-2 (passive cooling) and VF7-3 (control) transducers...... 64 3.8 Maximum ARFI displacements associated with the Duke-2 transducer using 3.33 MHz and 4.21 MHz radiation force excitation frequencies at various F/#s. Each excitation pulse was 47.5µs in duration focused at 1 cm using 55% of the maximum system voltage. The precision of these displacements are based on the standard deviation across 40 linesintheARFIimage...... 65 3.9 The maximum displacements associated with the VF7-3 transducer at different system voltages were compared to the displacement of the Duke-2 at 55% using a 3.33 MHz, 47.5 µs radiation force excitation focused at 1 cm, F/1. Means and standard deviations are across one image...... 66

xii 3.10 Backing layer and face heating for various M-mode sequences on the Duke-2 with a PRF of 8 kHz transmitting 3.33 MHz pulses at 50% of the maximum system voltage. ‘–’ represents data that was not acquired. VF7-3 face heating measurements were made using 30.7% system voltage, which yielded equivalent displacement to 50% voltage on the Duke-2. The face heating measurements were not repeaked between trials. The face heating measurement precision without repeaking is 8%. *These temperatures were recorded after element damage wasnoted...... 67 3.11 M-mode sequences resulting in visible damage to the Duke-2 transducer in B-mode images. The beam location of sequence 2 was spatially oﬀset from that of sequence 1 in order to have a completely independent aperture (no previously damaged elements). *The elements in beam 30 were located further away from the type-K thermocouple in the backing layer than those in beam 110. These experiments were performed only once; the precision of the type-K thermocouple is ± 2%...... 67 3.12 Sequence parameters for the thermal testing of the cMUT transducer. PW Doppler used apodized pulses, while M-mode did not. The PW sequences were also evaluated on the PH4-1...... 70 3.13 cMUT driving voltage, surface pressure, total acoustic power, and face heating for a sequence transmitting 8000 M-mode lines consisting of 10-cycle pulses at 2 MHz with a PRF of 7.9 kHz. cMUT transmit pressure simulations were used to estimate the peak positive surface pressure for all but the second row of data. The repeatability of the pressure, power, and temperature measurements for this experimental design was 8%. *This sequence tried to draw more power than the scannercouldsupply...... 72

3.14 Face heating on cMUT using PW Doppler sequences at 2 MHz, 3.75 cm, +/- 95V for 60 seconds. Voltages yielding equivalent surface pressures were used for measurements on the PH4-1 (100V = 100% system voltage). All sequences used 46 element and 25 cycles unless otherwise stated. A repeatability near 8% for the power and temperature measurementswasexpected...... 73

4.1 HIFU exposures used for liver ablation. Intensities are in water unless otherwise stated. (RFB = Reference speciﬁed that the measurement wasmadewitharadiationforcebalance.) ...... 83

xiii 4.2 Temperature rise, thermal dose, and damage associated with HIFU sequences. All ultrasound pulses are at 4.44 MHz. The intensity and power measurements are found by linear extrapolation from low system voltages. The intensity, power, and simulated temperature rises assumed no dead elements and were overestimates after damage occurred...... 97 4.3 M-mode sequences that are known to cause damage to standard, diagnostic transducers (see Chapter 3 for custom transducer damage) and should not be used without cooling measures. Note: The * sequence can be used if the bottom cm of the transducer is completely surrounded by room temperature (or colder) water...... 98

5.1 Heating sequence used for ex vivo and in vivo studies...... 110 5.2 Intensities associated with 4.44 MHz, F/1.5, unapodized, 10 cycle pulses focused at 5 cm. Intensities in each column are using linear extrapolation in water, assuming a 4-cm water path (with nonlinear losses) and 1 cm of chicken breast muscle (α=0.5 dB/cm/MHz), and assuming a 47-mm water path (with nonlinear losses), mouse skull (0.13 Np/MHz over skull thickness [21]), and 3 mm of brain (α=0.6 dB/cm/MHz), respectively. The in situ intensities are only given for the heating scenarios evaluated herein. These measurements were not repeaked, but the standard repeatability is 8%...... 112 5.3 Temperature rises associated with M-mode sequences transmitting 4.44 MHz, F/1.5, unapodized, 10 cycle pulses focused at 5 cm with a pulse repetition frequency of 7.1 kHz. These values were measured at a depth of 3 mm in chicken breast muscle through a waterbag interface in the nearﬁeld. The thermocouple was positioned using the needle holder/alignment apparatus at a depth of 3 mm below the muscle surface. Temperature deviations due to system variation are negligible.113

6.1 This table summarizes the exposure parameters investigated in this study along with the number of insonifications evaluated for each set of parameters. Each location was insonified for 30 seconds with a PRF of 10 Hz and an unapodized, F/1.5 configuration except the PW Doppler sequence (*) which used an 100 Hz PRF and an apodized, F/4configuration...... 124

xiv A.1 Carboxyfluorescein fluorescence and release at different temperatures as measured by a spectrofluorophotometer. The liposomes were exposed to each temperature for 1 hour except the full release condition at 50◦C, which only required 20 minutes. The precision of the spectrofluorophotometer was 0.6. One sample was evaluated per ± temperature...... 148 A.2 The ratio of carboxyfluorescein to rhodamine (release to amount of liposomes present) is given for the controls and ultrasonically heated liposomes. The mean, mean of the minimums (per sample), mean of the maximums (per sample), absolute minimum, and absolute maximum were taken across all sections in all samples. The p-value from atwo-samplet-testisshown...... 149

xv List of Figures

1.1 B-Mode (left), ARFI displacement (center), and pathology (right) images of an ex vivo bovine liver sample after HIFU ablation. The HIFU system used a 1 MHz, piston transducer transmitting continuous wave ultrasound for 10 seconds in order to generate the ablation lesion. This resulted in bubble formation in the B-mode image (bright white regions), and slight enhancement in echo signal from the center of the ablated region. In the matched ARFI displacement image (the grayscale units are microns), a darkened region of smaller displacements (e.g. stiﬀer region) corresponds to the ablation region. Note that the distal boundary is obscured by artifacts from bubbles, which could be avoided if a PDL were created. Areas of decorrelation (<0.99)areshowninblue...... 10

2.1 Normalized temperature rise as estimated by the bio-heat transfer equation at 2, 3, and 4 cm deep in liver assuming a constant transmit power. These curves represent the trade-oﬀ between nearﬁeld loss and focal gains as a function of frequency for an assumed α of 0.5 dB/cm/MHz[35]...... 20 2.2 (a) Standard deviation (in elevation dimension) in intensity across transducer face for PH4-1. (b) Acoustic power (measured by integrating Isppa values generated with 10% system voltage across the transducer face) for two curvilinear arrays (CH4-1=3.08 MHz, CH6- 2=4.44 MHz), one 1D phased array (PH4-1=2.67 MHz), and one 2D phased array (2 MHz). The error bars represent the 8% repeatability forhydrophonemeasurements...... 30 2.3 Focal plane (all with a focal depth of 37.5 mm) intensity plots (in dB) for the a) CH4-1 (3.08 MHz), b) CH6-2 (4.44 MHz), c) PH4-1 (2.67 MHz), and d) 2D (2 MHz) transducers measured with a membrane hydrophone. The color scale for each plot is -6 to 0 dB...... 31

xvi 2.4 Acoustic spatial peak and spatial average intensities at the (a) surface and (b) focus (37.5 mm) for two curvilinear arrays (CH4-1=3.08 MHz, CH6-2=4.44 MHz), one 1D phased array (PH4-1=2.67 MHz), and one 2Dphasedarray(2MHz)...... 32 2.5 Experimentally measured and FEM simulated displacements. Max- imum raw displacements within 25% of the 3.75 cm focus from ± acoustic radiation force (F/3, 180 µs interrogation) on the 1D probes. Effective displacements after normalization for the loss due to attenuation and increased energy absorption with frequency (CH4-1=3.08 MHz, CH6-2=4.44 MHz, PH4-1=2.67 MHz) are also shown. (These normalized displacements are divided by the maximum measured or simulated displacement before normalization to show all curves on sameplot.) ...... 33 2.6 Temperature rises at the focus for 1000 line M-mode sequences with a PRF of 7.9 kHz using 10-cycle pulses with an unapodized F/1.5 configuration through a (a) waterpath and (b) phantom material are shown (except CH4-1, which was in the noise floor of the thermocouple). Effective temperature rises after normalization for loss due to attenuation and increased energy absorption with frequency (CH4-1=3.08 MHz, CH6-2=4.44 MHz, PH4-1=2.67 MHz) are also shown. (These normalized temperatures are divided by the maximum measured or simulated displacement before normalization to show all curves on same plot.) FEM simulated temperature rises are also shown for the phantompropagationpathcase(b)...... 34 2.7 Summary of the therapeutic potential metric measurements obtained for the 1D arrays at their center frequencies (CH4-1=3.08 MHz, CH6- 2=4.4 MHz, PH4-1=2.67 MHz) a) without and b) with normalization (the normalization factors for focal gain and nearfield loss were applied). Each measurement was normalized by the maximum value across all probes in both plots. Note that the trends for Isppa and Isapa overlap...... 35 2.8 Summary of the therapeutic potential metric measurements obtained forthe1Darraysat2.67MHz...... 37

xvii 2.9 Normalized maximum temperature rise for the CH4-1 and PH4-1 transducers as calculated from their lateral-elevation intensity profiles (a) - intensity profile as measured, b) - intensity profiles normalized by total power, c) - intensity profiles normalized by maximum intensity) andthebio-heattransferequation...... 39 2.10 The fraction of the source intensity attenuated for different propagation depths in water and different transmit frequencies as determined by equations 2.4 and 2.5 [36]. Left - p0 (0.42 MPa) was as measured for 10% system voltage using the CH6-2 at 4.44 MHz; Right - Same asleftbutat98%systemvoltage(4.72MPa)...... 40

3.1 Face heating from a 1 cm aperture using a PZT5-H, single layer control array and a PZT-4, triple layer, composite array was measured by a 33 gauge, type-T, hypodermic needle thermocouple coupled with a thin layer of water to thermal tissue mimicking phantom material. Example temperature versus time curves are shown. Note: Due to better impedance matching, comparable Isppas were achieved with a lower driving voltage for the PZT-4 multilayer. Transmit frequen- 2 cies of a) 2.5 MHz (focal Isppa=240 W/cm ) and b) 3.6 MHz (focal 2 Isppa=607 W/cm ) are shown. Error bars show the standard deviation over four re-peaked measurements (with statistical outliers removed). 57 3.2 B-mode images of a RMI phantom including cysts and point targets using a) PZT5-H, single layer, control array (contrast: hyperechoic = -0.10, hypoechoic = 0.95) and b) PZT-4, triple layer, composite array (contrast: hyperechoic = -0.04, hypoechoic = 0.94). Note the similar imagequality...... 58

3.3 ARFI images of a lesion in a CIRS elastography phantom mimicking liver tissue 1.3 ms after radiation force excitation using 57% (a), 67% (b & c), 77% (d & e), and 87% (f) of the maximum system voltage for the radiation force excitations. Top Row: PZT-5H, single layer, control array, Bottom Row: PZT-4, triple layer, composite array. . . 59 3.4 B-mode image quality comparison between the a) 8L5 and b) VF10- 5. The hypoechoic lesion contrast was 0.42 for the 8L5 versus 0.21 for the VF10-5, while the hyperechoic lesion contrasts were -0.27 and -0.21,respectively...... 63

xviii 3.5 ”Golden Spike” images taken with a 7 MHz transducer of the acoustic stack of the Duke-2. Top - Axial image. Bottom - C-scan image at around 0.6µs in the axial scan. The dark red regions in the C-scan correspond to areas of matching layer delamination...... 68 3.6 Image quality comparison between the passively cooled Duke-2 (a) transducer and the VF7-3 control (b) using a CIRS elastography phantom containing a stiﬀ lesion. The B-mode contrast was 0.35 for the Duke-2 and 0.58 for the VF7-3, while the ARFI image contrasts were 0.40 and 0.57, respectively. The ARFI images are shown 0.6 ms aftertheradiationforceexcitation...... 69 3.7 (a) Intensity proﬁle (color scale in dB) of cMUT measured 8 mm from transducer surface transmitting 114 elements at 2 MHz with +/- 95V bias, and 21V AC voltage. The resulting power (integrated Isppa across the transducer face) was 0.6 W with a peak intensity of 0.37 W/cm2. (b) Fresnel pattern associated with cMUT (gray scale in volts). ... 71 3.8 B-mode image quality comparison between the cMUT and PH4-1 transducers using a CIRS elastography phantom. A very low contrast, 12-mm diameter lesion is centered at 0 mm lateral, 27 mm in depth. The contrast for the cMUT was 0.12 while the contrast for the PH4-1was0.28...... 74

4.1 Picture of setup for HIFU testing including cooling of the transducer (ice water circulated around perimeter of lens through tubing and ice water bag around handle), heated water bath (body temperature) for liver, and thermocouple for monitoring temperature rise (wire seen at frontoflivercontainer)...... 88 4.2 Normalized temperature rise (normalization factor = 7.2◦C) versus total insonification time as measured by the thermocouple at the focal depth and the CH6-2 transducer face (2% duty cycle, 55% system voltage (3025 V2)). Note that the temperatures increase almost linearly for the first 0.3 s at the focus and longer at the transducer face. However, the slopes are not equivalent due to the different thermal properties surrounding the transducer face (slope=6.98 ◦C/s) and focal point (slope=12.96 ◦C/s). Error bars represent the standard deviation of temperature readings over four trials with no repeaking. . . 92

xix 4.3 Normalized temperature (normalization factor = 2.6◦C) versus power measured by a thermocouple at the focus and at the CH6-2 transducer face (2% duty cycle, 0.12 s insonification time). The dashed line is extrapolated from 100 to 1600 V2 (10-40% system voltage) to show the expected trend for temperature with increasing power (slope=1.05e-3 ◦C/V2) in a linear medium compared to that at the face (slope=2.78e- 4 ◦C/V2) . The solid line indicates the expected trend from nonlinear saturation, as predicted by equation 4.5. Error bars represent the standard deviation of temperature readings over four trials with no repeaking...... 93 4.4 Normalized temperature rise (normalization factor = 11.3◦C) versus pulse repetition frequency (PRF) measured by a thermocouple at the CH6-2 transducer face (slope=0.42 ◦C/kHz) and focus (slope=0.84 ◦C/kHz) for a total duration of 0.36 s using 10-cycle pulses (55% system voltage (3025 V2)). Error bars represent the standard deviation of temperature readings over four trials with no repeaking...... 94 4.5 Normalized temperature rise (normalization factor = 1.1◦C simulation, 2.1◦C experiment) as the thermocouple is moved laterally away from the ultrasound beam focus as compared to FEM simulation for 5000 M-mode lines at 7.9 kHz PRF each with a 10 cycle, F/1.5, unapodized, 2 cm focus pulse transmitting at 4.44 MHz with 55% of the maximum system voltage in tissue mimicking material. The match with FEM breaks down at positions under the noise floor of the thermocouple. The vertical dashed lines indicate the -3 dB ultrasound beamwidth (F/# λ [64])...... 96 · 4.6 Normalized FEM temperature rises verus lateral position from the beam focus for M-mode sequences with 10 cycle, F/1.5, unapodized, 2 cm focus pulses transmitting at 4.44 MHz in liver. One sequence has a PRF of 14.9 kHz and an insonification time of 8 seconds (corresponds to high PRF sequence measured at 0.5 mm (circle)), while the other has a PRF of 7.9 kHz lasting for 9 seconds (corresponds to extended peaking and high power sequences measured at 1.2 mm (square)). The vertical dashed line indicates the -3 dB ultrasound beamwidth (F/# λ [64])...... 96 ·

xx 5.1 Moles of Doxorubicin released versus time in minutes from DPPC:MSPC(10%):DSPE- PEG(2000) (4%) vesicles at 30, 37, 39, 40, 41.3, 42 and 45◦C. The release of 3.86e−9 moles of drug corresponds to 100% release of contents. All data points represent the mean of three separate experiments. Open circles represent release at temperatures above the transition temperature (41.3◦C). This graph was reproduced from [86]...... 109 5.2 Temperature rise versus time for the M-mode sequence transmitting 4.44 MHz, F/1.5, unapodized, 10 cycle pulses focused at 5 cm with a pulse repetition frequency of 7.1 kHz at 35% system voltage for 90 seconds through 3 mm and 10 mm of chicken breast muscle. Ther- mocouple measurements have an error 2%...... 113 ≤ 5.3 a) X-ray of mouse 1 showing the placement of the thermocouple in the brain. b) Cooling proﬁle in mouse 1 after reaching a peak temperature of 42.7◦C. Baseline temeprature was 37.4◦C...... 114

5.4 a) X-ray of mouse 2 showing the placement of the thermocouple in the brain. b) Lateral proﬁle of the heating in the cerebellum of mouse 2. The line indicates the order in which the data points were taken starting at the circle and ending at the square. Baseline temperature was 37.8◦C...... 115

6.1 Example waveforms (a,c) and power spectra (b,d) of pulses with peak- to-peak pressures of 2.72 MPa (a,b) and 6.16 MPa (c,d). At these pressures, the waveforms demonstrate some nonlinearity. The corresponding MI (P−.3/√f) are 0.33 and 0.65, respectively, assuming propagationthrough2cmoftissue...... 123 6.2 (a) Anatomical sketch of a coronal slice of the brain with the insonification spots. Only the two most rostral spot positions were analyzed in the MR images. (b) Setup and transducer orientation relative to the mouse. Note: The water bag is not shown here...... 125 6.3 BBB opening with PW Doppler. 5.7-MHz, 7-µs ultrasound pulses repeated at 100 Hz with an apodized F/4 configuration yielding 2.72 MPapp were transmitted for 30 seconds immediately after a 30-µL Definity injection. The white arrow points to an opened spot in the brain. The black arrow points to the ventricle...... 127

xxi 6.4 Images showing a) B-mode ultrasound only (5.7 MHz), b) MR only, and c) structures seen in ultrasound (found by thresholding) overlaid in red on the MR image. The yellow + shows the intended center of the ultrasound focus based on the B-mode image. The white region surrounding the + on the right side of the MR image is indicative of T1 enhancement from Magnevist crossing the BBB. BBB opening was achieved using 5.7-MHz, 20-ms ultrasound pulses repeated at 10 Hz with an F/1.5 configuration, yielding pressures of 6.16 MPapp,ina 30-second insonification immediately after a 30-µL Definity injection. 128 6.5 BBB opening as a function of Definity dose. 5.7-MHz, 20-ms ultrasound pulses repeated at 10 Hz with an F/1.5 configuration, yielding pressures of 6.16 MPapp, were transmitted for 30 seconds immediately after Definity injection. Each * represents one animal and the dashed lineconnectsthemeanateachdose...... 128

6.6 Effect of delay between Definity injection and start of insonification on BBB opening from 0 to 120 s. 5.7-MHz, 20-ms ultrasound pulses repeated at 10 Hz with an F/1.5 configuration, yielding pressures of 6.16 MPapp, were transmitted for 30 seconds after a 30-µL Definity injection. Each * represents one animal and the dashed line connects the mean at each delay. CNR differences were insignificant (p>0.05) forthedelaysstudied...... 129 6.7 BBB opening for ultrasonic transmission frequencies from 5 to 8 MHz for the same system input voltage. 20-ms ultrasound pulses repeated at 10 Hz with an F/1.5 focal configuration were transmitted for 30 seconds immediately after a 30-µL Definity injection. At least two animals were tested per frequency. Non-derated and derated pres-

sures as well as MI (P−.3/√f) and MIin situ (P−in situ /√f, [79]) for each frequency are listed. The standard deviation of these pressure measurements are 1%...... 130 ≤

6.8 Effect of ultrasonic pressures from 1.05 to 6.16 MPapp (non-derated) on BBB opening. 5.7-MHz, 20-ms ultrasound pulses repeated at 10 Hz with an F/1.5 configuration were transmitted for 30 seconds immediately after a 30-µL Definity injection. Each * represents one animal and the dashed line connects the mean at each pressure...... 131

xxii 6.9 a) Effect of pulse durations of 0.35 µs (B-mode), 2 µs (Color Doppler), 70 µs (Acoustic Radiation Force Impulse Imaging), and 20 ms on BBB opening. 5.7-MHz ultrasound pulses repeated at 10 Hz with an F/1.5 configuration yielding 2.72 MPapp were transmitted for 30 seconds immediately after a 30-µL Definity injection. Each * represents one animal and the dashed line connects the mean at each pulse duration. b) The same data as in a) presented as a function of the total number of cycles in the insonification sequence. Note these are semi-log plots inx...... 132 6.10 H & E stained histology of a) blood cell extravasation caused by standard B-mode (MI=1.5, 0.35 µs, 5.7 MHz, 34.60 MPapp insonifying for five 30 second periods at a 36 Hz frame rate with 30-µL Definity) and b) no damage with the most aggressive experimental ultrasound exposure used for this study (MI=0.65, 5.7-MHz transmit frequency, 6.17-MPapp pressure (in water), F/1.5, and 20-ms pulse duration with 30-µLDefinity.)...... 132 6.11 Example of image guidance and system settings for PW Doppler mode BBBopening...... 133

7.1 Images of a canine brain using an Acunav transducer for a) B-mode imaging of gyri and sulci and b) Color doppler imaging of the internal carotidartery...... 143

A.1 B-mode before injection of liposomes (a) and subtraction image showing lipsome injectate (b) in chicken breast muscle. Dashed yellow line shows the M-mode line used for ultrasonic heating...... 149 A.2 Example images at 1.6x magnification (height=4.1 mm, width=5.4 mm) of liposomal release from control (a & c) and ultrasonically heated (b & d) conditions. Green shows carboxyfluorescein release, and red shows presence of lipid shell of liposomes. The ratio of carboxyfluorescein to rhodamine in these images were a) 0.44, b) 0.55, c) 2.34, and d) 3.28, and they represent the minimum (a & b) and maximum (c & d) ratios seen over all trials for the control and heated groups...... 150

xxiii Acknowledgments

I would like to thank Dr. Kathy Nightingale for providing a nurturing laboratory environment and lively discussions. I would also like to thank Dr. Gregg Trahey for his guidance along the way. I thank the other members of my dissertation committee for their valuable insights and resources. A dissertation is a collaborative eﬀort. I would like to thank my fellow lab members and those in the Trahey lab for always allowing an open exchange of ideas. In particular, Dr. Mark Palmeri was a great sounding board and ﬁnite element modeler extraordinaire. My experiments were also aided by some excellent Pratt fellows, Katherine Oldenburg and Stephen Rosenzweig. I thank my family for their support even when they had no idea what I was talking about. I would like to thank my friend, Laura Dyer, for keeping me sane, especially during my preliminary examination. Most importantly, I cannot thank my husband enough for all of his support, encouragement, and trips down from Maryland to visit. Thanks, Tom.

xxiv Chapter 1

Introduction

High-intensity focused ultrasound (HIFU) has progressed over the past decade to become a viable therapeutic method and is valuable as a non-invasive alternative to many surgical procedures [62]. Most current HIFU implementations require a separate imaging method or diagnostic ultrasound probe for localization of the region of interest and monitoring of the ablation, in addition to the therapy probe [14, 124]. Exact registration between images of the region requiring therapy and the administration of HIFU is of vast importance in order to avoid or minimize skin burns, nerve fiber damage, local pain, and other complications [129]. Wu et al. used concentric imaging and therapy transducers for image guidance and ablation of hepatocellular carcinomas, showing the effectiveness, safety, and feasibility of HIFU for this application in the clinic [129]. However, this approach does not offer the

flexibility of beamforming that a linear or phased array does. We hypothesized that a single, modified diagnostic transducer and imaging system may be suitable for some smaller therapy applications making co-registered conventional B-mode ultrasound and acoustic radiation force impulse (ARFI) visualization of the treatment area straightforward. ARFI imaging monitors the dynamic response of tissue to focused, impulse acoustic radiation force excitations and, thus, visualizes structures with varied stiffness, which are often not seen in conventional B-mode ultrasound, such as protein-denatured lesions created by HIFU [38,39,73].

1 Ultrasonic thermal therapies can also be used to release thermally sensitive liposomes encapsulating chemotherapeutic drugs. Chemotherapy is mainly used as an adjuvant to surgery and radiation for cancer instead of a primary treatment because of the inability of drugs to reach the DNA and RNA of the cancer and associated cells [88]. Therefore, a drug carrier, such as a liposome, which can load and retain the drug, evade the body’s defenses, target (passively and specifically) the interstitial tissue of tumors, and release the drug only where desired would be ideal to make chemotherapy more effective and less toxic [88]. A study was performed at the Na- tional Institutes of Health (NIH) to demonstrate that pulsed high-intensity focused ultrasound could effectively activate low temperature-sensitive liposomes (LTSL) to deliver doxorubicin (a chemotherapeutic agent) in tumors [34]. The acoustic output required for these studies was moderate compared to that for HIFU. Another therapeutic application of ultrasound involves opening the blood-brain barrier (BBB). In the brain, the BBB, a diffusion barrier consisting of endothelial cells, astrocyte end-feet, and pericytes, serves as another roadblock to the effective delivery of chemotherapeutic and other small molecule agents. Several methods, some involving ultrasound, have been investigated to increase the permeability of the BBB to drugs, antibodies, and gene transfer [66,115,116]. Therapeutic applications performed on a diagnostic system, the focus of this thesis, can be guided by standard ultrasound imaging (B-mode or Doppler) or with elastographic techniques, such as acoustic radiation force impulse (ARFI) imaging. The research herein tests the hypothesis that a diagnostic system can be used for combined imaging and therapeutic applications by:

1. Deﬁning the challenges associated with performing therapeutic applications using diagnostic transducers and systems,

2 2. Determining therapies that are feasible on diagnostic systems,

3. Evaluating the eﬀectiveness of or additional requirements necessary for performing various therapies with diagnostic systems, and

4. Where applicable, demonstrating localization and monitoring of therapies with

a diagnostic system.

In order to evaluate the eﬀectiveness of a diagnostic system for use in therapeutic applications, a set of non-destructive tests was developed that could predict the potential for high acoustic output. Chapter 2 formulates a rigorous, nondestructive

testing regimen for standard, diagnostic transducers to evaluate their potential for therapeutic use. Based on this work, transducer heating was identiﬁed as the largest challenge. Chapter 3 thus describes the design and evaluation of several custom diagnostic transducers with various modiﬁcations to reduce internal heating. These transducers are compared with diagnostic controls using image contrast, face heating,

hydrophone, and ARFI displacement measurements. There are several therapeutic applications that utilize ultrasound, including focused thermal ablation [73], targeted drug delivery and release [34], and, more recently, ultrasound combined with contrast agents has been reported to facilitate reversible opening of the blood-brain barrier [77]. Each therapy has diﬀerent acoustic output and imaging requirements. In this work, two thermal therapies and blood-brain barrier opening were evaluated. The potential for intra-operative, image- guided spot ablations for the treatment of early stage liver cancers using a diagnostic system was evaluated in Chapter 4. Chapter 5 demonstrates the potential of a diagnostic system for releasing agents from thermally activating low temperature sensitive liposomes. Chapter 6 investigates a mechanically-based therapeutic method for opening the BBB with ultrasonic contrast agent and specialized sonication regimes

3 that were localized using ultrasonic B-mode guidance. Chapter 7 summarizes the ﬁndings of this work and how they can be used to guide the development of diagnostic transducers for therapeutic use as well as deﬁning potential new applications for these and existing transducers.

1.1 Background

1.1.1 Transducer Heating Damage Mechanisms

There are many sources of heating in a diagnostic transducer that can lead to damage, which are discussed more thoroughly in Chapters 2 and 3. Transducer heating may also lead to a change in electrical impedance resulting in different electrical matching of the transducer with the transmitter and, consequently, a change in the output acoustic power [41]. The transducer can fracture if the applied power exceeds acceptable excitation levels [41]. The depolarizing field, intrinsic and extrinsic properties of the PZT (density, speed of sound, voltage constant, thermal conductivity, frequency constant, and loss factor), and maximum acceptable value of temperature differential across the PZT caused by self-heating from electro-mechanical conversion loss limit the maximum acoustic intensity able to be produced from a given transducer [51]. When designing a transducer to be used for both diagnostic and therapeutic purposes, there is typically a trade-off when choosing a PZT material. Hard PZT (e.g. PZT-4 and PZT-8, typically used for HIFU transducers) has lower mechanical and dielectric losses, which should be minimized for less internal heating; in contrast, soft PZT (e.g. PZT-5H typically used in diagnostic arrays) has higher dielectric and coupling constants, which should be maximized for better sensitivity and bandwidth [133]. When designing sequences for ultrasonic heating of tissue, it is also important to

4 consider the eﬀects of power delivery and timing with relation to internal transducer heating. The material properties of the acoustic stack should be considered for this evaluation. For instance, the resonant frequency and bandwidth, or coupling coef- ﬁcient, of a transducer are lowered with increasing temperature. The overall trans-

mitter impedance characteristics vary with operating frequency and temperature in a complex way that diﬀers between arrays. The tensile lap shear strengths of typical matching layer adhesives reach a maximum near room temperature and, therefore, have weaker bonds as the transducer heats. The dielectric constant increases grad- ually until about 120◦C and then dramatically until the Curie temperature. By

association, the dissipation factor follows the same general trend. At the same time, the strain constant also increases with temperature to yield a more eﬃcient transmitter. As a result, the acoustic output increases leading to more internal heating in the transducer due to absorption of acoustic energy, which in turn leads to a higher

loss tangent and more power dissipation in the form of heat. The resistivity of PZT- 5H generally decreases to a minimum resistance at the Curie temperature. As the resistance decreases, the power delivered to the PZT for a given voltage increases, which further exacerbates the internal heating [52]. A judiciously chosen transmit

sequence must, therefore, be designed to avoid transient temperature spikes and thermal runaway.

1.1.2 Dual-Mode Arrays Starting from HIFU

HIFU transducers are generally built with low loss materials capable of sustain- ing high transmit pressures. The traditional HIFU transducer is a large, concave, piston transducer; however, annular phased arrays capable of electronic focusing in depth have been made in recent years [112]. Large apertures ( 35x22mm) are ≥ used to generate the acoustic output and focal gains required for HIFU [112]. Air

5 backing is typically used to increase the acoustic energy transmitted out of the front of the array [112]; however, this lowers the bandwidth and, therefore, limits any imaging capabilities the transducer may have. Some piezo-composite materials have shown low electrical and acoustical cross-coupling, but low material eﬃcien-

cies require high driving powers for HIFU, which result in high face temperatures during operation [112]. PZT-8 and other hard ceramics (e.g. PZT-4, which was evaluated herein) can be used for HIFU, since they can handle high powers [106]. Thermally conductive matching layers can also be used on HIFU tranducers for maximum acoustic power transmission to the tissue [106]. On the other hand, diag-

nostic transducers are designed with superior imaging quality in mind. Inherent in this design is a broad bandwidth, which generally comes at the expense of a higher electro-mechanical loss except in certain cases, such as cMUTs [126]. Some compromises can be made between HIFU and diagnostic transducers to cre-

ate dual-mode ultrasound arrays (DMUA). Some piezo-composite transducer technologies allow high-power array transducers to be made with increased fractional bandwidth and very low cross coupling between array elements [37]. For piezocomposite arrays with smaller footprints, surface cooling may be required to make

up for the higher drive levels necessary to compensate for the loss in focal gain, especially with oﬀ-axis steering [112]. However, for Ebbini et al., optimization of array eﬃciency in the therapeutic mode resulted in coarsely sampled elements (2λ spacing) in order to maintain a large transducer surface area without increasing the system channel count [37]. Such coarse element spacing can generate strong grating lobes.

Furthermore, if smaller elements were used in a larger aperture array, they may be less eﬃcient in delivering acoustic power and harder to drive because of increased impedance [37]. Ebbini’s transducer was driven by a high-Q (quality factor) resonant matching circuit while in therapeutic mode and by low-Q series inductors while in

6 receive imaging mode to shorten the echo duration [37]. To improve the poor, traditional B-mode image quality, synthetic aperture imaging and single-transmit focus imaging have been used to better visualize HIFU-created lesions with DMUAs [37]. In another DMUA, piezocomposite technology was used in a rotating single element,

catheter probe with cylindrical focalization, air backing to prevent energy loss and maximize emission, and matching layers to improve transmission into the propagation medium [13]. In this case, a matching circuit was only used in the imaging mode to enhance imaging resolution and sensitivity [13]. Another form of DMUA consists of a dual mode array surrounded by elements for therapy only, which bene-

ﬁts from large focusing intensity gain in both transmission and reception [12,45,50]. Finally, several catheter DMUAs have been manufactured using non-composite materials (e.g. PZT 880 (APC International, Mackeyville, PA) and a PZT-4 equivalent ceramic) where all elements are used for imaging and therapy [75,125]. One of these

arrays operated at 3.1 MHz with air-backing and a composite/epoxy matching layer while being cooled by forced-ﬂow chilled water [75]. Most of the arrays mentioned above were capable of generating ablation lesions, however, their image quality was considerably inferior to that of comparable frequency diagnostic arrays.

1.1.3 Tissue Heating: Thermal Lesions and Drug Delivery

The thermal dose, or time required to ablate a speciﬁc type of tissue at a certain temperature, can be calculated using an empirically determined equation [29]:

T1−T2 t2 = t1/R (1.1) where t is time, T is temperature, R is 0.5 for T greater than 43◦C, the subscripts 1 and 2 represent two time-temperature combinations that result in lesion formation

7 ◦ (T1 is usually 43 C and t1 is the time at that temperature which depends upon tissue thermal properties and is empirically determined), and all parameters on the right hand side of the equation are known. Damianou et al. report the thermal dose of

◦ ◦ liver to be 45 to 60 minutes at 43 C [29] (i.e. t1=60 minutes, T1=43 C, R=0.5).

The potential for performing thermal ablation with a diagnostic system is explored in Chapter 4. Drug delivery is another application of ultrasonic heating requiring a much lower thermal dose than ablation. Ponce et al. reported temperature-sensitive liposomes containing doxorubicin (DOX), a chemotherapy drug, which release 100% of their payload locally within 10-20s at 41◦C [96]. A group at the National Institutes of Health (NIH) were able to activate these liposomes using a HIFU piston transducer at 1 MHz (42◦ for 2 min) to release DOX more rapidly and at a higher concentration within murine adenocarcinoma tumors [34]. The potential for thermal activation of liposomes for image-guided drug delivery with a diagnostic system is explored in Chapter 5. To determine the expected temperature in a given tissue, analytic solutions can be employed. Pennes’ bio-heat transfer equation is utilized in calculations of heat transport and temperature rise in perfused media [93]:

T˙ = κ 2T T/τ + q /c (1.2) ∇ − v v where the time derivative of temperature (T˙ ) is equal to the summation of a conductive term (κ is thermal diffusivity), a perfusion term (τ is the perfusion time constant), and a heat source term (qv is the rate of heat production per unit volume and cv is the volumetric specific heat). If there is no perfusion, τ is considered infinite and the second term disappears. Without the perfusion term, this equation

8 can also be used to predict transducer heating. For a step-function point source, Nyborg [93] presented a solution for the unperfused case in the following form:

T =2Cr−1erfc(R) (1.3)

C = qvdv/8πK (1.4)

R = r/(4κt)1/2 (1.5)

where dv is the volume of the small source, K is the thermal conductivity (K = κcv), and r is the distance from the source. If the bio-heat transfer equation is speciﬁcally applied to a case where the heat is generated from an ultrasonic source, then qv is [93]:

qv =2αIta (1.6)

where Ita is the derated temporal average intensity at the point of interest. This equation shows how heat is produced by the absorption of ultrasonic energy in tissue and is used throughout the work presented herein.

1.1.4 ARFI Visualization of HIFU Lesions

Two-dimensional acoustic radiation force impulse (ARFI) imaging, which is implemented with a high quality diagnostic imaging system, has been evaluated for

HIFU lesion visualization. Although gas bubbles created during HIFU are easily identiﬁable in B-mode, they do not portray the full extent of the treated area nor do they detect all lesions. Figure 1.1 portrays matched B-mode, ARFI, and pathology images of a HIFU-generated ablation lesion with some bubble artifact. For details on the HIFU heating and ARFI imaging parameters used, the reader is referred to [89].

9 Figure 1.1: B-Mode (left), ARFI displacement (center), and pathology (right) images of an ex vivo bovine liver sample after HIFU ablation. The HIFU system used a 1 MHz, piston transducer transmitting continuous wave ultrasound for 10 seconds in order to generate the ablation lesion. This resulted in bubble formation in the B-mode image (bright white regions), and slight enhancement in echo signal from the center of the ablated region. In the matched ARFI displacement image (the grayscale units are microns), a darkened region of smaller displacements (e.g. stiﬀer region) corresponds to the ablation region. Note that the distal boundary is obscured by artifacts from bubbles, which could be avoided if a PDL were created. Areas of decorrelation (<0.99) are shown in blue.

Note the enhanced lateral boundary deﬁnition of the ARFI image over B-mode. Protein-denaturing lesions (PDL) can be created with lower HIFU exposures not exceeding levels that produce extensive gaseous bodies through vaporization, tissue

degassing, or cavitation [73]. Because these lesions require less power from the transducer and, hence, less transducer heating, they are optimal for the tumor ablation proposed here. They are not necessarily visualized with conventional B-mode ultrasound but have been well visualized (when created by RF ablation) by ARFI imaging

without artifacts [38, 39]. One-dimensional ARFI imaging and localized harmonic motion imaging (LHMI) have also been investigated to visualize the axial extent of these lesions through time [49,73].

1.1.5 Blood-Brain Barrier Disruption

Mechanical stimulation with ultrasound can also be used for therapeutic pur-

poses. Blood-brain barrier (BBB) disruption can be achieved using low-pressure

10 ultrasound and ultrasonic contrast agent in order to increase the permeability of the BBB to drugs, antibodies, and gene transfer [66, 115, 116]. Studies where opening occurred with a temperature rise of only 0.025◦C suggest that thermal eﬀects are not a factor in this method of BBB opening [65]. Furthermore, this BBB disruption has occurred without the detection of wide-band emissions, the signature for inertial cavitation; this type of BBB disruption also showed no blood cell extravasation [65,77]. Consequently, the factors likely to be responsible for such BBB disruption are the oscillation of microbubbles as occurs with stable cavitation, acoustic streaming around the microbubbles, and radiation force on the microbubbles [31,65,77,100,115]. These phenomena may cause mechanical stretching of vessels that leads to opening of tight junctions or the triggering of biochemical reactions to open the BBB [115]. This stretch theory is further supported by work showing that microbubbles can induce a mechanical stretch, activating BKCa channels and leading to rapid hyperpolarization of the cell membrane potential that is in direct contact with the bubbles [119]. All of these studies have been performed with therapeutic transducers and driving electronics at lower (less than 2 MHz) frequencies. We demonstrate the feasibility of using a diagnostic system for combined image guidance and BBB disruption in mice in Chapter 6. Mechanical stimulation from ultrasound can also be used for other therapeutic applications, such as histotripsy (mechanical fractionation of tissue structure) and thrombolysis (dissolving of blood clots); these therapies were not explored with a

2 diagnostic system herein because of the high acoustic output (Isppa > 10 kW/cm ) and very low frequencies (near 120 kHz) required, respectively [31,131].

11 Chapter 2

Therapeutic Potential Metric

Part of this work was published in the Proceedings of the 2007 IEEE Ultrasonics Symposium.

2.1 Introduction

High-intensity focused ultrasound (HIFU) has progressed over the past decade to become a viable therapeutic method and is valuable as a non-invasive alternative to many surgical procedures. Current HIFU implementations require a separate imaging method or diagnostic ultrasound probe due to the lower frequency and narrowband nature of HIFU transducers [14,124]. Some groups have developed non- standard, combined therapy and imaging transducers in the same probe [25,73]. One such prototype device (a phased array, HIFU transducer) has undergone initial testing using the same elements for therapy and imaging. While it successfully ablated tissue, lower scattering, complex structures (e.g. small ablation lesions) were not well visualized [37]. Alternatively, we hypothesized that a single, modiﬁed diagnostic transducer and imaging system may be suitable for some smaller therapy applications making co-registered conventional B-mode ultrasound and acoustic radiation force impulse (ARFI) [38] visualization of the treatment area straightforward.

However, because diagnostic probes are prone to thermal damage at high acoustic

12 output, one cannot drive all candidate tranducers aggressively to compare them. A method of non-destructively evaluating the potential for high acoustic output from diagnostic probes is, therefore, desirable. The goal of the work in this chapter is to develop a non-destructive ‘therapeutic potential’ metric and experimental pro-

tocol to define the expected efficacy of diagnostic transducers for thermal therapy. This metric should be easy to measure, non-destructive, and would ideally include normalization methods to allow comparisons of different transducer configurations. As such, several existing and some novel methods for quantifying acoustic output were performed on several candidate transducers, and their efficacy for predicting

therapeutic potential was compared.

2.2 Background

In this work, several metrics for quantifying acoustic output are evaluated: intensity, acoustic power, thermal index, acoustic radiation force impulse (ARFI) imaging, and focal heating.

2.2.1 Intensity Measurements

The American Institute of Ultrasound in Medicine (AIUM) and the National Electrical Manufacturers Association (NEMA) recommend making intensity measurements with a membrane hydrophone for diagnostic ultrasound [2, 3]. Known challenges in making hydrophone measurements include: 1) ensuring the beam axis is normal to the hydrophone and 2) aligning the active element of the hydrophone with the beam [2]. The peak rarefractional pressure (pr), spatial-peak pulse-average intensity (Isppa), spatial-peak temporal-average intensity (Ispta), and pulse intensity integral (PII, Ita*pulse repetition period for a non-scanning mode) are all hydrophone

13 measurements commonly reported for clinical ultrasound. However, membrane hydrophones cannot be used to measure the acoustic ﬁeld of high-intensity focused ultrasound (HIFU), because a risk exists for damaging the membrane with cavitation at high pressures for long durations. Therefore, several non-standardized methods have been used to report HIFU intensities. Spatial-average linear intensity

(ISAL) is found by assuming a Gaussian beam proﬁle and measuring the beam pro- ﬁle at low pressures with a hydrophone and the power at high pressures (includes nonlinearities) using a radiation force balance [114], resulting in the equation [70]:

I 0.87W I = spta = (2.1) SAL 1.8 D2 where D is the -6-dB half-pressure maximum beam width and W is the total acoustic power. The constants are derived from the ﬁxed relationship between the total power in the beam and the intensity averaged over the -6 dB beam area assuming a Gaussian beam [114]. Because radiation force balances require a certain amount of time ( 4 ∼ s) per measurement cycle, this approach cannot be used with diagnostic transducers at high output, because the high pulse average intensities would likely damage the transducer over the 4 second period required. Also, due to risk of damaging the hydrophone at high pressures with cavitation, it is best to avoid making direct measurements of high pressures. Therefore, for this work, we employ the approach of linearly extrapolating intensities measured with a hydrophone at low system voltages to higher system voltages in order to estimate the Isppa for short duration, high acoustic output ultrasound sequences [4,16].

14 2.2.2 Power and Thermal Index

Acoustic power (W) is the temporal average power emitted in the form of ultrasonic radiation by the transducer assembly [3]. Acoustic power can be measured with a radiation force balance or hydrophone. Known challenges in making radiation force balance measurements include: 1) angle of incidence between the beam and absorbing target must be no more than 10 degrees from normal, 2) ultrasound heats the absorbing target in the force balance causing the measured weight of the absorbing target to drift with exposure time as buoyancy changes, and 3) a short delay time is necessary before taking data after power-on/power-oﬀ due to overshoot in the transitory responses during this transition compared the steady state values [3]. Even with these challenges, NEMA reports that the accuracy and precision of hydrophone power measurements will generally not be as high as those from a radiation force balance [3]. Nevertheless, the suggested method of measuring acoustic power with a hydrophone is to scan the plane at the depth of the maximum derated (0.3 dB/cm/MHz) pulse intensity integral (PII) for all locations where the PII exceeds 0.25% of that maximum and integrate these values [2,3]. The power measurements performed herein were made near the surface of the array and integrated over the active aperture. The thermal energy produced in the patient’s body during an exposure of duration t is equal to Wt [1]. The thermal index is a quantity related to the potential for thermal heating and is proportional to the measured or estimated temperature rise for the exposure conditions. It is designed to take into account the total energy output, scanning mode, shape of the ultrasound beam, position of the focus, center frequency, shape of the waveform, and frame rate, but it estimates a steady state temperature rise, and thus it does not incorporate insoniﬁcation time [7]. For

15 apertures greater than 1 cm2, as presented in this dissertation, it is deﬁned as:

2 max (min([W 3(z), I 3(z)x1cm ]))f T I = z . ta. c (2.2) 210

where W.3(z) and Ita.3(z) are the derated power and temporal average intensity

(along beam axis, peak in lateral-elevation plane) at depth z and fc is the center frequency [4].

2.2.3 Acoustic Radiation Force Measurements with ARFI Imaging

Radiation force is the basis for acoustic radiation force impulse (ARFI) imaging and is linearly related to acoustic output. Instantaneous radiation force, as in the case of an ARFI imaging pulse, can be expressed in terms of pulse-average intensity as [92]: 2αI~ F~ = pa (2.3) c where α is the attenuation of the medium and c is the speed of sound in the medium. The direction of the radiation force is along the gradient in particle velocity over time, which is the direction of the Poynting vector. The application of this force to tissue results in micron-scale displacements which can be ultrasonically tracked. The tissue displacement generated by acoustic radiation force is nonuniform and is strongly dependent on tissue properties and ultrasonic beam characteristics. The displacement is a dynamic event, consisting of shear waves that propagate away from the region of excitation (ROE). Shearing within the tracked volume results in displacement underestimation due to the averaging of displacements within the tracking beam point spread function (PSF) and reduced echo signal correlation that

16 increases the displacement estimate variance [95]. Even with these challenges, displacement will be linearly related to applied force and, thus, intensity. Due to the ease of ARFI implementation, we investigated the potential for predicting acoustic output based on comparing ARFI displacement estimates.

2.2.4 Thermal Measurements

Tissue temperature rises are proportional to acoustic output. Thus, measuring temperature rises for controlled sequences could predict acoustic output. However, there are challenges in temperature measurements. The artifacts associated with thermocouple measurements of ultrasonically induced temperature rises are viscous heating, reflections, and conduction along the wire. Viscous heating is the action of viscous forces between the wires of the thermocouple and the imbedding medium as the acoustic wave propagates over the thermocouple [44]. Reflections from the front face of the thermocouple produce a region of local heating with a thickness approximately equal to the region heated by viscous effects. These artifacts result in a more rapid temperature rise in the first few seconds of heating and are difficult to remove from data. Thermocouple conduction introduces perturbations in the temperature distribution causing a temperature depression at the point of maximum absorbed power density, and the highest temperatures typically occur on either side of the thermocouple tip [23]. Viscous heating has been shown to be significantly reduced if membrane (thin-film) thermocouples (TFTs) are employed [113]. At the focus, some of the high output sequences were shown to etch the TFTs; thus, fine wire thermocouples were used for the focal heating measurements. Among fine wire thermocouples, those with small diameters and sheathed in stainless steel have been reported to measure lower temperatures for the same ultrasound sequences (indicat- ing less viscous heating effects) [54]. It should also be noted that more self-heating

17 artifacts have been reported as ultrasonic frequency increases [54], which adds another layer of complexity to these measurements and further motivates the need for additional metrics to deﬁne whether a probe is capable of therapeutic heating levels.

2.2.5 Nonlinear Losses

Correlating hydrophone measurements and thermal measurements made through

a waterpath with anticipated output in tissue is complicated by appreciable nonlinear

losses in water. The attenuation due to nonlinear losses (αw) can be defined for an infinite plane wave as [36] σ α = z−1 (2.4) w 1+ σ where σ is the shock parameter (σ 1 for weak shock absorption [28]) and z is ≥ the distance of propagation from the ultrasonic source through water. The shock parameter (σ) is defined as [36]

2π σ = 3 [p0fzβ] (2.5) ρ0c0 where ρ0 is the density of water, c0 is the speed of sound in water, p0 is the pressure at the ultrasonic source, f is the ultrasonic frequency, and β is the coeﬃcient of nonlinearity (equal to 3.5 for water). From equations 2.4 and 2.5, it becomes clear that increasing frequency, propagation distance, or source pressure can greatly increase the losses in water due to weak shock absorption.

18 2.3 Methods

2.3.1 Diagnostic Systems

The initial goal for a diagnostically based ablation system is intrasurgical liver tumor ablation. Due to the competing effects of near field loss and absorption at the focus, an optimum frequency exists for a given focal depth and tissue attenuation (α). Therefore, the attenuation and thermal properties of liver were used in conjunction with the bio-heat transfer equation (Eqn. 1.2) to determine the optimum frequency range given a focal depth between 2 and 4 cm (α=0.5 dB/cm/MHz). A frequency range of 2.2 to 4.3 MHz was found to be optimal for our intended application, as shown in Figure 2.1. As a result, three standard, 1D, diagnostic (PZT-5H with backing and matching layers) imaging transducers (CH4-1, CH6-2, PH4-1) were chosen for testing on a modified, Siemens SonolineTM Antares system with a power supply output up to 100V. Additionally, a fourth experimental 2D (PZT-5H) array developed by Siemens for diagnostic cardiac imaging was tested on a Siemens prototype system with a pulsed-wave power supply output from 20 to 200V. The physical parameters associated with these transducers are listed in Table 2.1. All of these transducers were tested at their center frequencies (CH4-1=3.08 MHz, CH6-2=4.44

MHz, PH4-1=2.67 MHz, 2D Array=2 MHz). After this testing, it was deemed valuable to test the 1D arrays all at 2.67 MHz.

2.3.2 Automated Peaking

In order to address the challenges in making repeatable hydrophone and thermocouple measurements at the focus, an automated peaking system was developed.

Custom transducer holders were designed and manufactured in the Duke BME Ma- chine Shop that coupled directly to a 3-D stepper motor controlled translation stage

Liver (B) − 2 cm 1 Liver (B) − 3 cm Liver (B) − 4 cm

0.8

0.6

0.4 Normalized Temperature 0.2

0 0 1 2 3 4 5 6 Frequency (MHz)

Figure 2.1: Normalized temperature rise as estimated by the bio-heat transfer equation at 2, 3, and 4 cm deep in liver assuming a constant transmit power. These curves represent the trade-oﬀ between nearﬁeld loss and focal gains as a function of frequency for an assumed α of 0.5 dB/cm/MHz [35].

Transducer Footprint Frequency Elevation Pitch (mm) (MHz) Focus (mm) (mm) CH4-1 61.1x14 1.8-4 48.8 0.477 CH6-2 60.9x12 2.5-6.7 69 0.314 PH4-1 28.2x13.5 1.8-4 70 0.22 2D 19.2x14.4 2-2.5 all 0.4

Table 2.1: Physical properties of the diagnostic transducers evaluated in this study. The footprint reﬂects the full width and height of the array, but the same lateral aperture was used for comparing the transducers in the experiments presented herein.

(Newport, Irvine, CA). Membrane hydrophone holders were also designed such that the face of a given transducer in its holder and the membrane were in parallel. Cus- tom LabView (National Instruments, Austin, TX) programs were written to control the translation stage and the ultrasound scanner for automated peaking of the measurement device (hydrophone or thermocouple).

The peaking program consisted of three parts: a raster scan to ﬁnd a signal

20 above a user-deﬁned noise level, a dynamic peaking algorithm in the lateral-elevation plane, and a dynamic peaking algorithm in the axial dimension. The raster scan swept across a user-deﬁned 2D grid until an rms pressure or temperature rise above a set threshold was found. At this point, the dynamic peaking algorithm began

with a coarse translational step-size, where the transducer stepped in elevation as the measurement value increased and switched directions when it decreased until a maximum was found. This process was then repeated with a ﬁner step-size, usually 0.1 mm. The same dynamic peaking procedure was then done in the lateral dimension. Finally, dynamic peaking was performed in the axial dimension but only

within a user-defined range to avoid touching or compressing the transducer against the hydrophone or thermocouple. During peaking, the ultrasound scanner was also controlled by the LabView program. For the hydrophone measurements, the user input the frequency, number of cycles, focal depth, F/#, apodization, and system voltage for each pulse. LabView then called a command prompt program interfacing with the ultrasound scanner to initiate a M-mode sequence consisting of the user-defined pulses repeated at 76 Hz. When peaking was finished, the program automatically disabled imaging to preserve the life of the transducer and hydrophone. For the thermal measurements, a much higher system voltage must be used for each pulse; therefore, a continously running M-mode sequence cannot be used without risking damage to the transducer. Thus, the command prompt program was called to run a short, M-mode sequence previously made by the user offline. A delay of approximately 1 minute between each

M-mode sequence was enforced by the LabView program to allow for the temperature to return to baseline before the next acquisition. The automated peaking method resulted in repeatable measurements after repeaking. For the hydrophones, intensity measurements were repeatable within 8%. ±

21 For the thermocouples, temperature measurements were repeatable within 10%. ±

2.3.3 Intensity Measurements

Intensity measurements are a direct measure of acoustic output from a transducer. Surface measurements give the output before focusing, while focal measurements incorporate the transducer’s beamforming capabilities at the depth of interest.

These measurements are generally made with a hydrophone, which precludes direct measurement of high-intensity pulses over long durations. Intensity is proportional to the rate of heat production and can be used to quan- tify acoustic power in the determination of the thermal index; therefore, it was

measured as an indicator of potential heating in the TPM. Intensity measurements were made using a 0.2-mm spot size membrane hydrophone with a preampliﬁer (Onda Corporation, Sunnyvale, CA) for the TPM at transducer center frequencies and a 0.6-mm spot size membrane hydrophone (no preamp) (Sonic Technologies, Wyndmoor, PA) for the TPM for all transducers transmitting at 2.67 MHz. A 3-D

stepper motor controlled translation stage (Newport, Irvine, CA), operated using LabView (National Instruments, Austin, TX), was used to move the transducer in three dimensions. Pressure values for 6- to 10-cycle pulses were measured at low system voltages in 0.1-mm increments in two planes: approximately 3 mm from the

surface of the transducer and at the focal distance (as determined by ﬁnding the

depth of maximum root-mean-square pressure in the axial direction). The Isppa values were calculated from each measured pressure waveform according to the following equation [3]: T0 T0 1 1 1 2 Isppa = pvdt = p dt (2.6) T0 Z0 ρc T0 Z0 where T0 is the pulse length, p is pressure, v is particle velocity, ρ is tissue density,

22 and c is the speed of sound in water. The total acoustic power output was calculated at the surface of each transducer by integrating the pulse-average intensities across the active aperture. Because acoustic power is conserved in a lossless medium, this quantity was derated and used in the thermal index. The Isapa values, the spatial- average, pulse-average intensities, were calculated as the total acoustic power at the surface divided by the -6 dB beam area in the given lateral-elevation plane (at the focus in this study).

2.3.4 Thermal Measurements

Thermocouple measurements of focal heating are proportional to the derated intensity reaching the focus in a given medium. When relating these measurements directly to intensity, nonlinear propagation in water at the higher drive voltages generally used for temperature measurements, viscous heating artifacts associated with the thermocouple, thermal conductivity and insulation surrounding the point of peak heating, and heterogeneities in the phantom or tissue material can confound the results. Focal heating measurements were made as a standard for the other focal output metrics for heating. The associated focal heating was measured in two ways. First, a 33 gauge, hypodermic needle, type-T thermocouple (Omega Engineering, Inc.,

Stamford, CT) was inserted between two slabs of tissue mimicking material submerged in 2% ethanol such that the ultrasound passed through 3.25 cm of solution and 0.5 cm of phantom before reaching the thermocouple at the focal depth of 3.75 cm. Second, the thermocouple was inserted between two slabs of tissue mimicking material such that the ultrasound propagation path consisted entirely of phantom material. Focal heating was evaluated using an M-mode sequence of 2.3-µs pulses at 7.9 kHz for a total insoniﬁcation time of 0.13 s transmitting at 55% system voltage,

23 unless otherwise speciﬁed. The data were recorded using a 16-bit Personal DAQ 3000 acquisition system (Omega Engineering, Stamford, CT) at 1 kHz. To evaluate all the temperature data, a running average of 100 samples (100 ms) was taken. The baseline temperature

was calculated as the average of several seconds of acquisition before the onset of the M-mode sequence. The maximum temperature rise above baseline for each sequence was recorded. Viscous heating artifacts were considered negligible for the low temperatures measured [94]. Frequency normalization of the focal heating measurements was performed. The measured temperature rises were divided by frequency to account for increased absorptive heating with frequency in the phantom material (see equation 1.6). Addi- tionally, the focal heating was normalized by nearﬁeld loss due to attenuation, which resulted in a complete normalization factor deﬁned by:

αfexp( 2αfz) (2.7) − where α is the attenuation of the tissue or phantom material in Np/cm/MHz, f is the frequency in MHz, and z is the propagation path through tissue or phantom material (0.5 cm for the ﬁrst set of measurements and 3.75 cm for the second set).

2.3.5 ARFI Displacement Measurements

ARFI displacement estimates are also proportional to the derated focal intensity reaching the focus in a given medium. A direct relationship between these estimates and intensity is confounded by underestimation of the tracked displacements, inertial eﬀects, and heterogeneities in the phantom or tissue.

The volumetric rate of heating (qv) and radiation force magnitude are closely

24 related by equations 1.6 and 2.3. For the same measurement medium, the relative

displacements might be used as an indicator of the relative qv and, therefore, ARFI displacements were measured as part of the TPM, as another method of evaluating the acoustic output at the focus. Each of the 1D transducers were coupled with gel to

a homogeneous CIRS (Computerized Imaging Reference Systems, Inc., Norfolk, VA) elastography phantom (α=0.5 dB/cm/MHz, E=4.5 kPa) and focused at 3.75 cm. The ARFI sequence consisted of two B-mode reference pulses (F/2 transmit, dynamic receive), a 180 µs pushing pulse (F/3), and 50 B-mode tracking pulses (F/2 transmit, dynamic receive, 9.3 kHz PRF) repeated at 20 lateral locations all focused at 3.75 cm axially. (A larger F/# was used for these measurements versus the thermal measurements to reduce displacement underestimation in ARFI data.) Data were gathered by moving the transducer to four independent positions on the phantom. After using a normalized cross-correlation algorithm to calculate the displacement at all spatial locations and times, the mean of the 20 lateral locations was taken and the maximum displacement in time within 25% of the focal depth was determined. ± In order to have an unbiased evaluation of the overall transducer performance, the displacements were normalized by the nearﬁeld loss and focal gains associated with the diﬀerent frequencies, as performed for the focal heating measurements.

2.3.6 Field and FEM Simulations

Field II simulations are directly scaled by the experimentally measured intensity and give a linear approximation of the beam proﬁle. These simulations can then be used as inputs to ﬁnite element method (FEM) models in order to estimate induced displacements and temperature rises [94,95]. Field II: Comparisons of transducer performance for proposed designs and variable pa-

25 rameters can be performed with simulations. Thus, methods for simulating relevant quantities for input into the TPM were investigated. In order to evaluate the ARFI and thermal beam profiles of the transducers, Field II simulations [60] were performed over a 3-D field. These beam profiles were simulated over a rectangular,

solid mesh which extended 1.5 cm in elevation, 1.5 cm laterally, and 8 cm axially. The elements were uniform cubes with a node spacing of 0.3 mm. There were a total of 704871 nodes and 675000 elements in this mesh. Gaussian bandwidths centered at the transducer’s center frequency were converted to the time domain to use as the impulse response input to Field’s transducer deﬁnitions for the linear or curvilinear

array. The transmitted intensity ﬁeld was then calculated. Mechanical FEM: In order to simulate the mechanical response of the CIRS phantom to the ARFI push, the outputs from Field II were normalized, thresholded by 1% of the maximum,

and scaled by the peak in situ experimental pulse-average intensity measurements (linear extrapolated small signal measurments derated using simulated tissue attenuation) [95]. These intensities were then used to calculate the force distribution for a given transducer’s focal conﬁguration according to equation 2.3. These radiation body force values were converted to nodal point loads by concentrating the body force contributions over the element volume. The spatial extent of the mesh allowed displacements to decay to negligible levels before reaching the mesh bound- aries for the duration of the simulation. Quarter plane-symmetry was used and the top and bottom surfaces were fully constrained. A Young’s modulus of 4.5 kPa and a Poisson’s ration of 0.499 were entered as inputs to the model. The balance of linear momentum was then solved numerically with the ﬁnite element analysis package, LS- DYNA3D program (Livermore Software Technology Corporation, Livermore, CA), using an explicit, time-domain integration method. Single-point quadrature was per-

26 formed with hourglassing control to avoid element-locking and to reduce numerical artifacts. Post-processing of dynamic displacement and stress ﬁelds was performed using LS-PREPOST (Livermore Software Technology Corporation, Livermore, CA) and custom-written MATLAB code.

Ultrasonic Tracking: The eﬀect of ultrasonic tracking, particularly underestimation for these purposes, on the measured displacement was simulated [95]. A uniform scattering phantom was represented as randomly positioned point targets of equal echogenicity. A ran- dom RF-data tracking line was generated from these scatterer locations using Field

II. The displacement vector field data from the FEM simulations were used to repo- sition the scatterers according to the eight surrounding nodes in the FEM model. The corresponding RF line was simulated again using Field II and the process was repeated for each time step in the simulation. Windowed cross-correlation of the simulated RF signals was used to estimate the displacements and correlation coefficients as in normal ARFI imaging. For each transducer configuration, five scatterer (speckle) realizations were simulated with this method and averaged before finding the maximum tracked displacement.

Thermal FEM: The temperature rise associated with the sequences tested in the National Phys- ical Laboratory (Teddington, England) thermal tissue mimicking material experimentally were simulated using FEM thermal models. The thermal distribution from one pulse was calculated as an input to LS-DYNA3D (Livermore Software Technol- ogy Corporation, Livermore, CA) by converting the intensity distrution in Field II to temperature rises above ambient using the solution to the bio-heat transfer equation when the time period under consideration is short enough to neglect conduction

27 eﬀects. This equation is as follows:

q T = v ∆t (2.8) cv

where T is temperature, qv is the rate of heat production per unit volume, cv is ◦ 3 the volume specific heat, and t is time. The thermal properties (cv=3.9 MJ/ K/m , K=0.52 W/m/◦K) of the thermal tissue mimicking material were assumed to be independent of temperature and entered as inputs to the model, assuming that the tissue is a thermally homogenous, isotropic solid. Symmetry conditions were used on the zero-lateral and zero-elevation planes and the top model surface (in contact with the transducer) was held at a fixed temperature (0, ambient temperature). A time- domain, implicit semi-iterative solver was used to simulate the cooling profile for one ultrasound beam and the resulting dynamic temperature fields were processed using LS-PREPOST2 (Livermore Software Technology Corporation, Livermore, CA) and custom-written MATLAB code. (This modeling method has been previously validated in [94].) The cooling profile of one 2.3-µs pulse was then summed over

1000 time steps at a repetition frequency of 8 kHz to simulate the total temperature rise from the experimental sequence.

2.3.7 Impedance Measurements

Element impedance measurements can be used as an indicator of the electric power delivered to the transducer for a given system voltage. If a large mismatch occurs (usually in the form of high reactive impedance for our system), then less power is delivered to the transducer and a lower acoustic output could result. The interpretation of this measurement depends on the system to which the transducer will be connected. Furthermore, an impedance mismatch is not the only cause

28 of decreased acoustic output and, therefore, must be considered along with the eﬃciency of the acoustic stack. Element impedance measurements for the array plus the cable were performed with the transducer in air using a HP4194A impedance analyzer (Hewlett Packard,

Palo Alto, CA). After calibrating the analyzer, the leads from the analyzer were connected to a ground pin and one element pin on the system connector for the transducer. Two to three elements were tested for each transducer. A custom Lab- View (National Instruments, Austin, TX) program was used to save the magnitude and phase data, which was output directly from the impedance analyzer in ohms and degrees versus frequency. The data reported are the impedance magnitudes and 1/cos(phase) at the transmit frequency being evaluated in the TPM averaged over the elements tested for the given transducer. The latter quantity gives a ﬁrst order approximation of the possible reduction in the reactive power demand on the system power supply, which would increase the average acoustic power capable of being delivered.

2.4 Results

2.4.1 Experimental TPM Components

As shown in Figure 2.2, the PH4-1 has the highest power output followed by the CH6-2 and then the CH4-1. The powers calculated by linear extrapolation for 55% system voltage were 196 W for the CH4-1, 345 W for the CH6-2, and 427 W for the

PH4-1. The powers calculated by linear extrapolation for 98% system voltage were 624 W for the CH4-1, 1095 W for the CH6-2, and 1357 W for the PH4-1. After accounting for the duty cycle (1.8%), the resulting average powers (3.5-24.2 W) are an order of magnitude below the general range of powers reported in the HIFU

29 literature (discussed further in Chapter 4). It should be noted that the maximum average electric power output from the Antares is 40 W.

5 ) 2

Intensity (W/cm 2

0 −10 −5 0 5 10 Lateral (mm) (a) (b)

Figure 2.2: (a) Standard deviation (in elevation dimension) in intensity across transducer face for PH4-1. (b) Acoustic power (measured by integrating Isppa values generated with 10% system voltage across the transducer face) for two curvilinear arrays (CH4-1=3.08 MHz, CH6-2=4.44 MHz), one 1D phased array (PH4-1=2.67 MHz), and one 2D phased array (2 MHz). The error bars represent the 8% repeatability for hydrophone measurements.

The distribution of the acoustic intensity in the focal plane for each transducer is shown in Figure 2.3. The bimodal distribution seen in the CH6-2 and PH4-1 are most likely due to the large diﬀerence between the lateral and elevation foci for these arrays. This bimodal distribution was not seen in the PH4-1 simulation, but it was

present in the CH6-2 simulation at 4.44 MHz. The Isppa was the maximum intensity in these distributions, even if this was not centered in elevation as was the case for

the CH6-2 and PH4-1. Isppa and Isapa at the focus were highest for the CH6-2 as shown in Figure 2.4. According to equation 1.6, this is predictive of the highest focal

heating with the CH6-2. At their respective center frequencies, the PH4-1 and CH6-2 had the highest ARFI displacements within measurement error followed by the CH4-1 experimentally, as shown in Figure 2.5. After normlization for loss due to attenuation and

30 CH4−1 CH6−2 −5 −5

0 0 Elevation (mm) Elevation (mm)

5 5 −2 0 2 −2 0 2 Lateral (mm) Lateral (mm) (a) (b) PH4−1 2D −5 −5

0 0 Elevation (mm) Elevation (mm)

5 5 −2 0 2 −2 0 2 Lateral (mm) Lateral (mm) (c) (d)

Figure 2.3: Focal plane (all with a focal depth of 37.5 mm) intensity plots (in dB) for the a) CH4-1 (3.08 MHz), b) CH6-2 (4.44 MHz), c) PH4-1 (2.67 MHz), and d) 2D (2 MHz) transducers measured with a membrane hydrophone. The color scale for each plot is -6 to 0 dB.

31 6 300 I (surface) I (focus) sp sp I (surface) I (focus) 5 sa 250 sa ) )

2 4 2 200

3 150

Intensity (W/cm 2 Intensity (W/cm 100

1 50

0 0 CH41 CH62 PH41 2D CH41 CH62 PH41 2D (a) (b)

Figure 2.4: Acoustic spatial peak and spatial average intensities at the (a) surface and (b) focus (37.5 mm) for two curvilinear arrays (CH4-1=3.08 MHz, CH6-2=4.44 MHz), one 1D phased array (PH4-1=2.67 MHz), and one 2D phased array (2 MHz).

increased energy absorption, the CH6-2 had the highest displacement. The FEM simulations, however, predicted that the CH6-2 would have the highest displace-

ment both before and after normalization. At their center frequencies, the PH4-1 also had the highest focal heating through a waterpath and through phantom material, as shown in Figure 2.6. FEM simulations through phantom material predict the PH4-1 and CH6-2 will have similar heating, but the CH6-2 would have the highest heating after normalizing for increased loss in the phantom material and energy absorption at the focus with frequency.

2.4.2 TPM Summary

Because of the low acoustic output of the 2D array, a summary of the measurements made on only the 1D arrays is shown in Figure 2.7 and Table 2.2. The displacement and focal heating measurements are normalized as described in equation 2.7 (where z did not include the water path) in Figure 2.7. Using equation 2.2, the thermal index was shown to trend with acoustic power for these small focal beamwidths. Although the displacement and temperature metrics were normalized

32 20 Displacement (µm) 18 Frequency Normalized Displacement 16 FEM Displacement (µm) Frequency Normalized FEM Displacement 14

Displacement 8

2 CH41 CH62 PH41

Figure 2.5: Experimentally measured and FEM simulated displacements. Maximum raw displacements within 25% of the 3.75 cm focus from acoustic radiation force (F/3, 180 µs ± interrogation) on the 1D probes. Effective displacements after normalization for the loss due to attenuation and increased energy absorption with frequency (CH4-1=3.08 MHz, CH6-2=4.44 MHz, PH4-1=2.67 MHz) are also shown. (These normalized displacements are divided by the maximum measured or simulated displacement before normalization to show all curves on same plot.) for absorptive gain at the focus and nearfield attenuative losses due to frequency, nonlinear effects and measurement artifacts were difficult to accurately account for.

Because none of these metrics ranked the transducers in the same order as the experimentally measured focal heating, a second set of measurements were performed on the 1D arrays at the same transmit frequency to avoid complicated frequency- dependent diﬀerences.

Using a transmit frequency of 2.67 MHz, ARFI displacements, focal intensities, acoustic power, electrical impedance, beamwidths, and gradients (over the -6 dB beamwidth) were measured and/or simulated for the 1D arrays. As shown in Ta- ble 2.3 and Figure 2.8, all of the metrics showed similar trends within the repeatability errors of the measurements. When considering the electrical impedance of these

33 5 o Temperature Rise (oC) Temperature Rise ( C) 2 4.5 Frequency Normalized Temperature Frequency Normalized Temperature FEM Temperature Rise (oC) 4 FEM Frequency Normalized Temperature 3.5 1.5

2.5 1 Temperature Temperature 2

1.5 0.5

0.5 0 CH41 CH62 PH41 CH41 CH62 PH41 (a) (b)

Figure 2.6: Temperature rises at the focus for 1000 line M-mode sequences with a PRF of 7.9 kHz using 10-cycle pulses with an unapodized F/1.5 configuration through a (a) waterpath and (b) phantom material are shown (except CH4-1, which was in the noise floor of the thermocouple). Effective temperature rises after normalization for loss due to attenuation and increased energy absorption with frequency (CH4-1=3.08 MHz, CH6- 2=4.44 MHz, PH4-1=2.67 MHz) are also shown. (These normalized temperatures are divided by the maximum measured or simulated displacement before normalization to show all curves on same plot.) FEM simulated temperature rises are also shown for the phantom propagation path case (b). arrays, the higher value for 1/cos(phase) for the CH6-2 indicates that reactive power demand on the Antares power supply is higher for this array compared to the PH4-1 at 2.67 MHz. The effects of thermal insulation and focal configuration were further investigated by applying the bio-heat transfer equation 1.2 to the experimentally measured lateral-elevation plane intensity profiles for the CH4-1 and PH4-1 at 2.67 MHz.

Three cases were evaluated for the peak heating profile through time: 1) intensity profile as measured, 2) intensity profiles normalized to have the same total power, and 3) intensity profiles normalized to have the same maximum intensity. As shown in Figure 2.9, the PH4-1 intensity profile predicts a higher temperature rise (as con-

ﬁrmed by experiment), but the CH4-1 is higher when the proﬁles are normalized for total power and maximum intensity. These latter cases show the impact of thermal

0.8

0.6 Acoustic Power I (focus) 0.4 sppa

Normalized Value I (focus) sapa Displacement 0.2 Temp. (water) Temp. (phantom) Thermal Index 0 CH41 CH62 PH41 (a)

0.8

0.6 Acoustic Power I (focus) 0.4 sppa

Normalized Value I (focus) sapa Displacement 0.2 Temp. (water) Temp. (phantom) Thermal Index 0 CH41 CH62 PH41 (b)

Figure 2.7: Summary of the therapeutic potential metric measurements obtained for the 1D arrays at their center frequencies (CH4-1=3.08 MHz, CH6-2=4.4 MHz, PH4-1=2.67 MHz) a) without and b) with normalization (the normalization factors for focal gain and nearﬁeld loss were applied). Each measurement was normalized by the maximum value across all probes in both plots. Note that the trends for Isppa and Isapa overlap.

insulation due to the beam proﬁle shape on the maximum temperature rise. Diﬀerences in the shock parameter for the low system voltage (5%) used in the

35 Parameter CH4-1 CH6-2 PH4-1 (3.08 MHz) (4.44 MHz) (2.67 MHz) Max Displacements (µm) FEM (55%) 3.5 12.5 6.6 Tracked FEM (55%) 3.2 0.3 11.2 6.7 6.3 0.4 ± ± ± Experiment (55%) 4.8 0.3 6.5 0.4 7.0 0.3 ± ± ± Hydrophone ( 8%) 2 ± Focal I, F/1.5 (10%, W/cm ) 109 266 174 Focal I, F/3 (10%, W/cm2) 48 133 65 Surface Power, F/1.5 (10%, W) 6.5 11.1 13.8 TI (55%) 0.62 0.99 1.40 Temperature (◦C, 10%) ± Surface (55%) 0.27 0.60 0.29 Focal EtOH (30/40/55%) -/-/2.53 -/-/1.43 1.68/2.61/3.64 Focal Phantom (55%) – 0.64 1.32 Focal FEM (55%) 0.16 0.28 0.26 Electrical Impedance ( 3%) ± Magnitude (Ohms) 99 94 140 1/cos(phase) 1.0 1.0 1.5

Table 2.2: All data were acquired at the center frequency of each transducer. The system voltage at which each measurement is acquired is given. All ARFI data were acquired with an F/3 excitation configuration and an F/2 tracking configuration, while temperature measurements used an F/1.5 configuration. A ‘–’ indicates the data were not acquired. The attenuation for the displacement and thermal simulations and experiments was 0.5 dB/cm/MHz, and the phantom and simulation Young’s modulus was 4.5 kPa.

intensity measurements versus the high system voltages (30-55%) in the tempera-

ture measurements may impact these measurements leading to discrepancies. As indicated in Figure 2.10, an increase in pressure resulting from increasing the system voltage is associated with nonlinear propagation in water and has a signiﬁcant impact on the amount of loss in water due to weak shock absorption. For the CH6-2

at 4.44 MHz, this plot indicates that at a depth of 3.75 cm from the transducer, 0% and 90% of the source intensity was attenuated prior to the focus due to nonlinear propagation at 10% and 98% system voltage, respectively.

36 1.2

0.8

0.6 Acoustic Power I (focus) sppa

Normalized Value I (focus) sapa 0.4 Displacement Focal Temp. (water) 0.2 Focal Temp. (phantom) Thermal Index

CH41 CH62 PH41

Figure 2.8: Summary of the therapeutic potential metric measurements obtained for the 1D arrays at 2.67 MHz.

2.5 Discussion

Each of the metrics for predicting the transducer’s effectiveness at heating tissue have certain caveats associated with them. The total acoustic power is a good indicator of the capacity of a transducer to heat tissue without taking focusing effects into account. It should be noted that for the transducers evaluated, the change in acoustic power was not proportional to the differing aperture sizes in elevation and, therefore, was a function of different transducer output capabilities.

Isppa and Isapa predict the transducer’s ability to heat tissue while taking focusing into account; however, they are generally measured in water (a medium more prone to high nonlinearities and weak shock than tissue) and at lower powers (linear regime of the transducer’s output) than those utilized in heating. Radiation force creates a complex, dynamic response aﬀected by energy absorption in the tissue, the movement of surrounding tissue with the point of interest, and inertial eﬀects. In addition, using ultrasonic displacement monitoring of these excitations can be complicated by

37 Parameter CH4-1 CH6-2 PH4-1 Max Displacements (µm) FEM (55%) 2.5 5.0 6.6 Tracked FEM (55%) 2.4 1.7 4.8 2.0 6.3 0.4 ± ± ± Experiment (55%) 2.5 0.1 7.3 0.3 7.0 0.3 ± ± ± Hydrophone ( 8%) 2 ± Focal I, F/1.5 (5/30%, W/cm ) 27.4/– 49.5/2268 48.6/2007 2 Focal Isa, F/1.5 (5%, W/cm ) 14.3 29.1 26.0 Focal I, F/3 (5%/30%, W/cm2) 10.5/– –/760 15/668 Surface Power, F/1.5 (5%, W) 2.4 2.7 3.7 TI (55%) 0.85 2.17 1.75 Temperature (◦C, 10%) ± Focal EtOH (30/40/55%) 0.55/1.05/1.84 1.11/1.97/3.27 1.68/2.61/3.64 Focal phantom (55%) 0.38 1.21 1.32 Focal FEM (55%) 0.17 0.33 0.26 Electrical Impedance ( 3%) ± Magnitude (ohms) 110 98 140 1/cos(phase) 1.1 2.0 1.5 Field Simulations (elevation / lateral / axial) -6 dB Beamwidth (mm) 2.4/1.8/16.2 5.4/1.2/15.3 5.4/1.2/13.8 Mean Gradient (dB/mm) 4.6/9.3/0.7 2.5/4.8/0.8 2.1/5.1/0.8 Max Gradient (dB/mm) 7.7/14.5/2.0 5.0/4.8/2.2 3.8/5.1/2.4

Table 2.3: All data were acquired at 2.67 MHz. The system voltage at which each measurement is acquired is given. All ARFI data were acquired using an F/3 excitation configuration and an F/2 tracking configuration, while the temperature measurements and Field II (beamwidth and gradient) simulations used an F/1.5 configuration. A ‘–’ indicates the data was not taken. The attenuation for the displacement and thermal simulations and experiments was 0.5 dB/cm/MHz, and the phantom and simulation Young’s modulus was 4.5 kPa. The Field II simulations had 0.3-mm resolution in all three dimensions.

decorrelation that varies as a function of the relative size and shape of the point

spread functions of the pushing and tracking beams [95]. Simplifying assumptions and judicious selection of ARFI sequence parameters were employed to compare the displacement estimates to focal heating; however, we found that ARFI imaging could not be used as a consistently accurate indicator of focal heating in simulation or experiment (see Figures 2.7 and 2.8 and Tables 2.2 and 2.3).

Even with appropriate normalization factors, a robust TPM for diﬀerent frequencies proved to be quite challenging, as evidenced in Figure 2.7. For example, higher

38 (a)

(b)

(c)

Figure 2.9: Normalized maximum temperature rise for the CH4-1 and PH4-1 transducers as calculated from their lateral-elevation intensity profiles (a) - intensity profile as measured, b) - intensity profiles normalized by total power, c) - intensity profiles normalized by maximum intensity) and the bio-heat transfer equation.

frequencies result in weak shock formation and increased nearﬁeld energy loss in water before lower frequencies. Therefore, the CH6-2 at 4.44 MHz lost energy due

1 0.16 1 1 0.14 0.8 2 0.12 2

0.1 0.6 3 3 0.08

Depth (cm) 0.4 Depth (cm) 4 0.06 4

0.04 5 5 0.2 0.02

6 6 0 1 2 3 4 5 6 7 1 2 3 4 5 6 7 Frequency (MHz) Frequency (MHz)

Figure 2.10: The fraction of the source intensity attenuated for diﬀerent propagation depths in water and diﬀerent transmit frequencies as determined by equations 2.4 and 2.5 [36]. Left - p0 (0.42 MPa) was as measured for 10% system voltage using the CH6-2 at 4.44 MHz; Right - Same as left but at 98% system voltage (4.72 MPa).

to weak shock absorption before the other probes at equivalent pressures. This phe-

nomenon helps to explain the relatively high focal intensity at low system voltage for the CH6-2 but the low focal temperature at a higher intensity. Reasons for the discrepancy between the previously validated FEM simulations and the experimental displacement measurements have not yet been determined. The intensity

measurements used as inputs to the FEM simulations were repeatable within 8% (if repeaked), while the experimental displacement estimates were repeatable within 9% (for diﬀerent locations on a homogeneous phantom). Additionally, the experimental results suggest that the displacement estimates do

not have a direct relationship to the focal heating observed and should only be used as a rough indicator or in conjunction with another metric. This was also found to be true in the FEM simulation results. At the transducer center frequencies, the CH6-2 had the highest normalized simulated displacement by a large extent, while it had the highest normalized temperature rise by a smaller factor (see Figures 2.5 and 2.6).

At 2.67 MHz, while the PH4-1 had the highest simulated ARFI displacements, the

40 CH6-2 had the highest focal heating. This diﬀerence, which does not have ultrasonic tracking errors, viscous heating artifacts, or nonlinear losses through water, further supports the claim that ARFI displacement cannot be used to predict focal heating. The complexities of the mechanical and thermal behavior are prohibitive.

Better agreement between focal temperatures and the other predictors was observed when comparing diﬀerent probes all transmitting at the same frequency, instead of at the individual center frequencies. When transmitting at 2.67 MHz on all the arrays, the acoustic power measured near the surface of the transducers cor- rectly ranked the focal heating output. For tightly focused beams (beam area at a

given depth is < 1 cm2), this result agrees with the thermal index. However, the ratios between the focal heating of each transducer were better approximated by the intensity and displacement estimates, which include focal gains. The spatial distribution of the ultrasound beam plays a role in the peak heating achieved. For instance, small heated volumes with sharp gradients to body temperature have less thermal insulation to diffusion than larger volumes with gradual temperature fall-off. For the short insonification times used in the heating sequences herein, the spatial distribution has a negligible effect, but should be taken into account for longer heating regimens, as shown in Figure 2.9. Furthermore, the theory for weak shock absorption presented herein assumes an infinite plane wave. However, differences in focusing between arrays also play a role, where a more tightly focused beam has a greater weak shock absorption within a few millimeters of the focus than a beam with a weaker focusing gain [28]. Weak shock absorption also increases for greater source pressures. Therefore, the thermal measurements (through water) were made under conditions eliciting a larger shock parameter than those of the intensity measurements. Based on the complexities introduced by weak shock absorption and changing

41 beam proﬁles as the focal depth and frequency for a given transducer are changed, direct comparisons between the intensity measurements and thermal measurements (through a water path) were not possible. Furthermore, the eﬀect of the spatial distribution of the beam on the mechanical response of the tissue to a radiation

force excitation and the thermal insulation of the peak heating differs. As a result, a robust therapeutic potential metric could not be formulated based on these metrics. In contrast to the focal temperature and intensity measurements, the low impedance for the CH6-2 indicates that for the same applied voltage, the electrical power across each element will be greatest for the CH6-2. However, the ability of the transducer to efficiently convert electric to acoustic energy also determines the acoustic output. One of the challenges with creating a non-destructive and robust therapeutic potential metric is that this conversion efficiency can change between drive levels [13], such as the low drive levels used for testing purposes in the metric and high drive levels for therapeutic applications.

2.6 Conclusions

A robust therapeutic potential metric could not be formulated. However, insights into the challenges associated with comparing hydrophone, ARFI imaging, and temperature measurements were gained. The CH6-2 and PH4-1 probes were shown to have the most acoustic output in terms of ARFI displacement, intensity, and temperature. Another important ﬁnding was that ARFI displacements could not be used to accurately predict focal heating in experiment and simulation. This work did provide information for future studies. Because of the exponential loss in energy delivered to the focus with depth, a shallow depth of 2 cm was chosen for the liver ablation studies. By taking into account an increase in focal gain and

42 attenuation with frequency, the optimum frequency for this depth was found to be 4.3 MHz (as shown in Figure 2.1) which closely corresponds to the center frequency of the CH6-2. Therefore, this work provided a basis for choosing a commercial transducer for work in Chapter 4.

2.7 Acknowledgments

I would like to thank Siemens Medical Solutions, USA, Inc. for their system support, Dr. Gregg Trahey for his valuable insights, and Stephen Rosenzweig for his work on the automated peaking setup. This work was supported by an NSF Fellowship and NIH grants R01 EB002132 and R01 CA114075.

43 Chapter 3

Custom Diagnostic Transducers

In the previous chapter, the acoustic output of standard, diagnostic transducers was evaluated to determine the transducer with the highest potential for therapeutic

use. In this chapter, several design modiﬁcations are investigated for diagnostic transducers to increase their power output while avoiding thermal damage. These include 1) using a PZT-4 composite multilayer instead of PZT-5H, 2) employing an acoustically lossless lens material, 3) adding passive cooling, and 4) testing a low

loss tangent, capacitive micro-machined transducer.

3.1 Introduction

High-intensity focused ultrasound (HIFU) has progressed over the past decade to become a viable therapeutic method and is valuable as a non-invasive alternative to many surgical procedures [62]. Current HIFU implementations require a separate imaging method or diagnostic ultrasound probe for localization of the region of interest and monitoring of the ablation, in addition to the therapy probe [14, 124]. The imaging method must be co-registered with the therapy probe in order to accurately target the ablation regimen. Alternatively, co-registered imaging and therapy can be achieved using a non-standard, combined therapy transducer and imaging transducer in the same probe [25,73].

44 The design of a transducer capable of performing HIFU and imaging operations with the same elements has proven quite challenging. This stems from the fact that HIFU transducers are built to have high-transmit capabilities with low electromechanical loss [45]. To prevent energy loss behind the transducer and maximize emission for high-transmit, air backing is generally used [13, 112], which results in a longer impulse response and a narrower bandwidth. On the other hand, imaging transducers are designed for high imaging quality and therefore have a broad bandwidth [83], which allows for shorter pulses and hence higher resolution. The tradeoff between power efficiency and operating bandwidth is one of the main issues in designing a dual-mode array [37]. Dual-mode ultrasound array (DMUA) designs generally start from a HIFU array design and make alterations toward acceptable imaging. We hypothesize an alternative approach, where additions and modifications to a diagnostic transducer design

could be made to increase the power output or simply reduce the internal heating during high-power operation. This approach could lead to dual-mode transducers with higher image quality and, therefore, better localization and monitoring capabilities, while having the ability to perform therapeutic applications. Depending on the

desired lesion size, the required thermal dose, and ablation-inhibiting eﬀects (e.g., phase aberration caused by the skull or highly attenuating intervening structures), the preferred starting point for the design of a dual-mode transducer may vary. For small spot ablations in a low thermal dose tissue (e.g., liver) with no intervening structures (e.g., an intra-operative scenario), starting a dual-mode design from a

diagnostic transducer might be advantageous. This chapter serves to evaluate the effectiveness of several different modifications to diagnostic transducers in order to make them capable of therapeutic applications. Several custom, prototype transducers have been tested with features that could

45 be beneficial in therapeutic applications for reducing internal transducer heating, which is one of the primary limitations of using diagnostic transducers for therapy. As such, face heating, ARFI displacement, focal intensity, and surface pressure were metrics evaluated to determine the impact of each modification. However, these modifications should not come at the expense of image quality. Therefore, lesion contrasts in B-mode and ARFI images were metrics used to evaluate the image quality.

3.2 Background

In this section, the origins of heating in diagnostic transducers are reviewed as well as possible methods of reducing this heat.

3.2.1 PZT Transducer Heating

Ineﬃciencies in PZT lead to the production of heat during transducer operation. The dissipation factor, or loss tangent (tanδ, where δ is the phase angle between voltage and current in a capacitor), is a measure of the power lost in a dielectric, which is calculated as the ratio of the imaginary (related to dissipation of energy within the medium) to real (related to the stored energy within the medium) components of the permittivity constant [45,53]. A material with a high loss tangent will experience more internal heating than one with a low loss tangent. PZT-5H, typically used in diagnostic probes, has a loss tangent of 2.0-2.7%, while PZT-4 and PZT-8, used in some HIFU transducers, have loss tangents of only 0.2-0.4% and 0.1%, respectively [45, 133]. Furthermore, a high Curie temperature is desired to decrease the risk of depoling the elements [45]. However, the high impedance of PZT-4 and PZT-8, due to their low dielectric constants, makes it necessary to have larger ele-

46 ments to match the source impedance of conventional driving electronics [45], which has a negative impact on beamforming capabilities. One approach to increasing the dielectric constant of these materials is to utilize multilayer transducers that can be designed to reduce the element impedance mismatch [47]. The difference in heating between transducers constructed with multilayer PZT-4 versus single layer PZT-5H is evaluated herein and will be submitted for publication in IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control. Heating due to PZT inefficiencies or absorption of the ultrasound in the acoustic stack (backing layer, PZT elements, matching layer, and lens) can affect the bonds between different materials in the transducer, especially in the case of thermal shock [76]. Thermal shock is caused by transient temperature gradients yielding rapid rates of thermal stress application resulting from different coefficients of expansion in a composite material [76]. The likelihood of thermal shock can be reduced by more gradual heating of a material or using a material with a higher thermal conductivity (K, thermal diffusivity times specific heat), tensile strength, and toughness (i.e. amount of energy a material can absorb before rupturing, resistance to fracture when stressed, area under stress-strain curve) as well as a lower thermal coefficient of expansion (CE). For many commercial probes, epoxy is a common lens and matching layer material. An epoxy compound’s resin matrix has a significant influence on stress, adhesion, and thermal shock properties (among other things) [99]. In addition to thermal shock, delamination (separation of bonds) of the matching layers or lens, and depoling of the PZT are concerns caused by raised temperatures in diagnostic transducers. Once delamination occurs, the acoustic energy reverberates within the transducer leading to additional overheating [41]. Generally, the majority of electrical input power is dissipated as heat in the lens of a transducer [110]. While epoxy (usually doped with another substance, such as silver or tungsten) is one of the

47 more common lens and backing/matching layer materials for diagnostic transducers, other materials with high temperature tolerances can also be used, such as MacorTM (a machineable glass ceramic) for matching layers and TPX (polymethylpentene, a thermoplastic polymer) for the lens. The relevant thermal properties are listed in

Table 3.1 for common acoustic stack materials (as well as liver and muscle), which show PZT and Macor to be more resistant to thermal shock and easier to cool than epoxy and polystyrene due to their higher Ks and lower CEs. According to this table, polystyrene is clearly a poor material to use in the acoustic stack from a thermal standpoint as it will cool more slowly than the tissue that one might be

trying to heat in a HIFU application. Conversely, Macor is an excellent material. Because of signiﬁcant heat dissipation in the lens, a probe with a practically lossless lens material (proprietary) was tested in this work.

Material κ( 10−7m2/s) K (W/m/◦C) CE (1/◦C) × −5 Epoxy (M, B, L) 2.26 0.188-0.35 3.6-5.4 x 10 Polystyrene (L) 0.96 0.1-0.13 6.73 x 10−5 PZT 3.14 1.1 0.4-0.8 x 10−5 Macor (M) 18.48 1.46 0.93 x 10−5 TPX (L) – – 1.2 x 10−5 Liver 1.45 0.488 – Muscle 1.25 0.489 –

Table 3.1: Thermal diffusivities (κ), conductivities (K), and coefficients of expansion (CE) for common transducer materials and tissues. The first property affects the rate of heat diffusion (see equation 1.2 without perfusion). The latter two thermal properties relate to the likelihood of thermal shock. (M = matching layer, B = backing layer, L = lens)

3.2.2 Passive and Active Cooling

Passive cooling of transducers can involve any cooling mechanism not requiring the use of a pump or motor. Heat sinks, such as copper plates (used in the

48 Duke-2 transducer discussed herein), are among the more common methods of passive cooling. Another passive cooling design includes the use of a phase change material between the patient and the lens that is capable of absorbing heat from the lens through its own heat of fusion (heat required to convert solid to liquid)

without increasing in temperature [87]. Although passive cooling is suﬃcient in some applications, others require active cooling for maintenance of a reasonable lens temperature [74]. Active cooling makes use of a heat exchanger. This could include a closed loop circulating coolant system or a single pass ﬂow system passing through a heat ex-

changer, heat pipe, thermoelectric cooler, evaporative/condenser system, and/or phase change material [74]. Gas or liquid coolants can be used [74]. Deardorﬀ and Diederich [33] pumped water through thin-walled (0.1 mm) stainless steel tubing running directly behind a cylindrical PZT transducer in their DC interstitial ultra-

sound applicator to improve thermal performance without degrading the acoustic beam output. For extremely high output applications, such as HIFU, active cooling provides a means of controlling the internal transducer temperature to avoid transducer damage or excessive heating of the surface (usually the lens) in contact with

the patient. Active cooling was not explored herein.

3.2.3 cMUTs and Fresnel Focusing

Capacitive micro-fabricated ultrasonic transducers (cMUTs) are an alternative to traditional PZT transducers and can achieve 3-D imaging using delay and Fresnel focusing [26]. A cMUT consists of many metalized membranes supported above a bottom electrode on a substrate. When an alternating current is added to a direct current bias voltage applied between the electrodes, a sinusoidal membrane vibration is obtained. Similarly, if the biased membrane is exposed to an incoming acoustic

49 ﬁeld, electrical current is delivered to an external load [130]. These voltages must be limited to avoid collapse of the membrane onto the substrate, or snapdown [90]. For a plate capacitor, collapse is said to occur when the displacement reaches one- third of the original gap distance between the membrane and bottom electrode. The

electrostatic force acting on the top electrode is given by [90]:

2 ε0V S F = dc (3.1) 2g2

where ε0 is the dielectric constant, Vdc is the DC voltage, S is the surface area of the capacitor, and g is the eﬀective gap height. The stiﬀness across the membrane dictates the resulting displacement from this force, and thus the source pressure. If multiple DC bias lines run orthogonal to the azimuth elements instead of one large continuous electrode, then a pattern of bias voltages can be used as a Fresnel lens to focus in elevation [27]. The appropriate phase, φ, for a bias line at elevation y can be computed geometrically using [27]:

φ =2πf[ z2 + y2 z]/c (3.2) p −

where f is frequency, z is the focal depth, and c is the speed of sound. The phase

range of 0 to 2π is then split into discrete bins equaling the number of bias voltage levels available. The number of voltage polarity (or sign) transitions in the bias pattern should not exceed the number of cycles in the transmit pulse. With a limited number of bias voltages for Fresnel focusing, the side lobes can be quite high

compared to time-delay focusing; yet, when the available system channel count is limited, Fresnel focusing in the elevation dimension does allow for a much larger elevation aperture and smaller slice thicknesses compared to conventional lens and multi-row designs [27].

50 The coupling coeﬃcient, or electromechanical conversion, for capacitive microfabricated ultrasonic transducers is deﬁned by [130]

2 Emech 1 kT = = Eelec , (3.3) Etotal 1+ Emech

where Etotal = Eelec + Emech and the subscript T denotes clamping in the direction transverse to the electric ﬁeld. cMUTs are capable of large bandwidths because of

2 their high coupling coefficient (kT =0.85). In contrast, the coupling coefficient for 2 PZT transducers, k33, gives the relative portion of stored input electrical energy converted to stored mechanical energy, and vice versa, and has typical values of 0.56 in the absence of lateral restraint and 0.25 for clamped PZT [53]. If the transducer does not generate much heat, then the thermal diffusion and

shock become moot. Capacitive micromachined ultrasonic transducers fall into this category because they have low internal heating due to a low loss tangent (0.02%) [82]. Because of this low loss, we have investigated cMUTs herein. Further- more, cMUTs have long-term CW capabilities as well as high coupling coeﬃcients

and, thus, large bandwidths (encouraging use at higher frequencies for imaging resolution and lower frequency for therapy penetration). Some limitations of cMUTs include breakdown of the insulating layers from high voltages and failure of metal traces from high current densities, which can be reduced by increasing the size of

the gaps and thickness of the traces [126].

3.3 Methods

Several transducers were designed, built, and tested using diﬀerent heat manage- ment ideas. The designs 1) used a PZT-4 composite multilayer instead of PZT-5H,

51 2) replaced the lens with an acoustically lossless material, 3) added passive cooling, and 4) replaced PZT with a capacitive micro-machined transducer. The tests to evaluate improvements in output from these modiﬁcations were transducer heating, intensity measurements, acoustic radiation force displacements, and diagnostic im-

age contrast analysis. When transducer damage occured, the golden spike test (see section 3.3.5) was used to evaluate the cause of failure.

3.3.1 Transducer Heating

Transducer face heating was evaluated by placing a 33 gauge hypodermic (Cu/Cu- Ni) type-T thermocouple (Omega Engineering, Stamford, CT), or 36 gauge (Cu/Cu-

Ni) type T thermocouple (Omega) on top of porcine muscle or tissue mimicking

3 material (α=0.5 dB/cm/MHz, cv=3.9 MJ/K/m , K=0.52 W/m/K) and coupling it to the transducer face with gel or water, as speciﬁed in the results for the given transducer. The 33 gauge thermocouple data were recorded using a 16-bit Personal DAQ 3000 acquisition system (Omega Engineering, Stamford, CT) at 1 kHz. The 36 gauge thermocouple data were recorded using a 16-bit data acquisition system with a sampling frequency of 5 kHz (SuperLogics, Waltham, MA) that was precalibrated with cold junction compensation (CJC) and shielded from the environment with an aluminum cover. The setup was chosen based on the equipment available at the time of transducer evaluation. Peaking was only performed once for most face heating measurements. For a lensed array, the higher absorption due to the longer path in the middle of the lens would make that area slightly hotter than the edges. For an array without a lens, there might still be a similar, though smaller, gradient, because there is an inactive region on the edges of the elements. The ground connection can also serve as a heat sink. For a well-built transducer, the variation in heating or amplitude due to

52 bonding and other inconsistencies should be a good deal smaller than the gradient due to design and material property considerations [132]. To evaluate all the temperature data, a running average of 100 samples (100 ms for the 33 gauge thermocouple and 20 ms for the 36 gauge) was taken. The baseline

temperature was calculated as the average temperature over several seconds before the onset of the therapy sequence. The maximum temperature rise above baseline for each sequence was recorded.

3.3.2 Intensity Measurements

Surface Intensity:

Surface intensity measurements were made using a calibrated membrane hydrophone (0.2-mm spot size with a preamp from Onda Corporation, Sunnyvale, CA, or 0.6-mm spot size from Sonora, Longmont, CO) and 3-D stepper motor controlled translation stage (Newport, Irvine, CA) operated using LabView (National Instruments, Austin, TX). Peak positive pressures for 10-cycle pulses were measured in 0.5-mm increments across the lateral-elevation plane at a depth approximately 8 mm from the surface of the transducer. The Isppa values were calculated from each measured pressure waveform according to equation 2.6. The total acoustic power output was calculated at the surface of the transducer by integrating the pulse- average intensities across the active aperture. Focal Intensity: Focal intensity measurements were made with a calibrated membrane hydrophone (Sonic Technologies (Wyndmoor, PA) 0.6-mm spot size or a Sonora (Longmont, CO)

0.4-mm spot size or a Siemens-owned, calibrated membrane hydrophone). After using custom LabView software to automatically ﬁnd the location of peak pressure at the focus, the data were recorded at system voltages of 1-10%. To determine

53 the intensities at higher system voltages without risking damage to the hydrophone, a linear regression was applied to the measured intensities of the linear waveforms versus the squared system voltage. This method also serves to avoid the eﬀects of saturation in the measurement of higher acoustic outputs through water, which are less signiﬁcant in tissue.

3.3.3 Acoustic Radiation Force

Acoustic radiation force impulse (ARFI) imaging was used as another method of evaluating the acoustic output at the focus. The transducer was coupled with water to a CIRS (Computerized Imaging Reference Systems, Inc., Norfolk, VA) elastography phantom. The ARFI sequences consisted of one or two conventional, B- mode reference pulses (dynamic receive), a high-intensity, radiation force excitation pulse (unapodized), and several B-mode tracking pulses (dynamic receive) repeated at several lateral locations. The F/# used for the radiation force excitation, transmit frequencies, and focal depths used varied between transducers (and are provided with the results). Summed radio-frequency (RF) echo data were stored for oﬀ-line processing. A 1-D normalized cross-correlation algorithm (3 wavelength kernel and 0.385-mm unbiased search region) was used to calculate the displacement at all spatial locations and times. The maximum ARFI displacements were calculated by averaging all of the lateral locations in a region of homogeneous phantom and taking the maximum within 25% of the focal depth and through all and time, ± after removing displacement estimates with correlation coeﬃcients less than 0.98. The average displacement at a given time was determined in a similar way.

54 3.3.4 Diagnostic Image Quality Assessment

Because of the inherent trade-offs associated with design modifications to increase acoustic output in diagnostic transducers, the impact of these modifications on the

image quality must be assessed. Contrast in B-mode and ARFI images as well as qualititative B-mode resolution were evaluated for these purposes. An RMI 404GS (Gammex, Middleton, WI) phantom containing cysts and point targets was used to compare overall image quality and contrast in B-mode. In

ARFI, a lesion in a CIRS, tissue mimicking, elastography phantom was imaged. The contrast for B-mode and ARFI were calculated as:

D C =1 i (3.4) − Do

where Di is the brightness (B-mode) or displacement (ARFI) inside the lesion and

Do is an equal-sized region at the same depth outside the lesion.

3.3.5 Golden Spike

The golden spike test analyzes the integrity of the acoustic stack and identiﬁes heterogeneities in the layers, such as those caused by delamination. It was performed using a 0.5-mm diameter, 7-MHz transducer driven with short pulses by a Panametrics 5800 pulser/receiver (Waltham, MA). This transducer was aimed at

the lens of the transducer under test and acquired pulse-echo data over 2 µs in the acoustic stack. The 7-MHz transducer was scanned across the entire lens surface of the test transducer (0.5-mm increments in lateral and elevation directions). The golden spike usually has some internal energy circulation footprint in the time domain. To remove this, the golden spike was moved 10 mm away from the surface

of the test transducer and reference data associated with the golden spike footprint

55 was gathered over an area large enough to average out noise. The reference data was subtracted from the acoustic stack data to produce images of the acoustic stack in the axial-lateral and lateral-elevation planes.

3.4 Results

3.4.1 PZT-4 Multilayer Composite

PZT-4 has a lower loss tangent and thus should heat less than PZT-5H, but it also has a higher impedance making matching of driving electronics challenging. (We worked with W.L. Gore and Associates to investigate a multilayer PZT-4 technology to overcome this challenge. This work was funded by the NIH.) The Gore transducers consisted of 192 elements with a pitch of 0.03 cm, an elevation height of 1.2 cm, a lens focus of 7.1 cm, and bandwidths centered around 3.5 MHz. The PZT-4 composite was a 2-2 type with the piezo and polymer layers running perpendicular to the elements. When the array was diced, this resulted in an effective 1-3 structure, but with different fillers in elevation and azimuth [132]. The ARFI and thermal parameters used in this evaluation are shown in Table 3.2. The Gore transducers, PZT5-H single layer control and PZT-4 triple layer composite, were driven at system voltages (40% and 24%, respectively) resulting in equal focal intensities for a given frequency using a 6 cm lateral focus with a 1 cm aperture. Face heating measurements were performed using an M-mode sequence with a 7.06 kHz PRF and a 2.8% duty cycle for a total insonification time of 5 seconds, as shown in Figure 3.1. Automated peaking was utilized and error bars represent the standard deviation over four, re-peaked, independent measurements (with statistical outliers removed). These results indicate a decrease in heating in the PZT-4 array compared to the control array for two different driving frequencies within the

56 Test ARFI Thermal Phantom CIRS Liver NPL Phantom Push 2.5 2.5/3.6 Frequency (MHz) F/# 3.5 6 Focus (cm) 6 6 Track PRF (kHz) 4.1 7.06 Pulse 260 4 Length (µs) Track 8.5 5000 Time (ms) Locations 50 1 (24.4◦ lateral) (M-mode)

Table 3.2: Sequence parameters for the ARFI and thermal testing of the Gore PZT5-H single layer and PZT-4 composite triple layer transducers.

2.5 MHz 3.6 MHz 6 1 layer, PZT5−H (4.6 +/− 0.3C) 9 1 layer, PZT5−H (7.3 +/− 0.7C) 3 layer, PZT−4 (1.8 +/− 0.3C) 3 layer, PZT−4 (2.9 +/− 1.0C) 5 8 7 4 6

3 5 4 2 3 2

Temperature Rise (deg C) 1 Temperature Rise (deg C) 1 0 0

0 5 10 15 0 5 10 15 Time (s) Time (s) (a) (b)

Figure 3.1: Face heating from a 1 cm aperture using a PZT5-H, single layer control array and a PZT-4, triple layer, composite array was measured by a 33 gauge, type-T, hypodermic needle thermocouple coupled with a thin layer of water to thermal tissue mimicking phantom material. Example temperature versus time curves are shown. Note: Due to better impedance matching, comparable Isppas were achieved with a lower driving voltage for the PZT-4 multilayer. Transmit frequencies of a) 2.5 MHz (focal Isppa=240 2 2 W/cm ) and b) 3.6 MHz (focal Isppa=607 W/cm ) are shown. Error bars show the standard deviation over four re-peaked measurements (with statistical outliers removed). transducer bandwidth. The B-mode quality between these two arrays was similar, as shown in Figure 3.2.

57 For the images shown, the contrast of the hyperechoic lesion was -0.10 and -0.04 for the control and triple-layer array, while the hypoechoic lesion contrast was 0.95 and 0.94, respectively.

(a) (b)

Figure 3.2: B-mode images of a RMI phantom including cysts and point targets using a) PZT5-H, single layer, control array (contrast: hyperechoic = -0.10, hypoechoic = 0.95) and b) PZT-4, triple layer, composite array (contrast: hyperechoic = -0.04, hypoechoic = 0.94). Note the similar image quality.

These arrays were also compared using ARFI images of a CIRS liver tissue mimicking phantom (0.7 dB/cm/MHz, 5 kPa) with a stiffer lesion [40]. Each ARFI sequence consisted of 50 pushing locations spaced 0.48◦ apart to fully encompass a specific lesion and some surrounding background phantom material for contrast and displacement comparisons. Each push was a 2.5 MHz, unapodized, F/3.5, 160-µs pulse focused at 6 cm. Each 2.5 MHz tracking pulse used an F/2 configuration. Figure 3.3 shows ARFI images from these arrays when applying increasing system voltages for the ARFI pushing pulses. The increased jitter in (f) versus (e) is not due to any significant difference in B-mode contrast or SNR. Table 3.3 shows the lesion contrast for these images and the average displacement in the phantom material surrounding the lesion 1.3 ms after radiation force excitation (time point where good visualization of lesion occured). For 1quivalent focal pressures (57% transmit

58 voltage for the 3-layer probe and 70% transmit voltage for the control), as extrapolated from hydrophone measurements, the displacement was equivalent between the two probes (1.2 0.2 µm (multilayer) versus 1.4 0.2 µm (control)). ± ±

(a) (c) (e)

(b) (d) (f)

Figure 3.3: ARFI images of a lesion in a CIRS elastography phantom mimicking liver tissue 1.3 ms after radiation force excitation using 57% (a), 67% (b & c), 77% (d & e), and 87% (f) of the maximum system voltage for the radiation force excitations. Top Row: PZT-5H, single layer, control array, Bottom Row: PZT-4, triple layer, composite array.

In further testing, evidence of damage was seen with the 3-layer array but not the control. The thermal sequence at 2.5 MHz was transmitted at higher system

voltages (55% for control, 43% for 3-layer) than those shown in Figure 3.1. Although these voltages resulted in equivalent focal intensities, only the 3-layer array showed visible damage in B-mode (not shown). However, a higher system voltage was used with the 3-layer array during radiation force excitations without damage. These

59 Control 3-layer System Contrast Displacement System Contrast Displacement Voltage (%) (µm) Voltage (%) (µm) 67 0.7 1.2 57 0.7 1.2 77 0.7 1.6 67 0.7 1.7 87 0.7 2.1 77 0.7 2.1

Table 3.3: ARFI image lesion contrast and average displacement (1.3 ms after radiation force excitation) in homogeneous phantom material surrounding the lesion at depths from 5-6 cm for the Gore probes. Each radiation force excitation was a 2.5 MHz, unapodized, F/3.5, 160-µs pulse focused at 6 cm. The variations over 4 trials in contrast and displacement were negligible ( 0.02 for contrast and 0.07 µm for displacement). For comparable ≤ image metrics, lower system voltages were used for the 3 layer array. sequences had a lower duty cycle (3% instead of 4%) and varied active elements with the changing push locations. The PZT-4 composite multilayer worked well. For matched acoustic output, it had comparable image quality, less heating, and allowed for lower driving voltages than the control. However, this transducer was more fragile than the control. Dam- age occured while transmitting the same focal acoustic output from these arrays. In general, the results of this work would support using a PZT-4 composite multilayer transducer, however further investigation into the source of the damage is necessary.

3.4.2 Lossless Lens

Because a large percentage of transducer heating is in the lens [110, 121], the beneﬁts of a lossless lens material were investigated. The Siemens Acuson 8L5 transducer was constructed with a lens material with negligible acoustic attenuation; thereby, the heating normally associated with absorption in the lens should be greatly reduced compared to equivalent transducers with standard lens materials. The 8L5 tranducer has 128 elements with a pitch of 0.03 cm, an elevation height of 0.4 cm, a lens focus of 1.8 cm, and a 65% band-

60 width centered at 7.2 MHz. The Siemens VF10-5 transducer was used as a control, which is a linear array with a standard lossy Siemens lens material (proprietary). Comparisons were made both with respect to face heating and acoustic output as indicated by ARFI displacement (see Table 3.4). The VF10-5 has 192 elements with a pitch of 0.0201 cm, an elevation height of 0.5 cm, a lens focus of 2.0 cm, and a 51% bandwidth centered at 7.0 MHz.

Test ARFI Thermal Phantom CIRS Breast Porcine Muscle Push 5.71 5.71 Frequency (MHz) F/# 1.5 1.5,2,3 Focus (cm) 1.8 1.8 Track PRF (kHz) 17.6 7.9 Pulse 70 1.8 Length (µs) Track 4 1000,5000 Time (ms) Locations 40 1 (M-mode)

Table 3.4: Sequence parameters for the ARFI and thermal testing of the 8L5 (lossless lens) and VF10-5 (control) transducers.

To compare the acoustic output at the focus between the lossless lens (8L5) transducer and a control, ARFI imaging was utilized. Using a 0.7 dB/cm/MHz homogeneous, CIRS elastography phantom with a stiffness of 4 kPa, a 400-cycle, 5.71-MHz pulse focused at 1.85 cm was transmitted with an F/1.5 configuration at 40 lateral locations. The resulting maximum displacements at increasing system voltages are shown in Table 3.5. For the same system voltage, the VF10-5 displaced the phantom more than the 8L5. A linear fit between the maximum displacement and system voltage squared was employed to determine voltages for equivalent system output at which to compare

61 System 8L5 VF10-5 Voltage (%) (µm) (µm) 35 – 2.3 0.1 ± 45 2.9 0.1 3.6 0.1 ± ± 55 4.3 0.2 5.4 0.6 ± ± 65 5.7 0.2 6.8 0.2 ± ± 75 7.2 0.2 – ±

Table 3.5: Comparison of acoustic output at the focus of the 8L5 (lossless lens) transducer with a control using maximum ARFI displacements. Each radiation force excitation was a 5.7 MHz, unapodized, F/1.5, 70-µs pulse focused at 1.8 cm. The standard deviations were over 4 ARFI acquisitions at diﬀerent phantom locations. face heating between the 8L5 and VF10-5 transducers. Face heating was compared using M-mode sequences with 10-cycle, 5.7-MHz pulses transmitting at a PRF of 7.9 kHz for 1 or 5 seconds. Higher temperature rises for equivalent sequences were observed using the 8L5 transducer, as shown in Table 3.6.

8L5 VF10-5 F/# Time System Temperature System Temperature (s) Voltage (%) Rise (◦C) Voltage (%) Rise (◦C) 3 5 12 1.4 10 1.0 2 5 12 1.4 10 0.9 1.5 5 12 1.3 10 0.8 1.5 5 26.1 7.3 25 5.8 1.5 5 45 20.8 38.3 14.1 1.5 1 40.2 4.1 35 2.9 1.5 1 54.5 7.6 45 5.0 1.5 1 68.4 10.5 55 7.5

Table 3.6: Face heating comparison between the 8L5 (lossless lens) transducer and VF10- 5 (control) using M-mode sequences with 10-cycle, 5.7-MHz pulses transmitting at a PRF of 7.9 kHz for 1 or 5 seconds. The system voltages in each row are such that the focal acoustic output, according to ARFI displacement, was equal between the two arrays. Each temperature value is the average over 4 sequence ﬁrings (no repeaking). The precision of face heating measurements without repeaking is 8%. ±

The B-mode image quality of the 8L5 was compared to that of the VF10-5 using

62 the RMI 404GS phantom, as shown in Figure 3.4. The hypoechoic lesion contrast was 0.42 for the 8L5 versus 0.27 for the VF10-5, while the hyperechoic lesion contrasts were -0.27 and -0.21, respectively.

0 0

5 5

10 10

15 15

20 20 Axial (mm) Axial (mm) 25 25

30 30

35 35

−10 0 10 −10 0 10 Lateral (mm) Lateral (mm) (a) (b)

Figure 3.4: B-mode image quality comparison between the a) 8L5 and b) VF10-5. The hypoechoic lesion contrast was 0.42 for the 8L5 versus 0.21 for the VF10-5, while the hyperechoic lesion contrasts were -0.27 and -0.21, respectively.

The 8L5, lossless lens, transducer did not work well. A greater system voltage was required to yield the same acoustic output at the focus in terms of ARFI displacement. Using voltages that yielded equivalent focal output, the transducer face heating of the 8L5 transducer was more than that of the VF10-5 control array. As a result, this lossless lens design would not be recommended for therapeutic use. However, from a materials perspective, this lens improved B-mode image quality.

3.4.3 Passive Cooling

Another approach to decreasing transducer heating is to introduce passive cooling features. The Duke-2 transducer was designed in conjunction with Siemens Medical Sys-

63 tems with a thermally conductive backer and copper interconnects with low loss, high impedance matching layers in order to greatly reduce internal transducer heating [98]. This array has 192 elements with a 0.2-mm pitch, an elevation height of 6 mm, a lens focus of 3.75 cm, and a bandwidth ranging from 2.7-6.9 MHz (-6 dB). Its lens (RTV silicon) accounts for approximately 20 to 30% of the heat generated in a typical, diagnostic transducer according to empirical estimations made by the manufacturer (Siemens). However, it does provide a thermal resistance from the other areas where heat is generated, which encourages heat ﬂow along alternative paths, such as the foil interconnects, to ambient temperature. Without the lens, thermal resistance (proportional to thermal conductivity * area / path length) of the copper interconnects is higher than, or at best equal to, the shorter path resistance out of the front of the array [121]. This transducer was compared to the VF7-3, a 192 element linear array with a 0.203-mm pitch, an elevation height of 7.5 mm, a lens focus of 3.75 cm, and a bandwidth from 3.0-5.6 MHz (-6 dB). The ARFI and thermal parameters used in this evaluation are shown in Table 3.7.

Test ARFI Thermal ARFI contrast Phantom CIRS Breast Porcine Muscle CIRS Breast Push 3.33, 4.21 3.33 3.33 Frequency (MHz) F/# 0.3-1.0 0.26-0.75 1.5 Focus (cm) 1.0 1.0 1.4 Track PRF (kHz) 13.9 8.0 13.9 Pulse 47.5 3.0 90 Length (µs) Track 5 120-5000 5 Time (ms) Locations 40 1 50 (M-mode)

Table 3.7: Sequence parameters for the ARFI and thermal testing of the Duke-2 (passive cooling) and VF7-3 (control) transducers.

64 ARFI displacements were evaluated in a homogenous region of a 0.7 dB/cm/MHz CIRS phantom. Each ARFI imaging sequence consisted of 47.5µs excitation pulses transmitted at 55% of the maximum system voltage with a 1 cm focus at 40 lateral locations and tracked at 7.27 MHz. Excitation transmit frequencies of 3.33 MHz and

4.21 MHz were evaluated at focal conﬁgurations between F/0.3 and F/1.0, as shown in Table 3.8. As expected, lower F/#s are associated with larger displacements due to the increase in acoustic otuput with increasing numbers of elements. The 3.33 MHz, F/1 conﬁguration from the Duke-2 was compared to that of the VF7-3. The Duke-2 required a much higher system voltage to yield equivalent displacements to those of the VF7-3, as shown in Table 3.9.

Frequency (MHz) F/# Displacement (µm) 3.33 0.3 12.8 1.2 ± 3.33 0.4 10.1 0.9 ± 3.33 0.5 7.9 0.7 ± 3.33 0.6 5.3 0.5 ± 3.33 0.7 4.4 0.4 ± 3.33 0.8 4.0 0.4 ± 3.33 0.9 3.6 0.4 ± 3.33 1.0 3.3 0.3 ± 4.21 0.5 5.1 0.5 ± 4.21 1.0 4.2 0.4 ± Table 3.8: Maximum ARFI displacements associated with the Duke-2 transducer using 3.33 MHz and 4.21 MHz radiation force excitation frequencies at various F/#s. Each excitation pulse was 47.5µs in duration focused at 1 cm using 55% of the maximum system voltage. The precision of these displacements are based on the standard deviation across 40 lines in the ARFI image.

Transducer heating measurements were acquired with the Duke-2 transducer in

two ways. A type-K thermocouple was placed in the backing layer of this array during construction and a 36 gauge type-T thermocouple was also sandwiched in between the lens (center of the active aperture) and porcine muscle with gel using the setup described in section 3.3.1. The temperature rises associated with several

65 System Duke-2 VF7-3 Voltage (%) (µm) (µm) 25 – 1.7 0.5 ± 35 – 3.5 0.9 ± 45 – 6.5 2.1 ± 55 3.3 0.3 16.8 8.0 ± ±

Table 3.9: The maximum displacements associated with the VF7-3 transducer at diﬀerent system voltages were compared to the displacement of the Duke-2 at 55% using a 3.33 MHz, 47.5 µs radiation force excitation focused at 1 cm, F/1. Means and standard deviations are across one image.

M-mode sequences were measured on the Duke-2 and compared with the VF7-3 transducer as a control, as shown in Table 3.10. The system voltage used with the VF7-3 was chosen by applying a linear regression to the maximum displacement versus system voltage squared data and selecting the system voltage (30.7%) yielding

an equivalent displacement to the Duke-2 driven at 50%. In most cases, the face heating of the Duke-2 was lower than that of the VF7-3. Based on these results, two aggressive sequences were performed to evaluate the robustness of the design. The conﬁguration and temperature rises associated with these sequences are shown in Table 3.11. Both sequences resulted in damage to the Duke-2 transducer, which was evident in B-mode images. The beam location was moved for the second location (beam 30) in order to have an active aperture completely independent of the one used for the ﬁrst damaging sequence (beam 110).

Subsequent to these thermal measurements, the acoustic stack was imaged using a ”Golden Spike” test at Siemens with a 7 MHz transducer, as shown in Figure 3.5. Analysis of the image showed delamination in the matching layers to be the source of damage and image degradation. The image quality between the Duke-2 and VF7-3 transducers were compared using B-mode and ARFI image contrast of a stiﬀ, hypoechoic lesion in a CIRS

66 Cycles F/# M-mode Duke-2 Backing Duke-2 Face VF7-3 Face lines Temp. (◦C) Temp. (◦C) Temp. (◦C) 40 0.75 1000 *1.48 – 1.39 10 0.75 4000 0.92 1.78 1.66 10 0.75 24000 *8.28 – 7.31 10 0.75 40000 *13.23 – 11.41 30 0.26 2000 1.08 – – 10 0.75 6000 1.40 2.20 2.24 30 0.26 3000 1.63 – – 10 0.75 8000 1.90 2.42 2.77 30 0.75 4000 3.43 2.88 3.76 30 0.75 5000 4.72 4.02 4.55 10 0.50 1000 0.33 0.44 0.80 10 0.26 1000 0.36 0.40 –

Table 3.10: Backing layer and face heating for various M-mode sequences on the Duke- 2 with a PRF of 8 kHz transmitting 3.33 MHz pulses at 50% of the maximum system voltage. ‘–’ represents data that was not acquired. VF7-3 face heating measurements were made using 30.7% system voltage, which yielded equivalent displacement to 50% voltage on the Duke-2. The face heating measurements were not repeaked between trials. The face heating measurement precision without repeaking is 8%. *These temperatures were recorded after element damage was noted.

Sequence 1 Sequence 2 Beam 110 30 Frequency (MHz) 3.33 3.33 Cycles 30 20 F/# 0.75 1.0 M-mode lines 8000 5280 PRF (kHz) 8 15 System Voltage (%) 55 55 Backing Temperature (◦C) 8.09 0.5*

Table 3.11: M-mode sequences resulting in visible damage to the Duke-2 transducer in B-mode images. The beam location of sequence 2 was spatially oﬀset from that of sequence 1 in order to have a completely independent aperture (no previously damaged elements). *The elements in beam 30 were located further away from the type-K thermocouple in the backing layer than those in beam 110. These experiments were performed only once; the precision of the type-K thermocouple is 2%. ±

67 Figure 3.5: ”Golden Spike” images taken with a 7 MHz transducer of the acoustic stack of the Duke-2. Top - Axial image. Bottom - C-scan image at around 0.6µs in the axial scan. The dark red regions in the C-scan correspond to areas of matching layer delamination. elastography phantom, as shown in Figure 3.6. The B-mode contrast was 0.35 for the Duke-2 and 0.58 for the VF7-3, while the ARFI image contrasts were 0.40 and 0.57, respectively.

The passively cooled Duke-2 transducer had pros and cons. It required higher system voltages to yield similar acoustic outputs at the focus in terms of ARFI displacement. However, the face heating for the same focal output was slightly lower, in general, for this array than the control (VF7-3). Damage was observed for more aggressive sequences only tried on the Duke-2. The Duke-2 image quality was found to be inferior in B-mode and ARFI as compared to the VF7-3. Overall, although this array did demonstrate some reduced heating capabilities, they were

68 1 1 6 6 8 0.8 8 0.8 10 10 0.6 0.6 12 12 14 0.4 14 0.4 Axial (mm) 16 16 0.2 0.2 18 18 0 0 −4 −2 0 2 4 −4 −2 0 2 4 Lateral (mm) (a) 1 1 6 6 8 0.8 8 0.8 10 10 0.6 0.6 12 12 14 0.4 14 0.4 Axial (mm)

16 0.2 16 0.2 18 18 0 0 −4 −2 0 2 4 −4 −2 0 2 4 Lateral (mm) (b)

Figure 3.6: Image quality comparison between the passively cooled Duke-2 (a) transducer and the VF7-3 control (b) using a CIRS elastography phantom containing a stiﬀ lesion. The B-mode contrast was 0.35 for the Duke-2 and 0.58 for the VF7-3, while the ARFI image contrasts were 0.40 and 0.57, respectively. The ARFI images are shown 0.6 ms after the radiation force excitation. not deemed signiﬁcant enough to prevent damage to the array during ablation and, thus, were not pursued further in this work.

3.4.4 Capacitive Micromachined Ultrasound Transducer

Another approach to decreasing transducer heating is to use cMUTs, rather than PZT. We compared face heating and acoustic output of a custom cMUT array designed by Siemens with that of a standard array. The cMUT tested herein was designed for transmitting at a low frequency (2 or

69 2.5 MHz) for deep B-mode penetration. It consisted of 192 elements with a pitch of 0.02 cm and 192 bias lines spaced the same distance apart in elevation (no lens required due to Fresnel focusing). The PH4-1 transducer was used as a PZT control transducer for comparison. The PH4-1 has 128 elements with a pitch of 0.022 cm and an elevation height of 1.35 cm. These arrays were compared in terms of acoustic output near the surface of the array and face heating resulting from the transmission of PW Doppler sequences (see Table 3.12).

Test Thermal Thermal Media NPL Phantom NPL Phantom Frequency (MHz) 2.0 2.0 F/# 1.6 1.6,4.1 Focus (cm) 3.75 3.75 PRF (kHz) 7.9 2.4,28.4 Pulse Length (µs) 5.0 3,12.5 Time (s) 1 60 Mode M-mode PW Doppler

Table 3.12: Sequence parameters for the thermal testing of the cMUT transducer. PW Doppler used apodized pulses, while M-mode did not. The PW sequences were also evaluated on the PH4-1.

Because the PH4-1 has a ﬁxed focus of 7 cm in elevation and a smaller total aperture, the acoustic output was compared using surface pressures. The surface acoustic (instantaneous) power of the cMUT transducer was measured with an ONDA, 0.2- mm spot size, calibrated hydrophone with preampliﬁer, while the PH4-1 transducer measurements were made with a Sonic Technologies, 0.6-mm spot size, calibrated hydrophone. Both were measured 8 mm from the surface of each transducer. The acoustic power of the cMUT was 0.6 W using a transmit frequency of 2 MHz focused at 37.5 mm with 114 elements with +/- 95V DC bias (optimal voltage for increased acoustic output) and 21V AC (22% of the collapse voltage for the DC bias used) driving voltages applied. The peak intensity and peak positive pressure were

70 0.37 W/cm2 and 152 kPa, respectively. A plot of the intensity proﬁle is shown in Figure 3.7, where the Fresnel pattern can be seen in elevation.

−10 0

−5 −5 0 −10 5 Elevation (mm)

10 −15 −20 −10 0 10 20 Lateral (mm) (a) −10 95

−5

0 0

5 Elevation (mm) 10 −95 −20 −10 0 10 20 Lateral (mm) (b)

Figure 3.7: (a) Intensity proﬁle (color scale in dB) of cMUT measured 8 mm from transducer surface transmitting 114 elements at 2 MHz with +/- 95V bias, and 21V AC voltage. The resulting power (integrated Isppa across the transducer face) was 0.6 W with a peak intensity of 0.37 W/cm2. (b) Fresnel pattern associated with cMUT (gray scale in volts).

The face heating associated with short (1.01 s) M-mode sequences (7.9 kHz PRF,

8000 lines, 10 cyc, 2 MHz transmit, similar to those used on the Duke-2) were evaluated with a 36 gauge thermocouple coupled directly to the thermal tissue mimicking phantom material submerged in 2% ethanol. The measured face temperature rises and associated peak positive surface pressures (estimated by cMUT transmit pres-

sure simulations at Siemens) are shown in Table 3.13. It should be noted that the

71 scanner was incapable of supplying enough power for all 8000 M-mode lines when the system voltage was raised to 60% of the cMUT collapse voltage (57 V). The surface pressures measured on the PH4-1 (PZT control) for the same AC voltages are also shown.

cMUT PH4-1 ◦ Vac (V) Psurface (kPa) Total Power (W) ∆T( C) Psurface (kPa) 19 137 0.5 0.03 764 21 152 0.6 – 840 28.5 209 1.1 0.06 1127 38 283 2.1 0.1 1490 47.5 362 3.4 0.14 1853 57 446 5.2 * 2216 90 803 16.8 0.05 3478 (only 1000 lines)

Table 3.13: cMUT driving voltage, surface pressure, total acoustic power, and face heating for a sequence transmitting 8000 M-mode lines consisting of 10-cycle pulses at 2 MHz with a PRF of 7.9 kHz. cMUT transmit pressure simulations were used to estimate the peak positive surface pressure for all but the second row of data. The repeatability of the pressure, power, and temperature measurements for this experimental design was 8%. *This sequence tried to draw more power than the scanner could supply.

In order to see if any signiﬁcant face heating occurred with the cMUTs, the temperature rise associated with several PW Doppler sequences was measured (33 gauge thermocouple coupled to thermal tissue mimicking phantom with gel and positioned at the approximate center of the active aperture), as shown in Table 3.14. The maximum temperature rise observed was 2.36◦C using a 46 element (F/4) lateral aperture focused at 3.75 cm with 84.7V (89% collapse) AC driving voltage and +/- 95V DC bias voltage with 25 cycles per pulse at a PRF of 2.4 kHz. Note that the system voltage was raised above 57 V for a longer time than shown in Table 3.13 but the PRF and number of elements in the aperture were decreased. A linear regression was applied to the surface pressures measured for the cMUT and PH4-1 transducers in order to determine the system voltages at which to drive the PH4-1

72 with equivalent acoustic output. Less heating was observed for the cMUT than the PH4-1 for the sequences evaluated (Table 3.14). cMUT PH4-1 PRF Vac Expected Acoustic ∆ T Vac ∆ T (kHz) (V) Power per Pulse (W) (◦C) (V) (◦C) 28.4 21 0.242 0.73 2.6 1.23 (6 cycles) 2.4 21 0.242 0.26 2.6 0.72 2.4 30 0.508 0.43 4.8 0.91 2.4 47.6 1.372 1.00 9.1 1.63 2.4 67.3 3.559 1.64 13.9 2.95 2.4 84.7 6.027 2.36 18.1 4.55

Table 3.14: Face heating on cMUT using PW Doppler sequences at 2 MHz, 3.75 cm, +/- 95V for 60 seconds. Voltages yielding equivalent surface pressures were used for measurements on the PH4-1 (100V = 100% system voltage). All sequences used 46 element and 25 cycles unless otherwise stated. A repeatability near 8% for the power and temperature measurements was expected.

As a ﬁnal note, the acoustic output of the cMUT at the focus was low. For a 31Vpp AC with +/- 85V bias, the highest intensity measured was for an F/2 conﬁguration focused at 4 cm with a 2 MHz transmit resulting in 22.5 W/cm2.

B-mode image quality was compared between the cMUT and PH4-1 transducers using a CIRS, liver-mimicking elastography phantom containing a stiﬀer concentric lesion (Figure 3.8). The lesion contrast for the cMUT was 0.12 while the contrast for the PH4-1 was 0.28.

The cMUT did not work well. For equivalent system voltages, the cMUT yielded much lower surface pressures than the PH4-1, and the cMUT’s maximum acoustic output was much lower than what has been used for liver ablation in the HIFU literature. However, the cMUT yielded lower face temperatures than the PH4-1. The image quality was degraded for the cMUT compared to the PH4-1. This prototype cMUT design would not be recommended for future therapeutic use; however, it should be noted that this was one of the ﬁrst prototype cMUT transducers, and thus

73 0

Axial (mm) 40

−50 0 50 Lateral (mm) (a) 0

Axial (mm) 40

−50 0 50 Lateral (mm) (b)

Figure 3.8: B-mode image quality comparison between the cMUT and PH4-1 transducers using a CIRS elastography phantom. A very low contrast, 12-mm diameter lesion is centered at 0 mm lateral, 27 mm in depth. The contrast for the cMUT was 0.12 while the contrast for the PH4-1 was 0.28.

fabrication, system impedance matching, and other design issues may be signiﬁcant contributing factors to its poor performance.

3.5 Discussion

Diagnostic arrays designed only for imaging performance demonstrate good bandwidth and sensitivity, but are generally limited in their acoustic power output capability due to internal heating. A lower loss piezoceramic, such as PZT4, has lower coupling (by 10%) and dielectric constants than PZT5-H and, therefore, results in

tradeoﬀs between sensitivity and bandwidth. Multilayer structures are an eﬀective

74 way of compensating for the lower dielectric constant [47]. Piezocomposites can be used to further reduce the acoustic impedance and improve bandwidth. Our data show that a multilayer PZT-4 composite array generated comparable B-mode and ARFI image quality, with considerably lower driving voltage utilized for the

multilayer array. This supports the use of this design. However, composites have several special thermal properties which must be considered. When the working temperature of a piezocomposite reaches the glass-rubber transition temperature (Tg) of resin, transducer failure can occur because of the rapid increase in the polymer’s mechanical losses due to its viscoelastic behavior. For composites cured at high temperatures, the composite thermal avalanche temperature (point where the cooling conditions can no longer compensate for the composite losses) is lower [103]. Furthermore, PZT mechanical losses vary more with piezocomposites with a lower PZT volume fraction [103], making composites with higher volume fractions (67% for the array described herein) more efficient at higher operating temperatures. Nevertheless, PZT5-H losses increase considerably more than PZT-4 as the electric field, temperature, or strain is increased [132]. Evidence of this property was seen in the lower face heating recorded for the PZT-4 composite compared to the PZT5-H control array, as shown in Figure 3.1. Although the multilayer appeared to be superior to the control in both face heating and system voltage requirements, delamination and dead elements were seen in the composite multilayer which was damaged during more aggressive thermal testing. As a result, a different bonding material between layers, passive or active cooling, a cable with a different impedance, or another design modification may be required to sustain repeated ARFI or therapy sequences and take full advantage of this technology. The lower face heating with the PZT-4 triple layer composite transducer does indicate that this could be worth exploring further in the context

75 of therapy with a diagnostic transducer. A PZT5-H array, 8L5 transducer, with a virtually lossless lens (negligible acoustic attenuation) was tested as well and compared with a conventional array. This transducer required a higher system driving voltage to yield equivalent ARFI displacements, and it demonstrated increased face heating over the VF10-5 using drive levels for equivalent focal output. Using ARFI displacement as an indicator of acoustic output at the focus, the need for higher driving voltages to achieve equivalent acoustic outputs may have been due to impedance matching diﬀerences in the two arrays. The 8L5, a probe designed for the Acuson SequoiaTM system, required an adapter to mate its connector to the Antares system. Although this transducer would not be recommended due to its lower output capabilities and higher face heating on the Antares, its superior B-mode image quality to the VF10-5 indicates that this lossless lens material could be used for diagnostic purposes.

The Duke-2 transducer contained an abundance of passive cooling. This cooling was designed with the intent of being able to compensate for any transducer heating that could be generated by power levels achievable on the Antares. On average, the face heating recorded for this transducer was 8% less than that of the VF7-3 for an equivalent acoustic output, as determined by ARFI displacements. However, matching layer delamination was seen after implementation of only two aggressive thermal therapy-like sequences. These results suggest that passive cooling is not sufficient by itself to reduce transducer heating enough to perform ablation. Furthermore, the ARFI displacements for this transducer (not requiring an adapter) were significantly lower than those produced by the VF7-3, which indicates that this particular design requires rework in terms of tuning to the system impedance. Finally, the image quality of the Duke-2 did suffer in terms of B-mode and ARFI image contrast as well as the general clarity of structures (such as the lesion shown in Figure 3.6)

76 resolvable in B-mode. Although some reduced face heating was observed in most cases, further investigation into designs using only passive cooling as a means to performing therapeutic applications on diagnostic transducers is not recommended. If passive cooling were combined with active cooling, therapeutic applications may become possible; however, the degredation in image quality suggests that passive cooling should remain around the perimeter of the array and not interspersed in the backing layer as ﬁns. Negligible internal heating is expected from cMUT transducers due to their low loss tangent. However, the one tested herein showed non-negligible face heating for

PW Doppler sequence, though this face heating was less than that of the PH4-1. This heating could have been due to absorption in the matching layers (there was no lens) between the cMUT elements and the thermocouple; it could also be due to breakdown of the insulating layers from high voltages or heating in the electronic traces contained in the micro-machined silicon wafer. Nevertheless, this transducer had a peak positive surface pressure of only 803 kPa at 95% of the maximum voltage attainable without reaching the collapse voltage (95V for the recommended bias voltage). Comparatively, the PH4-1 (phased, 1D array) transducer has a surface pressure of 3478 kPa at the same system voltage. Wong et al. have reported that a cMUT with a focal gain of 10 requires a peak CW surface pressure of 1 MPa to generate the hundreds of W/cm2 needed at the focal point for ablation [126]. Thus, the instantaneous acoustic output of this transducer was determined to be too low for therapeutic applications, and this ﬁnding was further supported by its low focal intensities. From a system point of view, the average acoustic output necessary for therapy could not be achieved with our system (two DC power supplies for bias lines and Antares for AC voltage) for this transducer. This was shown when a M-mode se-

77 quence with an F/1.6 conﬁguration focused at 3.75 cm drew too much power from the Antares system in less than 1 second with only a 4% duty cycle driven at 60% of the collapse voltage. Overall, the instantaneous acoustic output at the focus for this cMUT would need to be increased by at least 1.25 times, while the power supply

in the diagnostic scanner would need to be able to handle a 25-fold higher duty cycle. Therefore, this cMUT is not predicted to be capable of creating ablation lesions given any driving voltage. However, the power available from our diagnostic scanner was more limiting than the acoustic output achievable by the cMUT, which may be improved with better impedance matching between the transducer and scan-

ner. Taking all of these observations into account, cMUTs are not recommended for therapeutic use on a diagnostic system without design modiﬁcations.

3.6 Conclusions

Transducers for ARFI imaging and therapeutic applications must be able to operate under high driving conditions. A lower loss PZT multilayer composite material, a lossless lens, passive cooling, and a cMUT transducer were evaluated for their potential eﬀectiveness in these regimes. Some of these prototype transducers exhibited lower face heating; however, some exhibited evidence of damage after more aggressive thermal therapy-like sequences, while others did not produce adequate acoustic output at the focus. Clearly, the design of custom diagnostic transducers to reduce the internal heating resulting in transducer damage is challenging and must undergo several iterations to ﬁnd the right balance of design parameters to achieve enough acoustic output at the focus in the appropriate frequency range with reduced face/internal heating. Therefore, the more established designs of standard diagnostic transducers were used to test therapeutic applications in subsequent chapters.

78 Passive cooling would not be expected to cool the array enough for ablation (only 8% improvement as compared to control). However, active cooling, such as ∼ water circulation around the acoustic stack (PZT, backing layer, matching layer, and lens) might reduce internal transducer heating enough to deter damage and allow for the high image quality design of the diagnostic transducer to remain intact. If not, replacing the PZT-5H with multilayer composite PZT-4 and selecting more thermally tolerant matching layers and adhesives may be options from a materials perspective. For 1D arrays, removing the lens (where the majority of acoustic energy is dissipated as heat [110]) might be feasible if the elements are curved in elevation to physically focus the array. Based upon the ﬁndings herein, the most promising design is hypothesized to be an actively-cooled, PZT-4 multilayer composite transducer with passive copper cooling as well.

3.7 Acknowledgments

I would like to thank Stephen Barnes, Michael Zipparo, Timothy Proulx, and

Worth Walters for helpful discussions.

79 Chapter 4

Liver Ablation

In Chapter 2, the CH6-2 transducer was determined to have the highest therapeutic potential for our application due to its superior acoustic output compared to other transducers operating with a similar frequency range at the same focal depth. In this chapter, sequences are optimized and developed for the purpose of creating ablation lesions in liver with this transducer. While it was not possible to generate enough output for the required duration to create an ablation lesion in liver, the additional acoustic output requirements necessary to form a lesion with a diagnostic transducer are discussed. Part of the work presented in this chapter was published in the Proceedings of the 2005 IEEE Ultrasonics Symposium.

4.1 Introduction

Localization and treatment monitoring during HIFU treatment is of key significance for a successful clinical implementation. Exact registration between images of the region requiring therapy and the administration of HIFU is of vast importance to avoid or to minimize skin burns, nerve ﬁber damage, local pain, and other complications [129]. In one case, clinical trials for HIFU treatment of Parkinson’s disease had successful outcomes but were terminated due to the diﬃculty of visual-

80 izing and monitoring the treatment area [10]. Reducing or eliminating damage to surrounding tissue is essential to patient comfort and the wide-spread acceptance of HIFU as an alternative to surgery, RF ablation, etc [127]. As a result, automatically co-registered visualization of HIFU ablation using the same transducer for imaging

as therapy would be immensely useful and reduce costs. Most clinical HIFU trials use MRI temperature monitoring [63, 129], which is very expensive. An ultrasonic visualization method would be better, if comparable monitoring were possible. In this chapter, the possibility of using a diagnostic transducer and system, which has been demonstrated to be capable of visualizing HIFU lesions, to make small ( 10 ≈ mm3) spot ablations in liver is investigated. The National Cancer Institute estimates 21,370 new cases of liver and intrahepatic bile duct cancer in the United States in 2008 with 18,410 of these cases resulting in death [58]. Wu et al. used concentric imaging and therapy transducers for image guidance and ablation of hepatocellular carcinomas, showing the effectiveness, safety, and feasibility of HIFU for this application [129]. Survival rates were shown to be significantly higher for stage II (one tumor having spread to nearby vessels or more than one tumor with none larger than 5 cm) than stage III (more than one tumor larger than 5 cm, involvement of major blood vessels, spreading to nearby organs, broken lining of peritoneal cavity, and/or spreading to nearby lymph nodes) cancers [129]. The goal of the work in this chapter is to evaluate the potential for performing spot ablations in liver, which potentially would be applicable to stage I (one tumor that has not spread to nearby blood vessels) and possibly stage II liver cancers intraoperatively. This system would be most effective as a semi-invasive, intraoperative alternative (compared to alcohol injections or RF ablations) for small, nonresectable hepatocellular tumors and for some types of metastatic liver cancers. Although a lesion was not achieved with this system, the additional output capabilities necessary for ablation were deter-

81 mined. Furthermore, diagnostic transducer design modiﬁcations for therapeutic use are suggested.

4.2 Background

4.2.1 Typical HIFU Exposures

Studies with traditional HIFU transducers report a variety of powers, temperatures, and durations to achieve lesions depending on the accuracy of their monitoring scheme and whether a protein-denatured lesion or over-exposed lesion is desired. Ta- ble 4.1 shows a variety of HIFU exposures used successfully to ablate liver. The in situ intensity * frequency column indicates the measured temporal-average focal intensity (spatial-peak or spatial-average) derated by the propagation distance (α=0.5 dB/cm/MHz) and multiplied by frequency. For the majority of these reports, the exposure time is less than 20 seconds per location. The thermal dose equation

(1.1) predicts the temperatures and associated durations necessary for tissue ablation [30, 93]. For the studies in Table 4.1 that reported their temperatures and durations, the minimum thermal dose in equivalent minutes at 43◦C is on the order of 103, with the majority using values greater than 1010.

4.2.2 Nonlinear Acoustics

As introduced in Chapter 2 for a plane wave, nonlinear losses at higher source pressures can also be predicted for focused beams. By assuming a Gaussian beam, the shock parameter and associated weak shock absorption can be determined. Weak shock absorption is deﬁned for shock parameters greater than one. The shock parameter can be deﬁned as [28]:

82 Acoustic Power, Duration Frequency, In Situ Thermal Ref. Temperature, Focus Intensity * Equivalent Intensity Frequency Dose max of 90◦C, 20 s 1.1 MHz 137 1012 [19] 500 W/cm2 11 cm W-MHz/cm2 3.8-24 W (RFB), 2-20 s 1.7 MHz 114-723 106 1015 [17] 2 2 − 67-425 W/cm 14 cm W-MHz/cm (ISAL in situ), (ISAL), 121-765 W/cm2 206-1301 2 (Isp in situ), W-MHz/cm ◦ 70-100 C for (Isp) 106-266 W/cm2 (ISAL in situ) linear rise to 100◦C, 7 s 1.1 MHz 137 1015 [18] 500 W/cm2 11 cm W-MHz/cm2 144.4-370.9 W, 5-12.3 s – – 103 109 [68] ◦ − 60-80 C 140-260 W, – 0.8-1.6 MHz 1519-3019 – [63] 5-15 kW/cm2 9-15 cm W-MHz/cm2 2 5-20 kW/cm (ISAL) 0.5-3 mm/s 0.8-3.2 MHz 177-3672 – [128] (RFB) (scan) 9-16 cm W-MHz/cm2 34.6-72.2 W, 8-20 s 1.5 MHz 1125-2348 – [104] 750-1565 W/cm2 8-10 cm W-MHz/cm2 (ISAL in situ) 3000 W/cm2 800 ms 4.09 MHz 7666 – [73] 2 (Isppa, water) 9 cm W-MHz/cm (1 cm in liver)

Table 4.1: HIFU exposures used for liver ablation. Intensities are in water unless otherwise stated. (RFB = Reference speciﬁed that the measurement was made with a radiation force balance.)

2πp0fβz0Gf 2 2 σ = ln[Gf + G 1] + ln[R + √1+ R ] (4.1) 3 2 { q f − } ρ0c0 Gf 1 q − where σ is the shock parameter for a focused Gaussian beam, ρ0 is the density of

water, c0 is the speed of sound in water, p0 is the pressure at the ultrasonic source, f is

the ultrasonic frequency, β is the coeﬃcient of nonlinearity (equal to 3.5 for water),

2 z0 is the focal depth, Gf is the focal amplitude gain, and R is (zr/z0) Gf 1 − q − (where zr is the distance from the depth of interest to the focus). Weak shock

83 absorption also contains a purely geometrical term (F ) that, for a Gaussian beam, can be deﬁned as [28]:

2 z0√1+ R 2 2 F = ln[Gf + G 1] + ln[R + √1+ R ] . (4.2) 2 { q f − } Gf 1 q − These two terms can then be combined into an expression for weak shock absorption

associated with σ 1 as [28,36]: ≥

σ α = F −1. (4.3) w 1+ σ

Weak shock absorption refers to the losses occurring because of the generation and absorption of high frequencies during shock development. Typical values for weak shock absorption are between 0.1 and 1.0 Np/cm. These losses describe the absorption of a medium as long as the linear absorption coeﬃcient at the fundamental

frequency is small, such as in water [28]. In the parametric experiments for ablation sequence optimization presented in this chapter, the measured temperature rise is aﬀected by attenuation through the waterpath and absorption at the focus due to the weak shock. In a linear propagation medium for short time periods, the solution

to the the bio-heat transfer equation (1.2) simpliﬁes to:

2α I0exp( 2α z0)∆t T = t − t (4.4) cv where αt is the linear tissue attenuation in Np/cm, I0 is the non-derated intensity at the focus (which is linearly extrapolated from lower driving voltages if not directly

measured), and cv is the volumetric speciﬁc heat. Therefore, when the temperature is measured at the interface between the waterpath and tissue, the presence of weak

shock is associated with increased harmonic content and, thus, greater absorption.

84 This temperature can be considered proportional to the following expression:

N T (α + α (z0))I0exp( 2α (z )∆z) (4.5) ∝ t w − w n Xn=1 where the focal depth is divided into N units that are ∆z in length, each with a diﬀerent depth-dependent αw.

4.3 Methods

Thermal sequences were designed and implemented with a Siemens SonolineTM Antares scanner and a CH6-2 curvilinear transducer (Siemens Medical Solutions

USA, Inc., Ultrasound Division, Issaquah, WA). The scanner was modified to allow user control of the acoustic beam sequences and intensities, access to the raw Radio-Frequency (RF) data, and circumvention of “safety limits” on instantaneous output power. Studies were first conducted in order to evaluate how changing various ultrasound parameters affected heating. With this knowledge and HIFU dosages from the literature, several aggressive ultrasound sequences were designed for liver ablation and tested to determine the capability of the diagnostic system for creating lesions, with thresholds for transducer damage being identified.

4.3.1 Parametric Analysis

Heating sequences were developed and implemented using M-mode pulses focused at 5 cm with a transmit frequency of 4.44 MHz. Total insonification time, power, and duty cycle were manipulated in order to nondestructively evaluate different heating scenarios for this system. The efficiency of these heating regimes were evaluated by comparing the focal heating achieved relative to the resulting face heating (related

85 to the transducer internal heating). The sequences for testing the eﬀects of total insoniﬁcation time used a 2% duty cycle (10 cycles, 7.9 kHz PRF), 55% voltage (3025 V2), and durations ranging from 0.12 to 0.88 s. Thermal constants for typical

2 3 ◦ epoxy (κ=2.26e-7 m /s and cv=1.55e6 J/m / C) and porcine muscle (κ=1.25e-7 2 3 ◦ m /s and cv=4.2e6 J/m / C) [35,93] were used to calculate the analytic solution to the bio-heat transfer equation (1.5) for comparison. Using the 0.12 s sequence, the power was increased ( 100-9604 V2) by using 10 to 98% of the maximum system ∝

voltage. These settings yielded non-derated Isppa values from approximately 685 to 68015 W/cm2, as calculated from linear extrapolation of intensity measurements

from low system voltages. Duty cycle was changed by increasing the pulse repetition frequency (1 to 14 kHz) while maintaining a total insoniﬁcation time of 0.36 s. Transducer face and focal heating temperature measurements were made by placing a type-T (Cu/Cu-Ni), 36-gauge thermocouple (Omega Engineering, Stamford,

CT) along the elevation dimension of the transducer. Face temperature measurements were made by sandwiching the thermocouple with gel beteen the lens and porcine muscle. For focal heating measurements, the transducer was coupled to a stepping motor controlled translation stage (Newport Corporation, Irvine, CA) and

aimed through a water path to the focus, where the thermocouple was aﬁxed on top of a piece of porcine muscle with sound absorbing material underneath. The thermocouple data was recorded using a 16-bit data acquisition system with a sampling frequency of 5 kHz (SuperLogics, Waltham, MA) that was precalibrated with cold junction compensation (CJC) and shielded from the environment with an alu-

minum cover. For focal heating, the thermocouple was manually (prior to automated peaking implementation), thermally peaked in all three spatial dimensions before ac- quiring data in order to ensure the tip was centered within the ultrasound beam [94]. To do this, the transducer was moved in 0.1-mm increments and the temperature

86 rise at each location was measured for a non-aggressive, peaking sequence (1000 M-mode lines at 7.9 kHz PRF each with a 10-cycle, F/1.5, unapodized, 5-cm focus pulse transmitting at 4.44 MHz with 55% of the maximum system voltage). Because the data for each individual parametric plot were acquired after a single peaking pro-

cedure, the additional error introduced from the manual peaking routine over the automated routine described in Chapter 2 did not aﬀect the conclusions made from the parametric analysis. To evaluate the temperature data, a running average of 100 samples (20 ms) was taken. The baseline temperature was calculated as the average of several seconds of acquisition before the onset of the therapy sequence. The maximum temperature rise above baseline for each sequence was recorded. Experiments were repeated 4 times and error bars represent mean standard deviation of the thermocouple readings ± without repositioning.

4.3.2 HIFU Testing

Based on the conduction times along the transducer face and at the focus in tissue (see Figure 4.2) and the short durations used in the literature (see Table 4.1), the developed HIFU sequences were targeted to be 20 seconds. ≤ In the liver experiments, ice water was circulated around the perimeter of the transducer lens through thin plastic tubing. This served to provide a heat sink along the edge of the lens while not disrupting the acoustic ﬁeld or adding extra losses (such as in the case of a waterbag interface). Ice water was also placed around the handle of the transducer to keep the handle electronics cool. With this setup

(see Figure 4.1), extended duration peaking (55% voltage, 10 cycles, 7.9 kHz PRF, 4.44 MHz transmit, 9 s insoniﬁcation), high power (98% voltage, 10 cycles, 7.9 kHz PRF, 4.44 MHz transmit, 9 s insoniﬁcation) and high PRF (14.9 kHz, 50% voltage,

87 Ice Water Bag Heated Water Circulator

Circulating Ice Water

Liver

Thermocouple

Figure 4.1: Picture of setup for HIFU testing including cooling of the transducer (ice water circulated around perimeter of lens through tubing and ice water bag around handle), heated water bath (body temperature) for liver, and thermocouple for monitoring temperature rise (wire seen at front of liver container).

10 cycles, 4.44 MHz transmit, 8 s insonifcation) sequences were evaluated. The focal heating was measured by inserting a 33-gauge, hypodermic needle, type-T thermocouple (Omega Engineering, Inc., Stamford, CT) at the focal depth in bovine liver in a 37◦C waterbath with the transducer directly coupled to the liver. Thermal peaking was automated for these experiments using the stepping motor translation stage (Newport Corpoaration, Irvine, CA) controlled by a custom LabView program (National Instruments, Austin, TX) as described in section 2.3.2.

The thermal peaking sequence used in this setup consisted of 5000 M-mode lines at 7.9 kHz PRF each with a 10-cycle, F/1, unapodized, 2-cm focus pulse transmitting at 4.44 MHz with 55% of the maximum system voltage. The spatial temperature profile was measured using the thermal peaking sequence (low temperature rise so little viscous heating was expected [94]) from the peak to at least 1 mm laterally offset from the peak in order to avoid viscous heating artifacts during the high output sequences [23]. The transducer was then repositioned so that the thermocouple was outside the -3 dB beamwidth (defined as the product of F/# and λ [64]) but still

88 had good SNR. Heating from HIFU sequences was then estimated by running a finite element method (FEM) simulation (see section 4.3.3) of the heating induced by the ultrasound sequence in liver to determine the profile of heat conduction in the lateral direction. The ratio of the simulated peak temperature rise to the rise at the offset position of the thermocouple in the experiment was then used to scale the measured temperature to estimate the actual temperature rise achieved at the focus. To compensate for variations between liver samples and aid in comparisons between HIFU sequences, each temperature rise was then multiplied by the ratio of the maximum temperature rise achieved among the peaking sequences associated with all the liver samples evaluated to the temperature rise associated with the peaking sequence for the liver used for the HIFU sequence evaluation of interest (scale factors ranged from 1 to 2.4). These temperature data were recorded at a sampling rate of 1000 Hz with a Personal DAQ 3000 (Omega Engineering, Inc.) and evaluated in the same manner as in the parametric studies.

4.3.3 Finite Element Thermal Modeling

A thermal finite element method (FEM) model was used to validate the lateral offset procedure by simulating the peaking sequence (5000 M-mode lines at 7.9 kHz PRF each with a 10-cycle, F/1.5, unapodized, 2-cm focus pulse transmitting at 4.44 MHz with 55% of the maximum system voltage) in thermal tissue mimicking material (National Physical Laboratory, Teddington, UK). The thermal profile (temperature versus lateral position) was simulated in LS-DYNA3D (Livermore Soft- ware Technology Corporation, Livermore, CA). First, the intensity distribution in

Field II (as described in Chapter 2) was converted to initial temperature rises above ambient using the solution to the bio-heat transfer equation when the time period under consideration is short enough to neglect conduction eﬀects. This equation is

89 as follows: q T = v ∆t (4.6) cv where T is temperature, qv is the rate of heat production per unit volume, cv is ◦ 3 the volume speciﬁc heat, and t is time. The thermal properties (cv=3.9 MJ/ K/m , K=0.52 W/m/◦K) of the thermal tissue mimicking material were assumed to be independent of temperature and entered as inputs to the model, assuming that the tissue is a thermally homogenous, isotropic solid. The mesh was comprised of

211,680 trilinear dynamic thermal elements with 223,729 nodes spaced 0.3 mm apart laterally, axially, and in elevation. The mesh extended 1.26 cm laterally, 1.26 cm elevation, and 3.76 cm axially. Symmetry conditions were used on the zero-lateral and zero-elevation planes and the top model surface (in contact with the transducer) was held at a fixed temperature (0, ambient temperature). A time-domain, implicit semi-iterative solver was used to simulate the cooling profile for one ultrasound beam and the resulting dynamic temperature fields were processed using LS-PREPOST2 (Livermore Software Technology Corporation, Livermore, CA) and custom-written MATLAB code. (This model has been previously validated in [94].) The cooling profile of one 10-cycle pulse was then summed over 5000 time steps at a repetition frequency of 8 kHz to simulate the peaking sequence and spline interpolation was used to achieve a resolution of 0.1 mm in the lateral dimension. The normalized temperature versus lateral position simulation results were then compared to the corresponding experimental data in the phantom material. The cooling profile of the HIFU sequences were simulated with the same procedure while using the thermal

◦ 3 ◦ properties (cv=3.4 MJ/ K/m , K=0.50 W/m/ K) for liver as inputs and summing over the appropriate duration (corresponding to the length of the HIFU sequence of interest). From these simulations, the peak experimental temperature rise could be

90 estimated from the rise measured at the laterally oﬀset thermocouple position, thus avoiding viscous heating artifacts.

4.3.4 Transducer Evaluation

Focal intensity and acoustic power as well as element impedance measurements were performed according to the methods described in sections 2.3.3 and 2.3.7, respectively.

4.4 Results

4.4.1 Parametric Evaluation

It was hypothesized that delivering heat as fast as possible to the tissue would be most efficient to reduce the effects of conduction. Therefore, the heating of porcine muscle and the transducer face was investigated as a function of duration of excitation, as shown in Figure 4.2. It was found that the temperature rise in the tissue increased linearly over approximately 0.3 seconds and then began to plateau. This leads to the conclusion that rapid (<0.3s) delivery of acoustic energy is the most efficient approach. Furthermore, the face heating was found to increase at a slower rate within this time. Assuming negligible conduction in the interval between pulses (0.13 ms), equation 1.3 and its superposed version over a rectangular source showed good agreement to the experimental data at the focal point in the tissue and the transducer face, respectively. (The parameter r in equation 1.3 was set to the radius of the thermocouple (0.2 mm) for the solution at the focus and to the approximated distance between each element and the thermocouple (0.4 mm for center element) for the solu-

91 tion at the transducer face.) By extrapolating the analytic solutions, the transducer face temperature rise is shown to equal that of the focus shortly after 2 seconds of insoniﬁcation for these sequence parameters. Further extrapolation of the analytic transducer heating solution (not shown) indicates a plateau in the temperature rise at the face much later in time.

2 Thermocouple at Transducer Face Thermocouple at Focus Analytic Solution: 2.26e−07 m2/s 1.5 Analytic Solution: 1.25e−07 m2/s

0.5 Normalized Temperature Rise

0 0 0.5 1 1.5 2 Time (sec)

Figure 4.2: Normalized temperature rise (normalization factor = 7.2◦C) versus total insonification time as measured by the thermocouple at the focal depth and the CH6-2 transducer face (2% duty cycle, 55% system voltage (3025 V2)). Note that the temperatures increase almost linearly for the first 0.3 s at the focus and longer at the transducer face. However, the slopes are not equivalent due to the different thermal properties surrounding the transducer face (slope=6.98 ◦C/s) and focal point (slope=12.96 ◦C/s). Error bars represent the standard deviation of temperature readings over four trials with no repeaking.

System nonlinearities in the scanner power supply and increasing transducer inefficiencies at higher drive levels were hypothesized to impact the heating at higher system voltages such that the theoretical linear relationship between the power delivered to the transducer and temperature rise will overestimate the temperature rise at the focus. Therefore, the effect of increasing power on the temperature rise was investigated using a 2% duty cycle (10 cycles, 7.9 kHz PRF), powers proportional to 100 to 9604 V2 (10 to 98% system voltage), and a 0.12 s insonification time. A plateau at higher powers, as shown in Figure 4.3, was seen at the focus,

92 while the transducer face heated linearly. Because a waterpath to the focus was used in these measurements, nonlinear propagation effects were also considered. Using equation 4.5, the predicted saturation in water for increasing power was calculated, as shown in Figure 4.3. Note that a Gaussian beam profile (0.5-mm radius at the e−1 level, as prescribed in [28]) assumption is used in these calculations, but the actual beam differs slightly from this assumption due to differences in the lateral and elevation focal configuration. A comparison of this predicted curve to the thermocouple data indicated the plateau at higher powers was largely, if not completely, due to saturation in water and not to system nonlinearities or transducer inefficiencies over the short insonification times tested. If the latter phenomena were involved, the focal temperatures would be expected to plateau faster at higher powers, and the linear increase in face temperature with increasing system power would not be expected.

1.6 Transducer Face 1.4 Focus Expected Linear Fit 1.2 Expected Saturation

0.8

0.6

0.4 Normalized Temperature Rise 0.2

0 0 2000 4000 6000 8000 Voltage2

Figure 4.3: Normalized temperature (normalization factor = 2.6◦C) versus power measured by a thermocouple at the focus and at the CH6-2 transducer face (2% duty cycle, 0.12 s insoniﬁcation time). The dashed line is extrapolated from 100 to 1600 V2 (10-40% system voltage) to show the expected trend for temperature with increasing power (slope=1.05e-3 ◦C/V2) in a linear medium compared to that at the face (slope=2.78e-4 ◦C/V2) . The solid line indicates the expected trend from nonlinear saturation, as predicted by equation 4.5. Error bars represent the standard deviation of temperature readings over four trials with no repeaking.

It was hypothesized that a linear relationship between temperature rise and the

93 duty cycle (or PRF, if the pulse duration is held constant) would exist such that a higher PRF would be preferable for heat delivery until system nonlinearities became signiﬁcant or the propagation time to the focus limited the minimum pulse-repetition interval allowable with the Antares system. Therefore, the eﬀect of increasing duty

cycle (0.2 to 3%) on the temperature rise was investigated by varying the PRF from 1 to 14 kHz using a 10 cycle pulse, 55% system voltage (3025 V2), and a total insoniﬁcation time of 0.36 s. As expected, over this range of PRFs, a linear relationship between temperature rise and PRF was found for the transducer and tissue, as shown in Figure 4.4.

1.2 Transducer Face Focus 1

0.8

0.6

0.4

0.2

Normalized Temperature Rise 0

−0.2 0 5 10 15 PRF (kHz)

Figure 4.4: Normalized temperature rise (normalization factor = 11.3◦C) versus pulse repetition frequency (PRF) measured by a thermocouple at the CH6-2 transducer face (slope=0.42 ◦C/kHz) and focus (slope=0.84 ◦C/kHz) for a total duration of 0.36 s using 10- cycle pulses (55% system voltage (3025 V2)). Error bars represent the standard deviation of temperature readings over four trials with no repeaking.

4.4.2 HIFU Studies

In order to validate the FEM procedure for the lateral oﬀset peak prediction method, a phantom mimicking the thermal properties of tissue(α=0.5 dB/cm/MHz,

c=3.9 MJ/K/m3, k=0.52 W/m/K, National Physical Laboratory, Teddington, Eng-

94 land) was compared to FEM simulation results using a sequence with a low temperature rise and, therefore, negligible viscous heating [94]. Figure 4.5 shows the agreement between thermocouple measurements (entire acoustic path is phantom material) and a thermal FEM model of the matched thermal peaking sequence (5000

M-mode lines at 7.9 kHz PRF each with a 10 cycle, F/1.5, unapodized, 2 cm focus pulse transmitting at 4.44 MHz with 55% of the maximum system voltage) at different lateral locations. At the low temperatures of the thermal peaking sequence, the experiment and simulation lateral thermal profiles match well (until the thermocouple noise floor is reached). Therefore, the lateral conduction profiles from FEM simulations of the more aggressive heating sequences (subject to viscous heating artifacts at the peak) were used in conjunction with the laterally offset thermocouple measurements to estimate the heating at the peak without artifact. These simulated conduction profiles are shown in Figure 4.6 along with the ultrasound beamwidth and the laterally offset positions where thermocouple measurements were made. Based on effective ablation procedures presented in the literature (see Table 4.1), several ablation sequences were developed for the CH6-2 transducer and Antares scanner with durations of 9 seconds or less. The sequences were designed to test the following scenarios: a non-destructive peaking sequence for an extended duration, a sequence with a high PRF, and a sequence transmitted near the maximum system voltage. Based on the parametric analysis, a PRF of 7.9 kHz was chosen as the minimum PRF for these sequences in order to take advantage of the diverg- ing temperature rises between the focus and the transducer face at higher PRFs.

The saturation seen while increasing the system voltage in the parametric analysis validated the weak shock theory reasoning for not using a waterpath when trying to create a lesion. Finally, although a total insoniﬁcation near 0.3 seconds would have been optimal from a conduction standpoint, it was previously determined that

FEM Experiment 1

0.8

0.6

0.4

Normalized Temperature Rise 0.2

0 0 0.5 1 Lateral Position (mm)

Figure 4.5: Normalized temperature rise (normalization factor = 1.1◦C simulation, 2.1◦C experiment) as the thermocouple is moved laterally away from the ultrasound beam focus as compared to FEM simulation for 5000 M-mode lines at 7.9 kHz PRF each with a 10 cycle, F/1.5, unapodized, 2 cm focus pulse transmitting at 4.44 MHz with 55% of the maximum system voltage in tissue mimicking material. The match with FEM breaks down at positions under the noise ﬂoor of the thermocouple. The vertical dashed lines indicate the -3 dB ultrasound beamwidth (F/# λ [64]). ·

1 8 sec 9 sec 0.8

0.6

0.4

0.2 Normalized Temperature Rise

0 0 0.5 1 1.5 Lateral Position (mm)

Figure 4.6: Normalized FEM temperature rises verus lateral position from the beam focus for M-mode sequences with 10 cycle, F/1.5, unapodized, 2 cm focus pulses transmitting at 4.44 MHz in liver. One sequence has a PRF of 14.9 kHz and an insoniﬁcation time of 8 seconds (corresponds to high PRF sequence measured at 0.5 mm (circle)), while the other has a PRF of 7.9 kHz lasting for 9 seconds (corresponds to extended peaking and high power sequences measured at 1.2 mm (square)). The vertical dashed line indicates the -3 dB ultrasound beamwidth (F/# λ [64]). ·

96 the transducer could not deliver the required thermal dose for ablation in such a short time span (see Table 4.3), and so a longer time course was chosen. The focal conﬁguration and timing of these sequences along with the experimentally measured acoustic output and temperature at the focus are shown in Table 4.2. None of these sequences resulted in a lesion visible in ARFI or by eye after dissection. The most damage to the transducer occurred during a sequence where high voltage was applied to the transducer. Furthermore, only one sequence (column 1: 10 cyc, 55%, 2 cm, F/1, 7.9 kHz) did not result in any damaged elements, which was the least aggressive sequence.

Extended Peaking High PRF High Power (10 cyc, 55%, 2cm, 10 cyc, 50%, 2cm, 10 cyc, 98%, 2cm, F/1, 7.9 kHz) F/1, 14.9 kHz) F/1, 7.9 kHz) Measured Temp. 1.9◦C @ 1.2 mm 6.0◦C @ 0.5 mm 1.2◦C @ 1.2 mm @ Oﬀset Estimated Temp. 6.8◦C 8.6◦C 4.3◦C @ Peak Sequence Duration 9s 8s 9s Thermal Equivalent 0.33 0.81 0.01 Minutes Simulated Temp. 9.7 13.6 30.8 Rise 2 2 2 Isppa 4785 W/cm 3954 W/cm 15201 W/cm (linearly extrapolated) 2 2 2 Ispta in situ 31 W/cm 48 W/cm 98 W/cm 2 2 2 Ispta in situ * 138 W-MHz/cm 213 W-MHz/cm 435 W-MHz/cm Frequency Power/pulse 275 W 228 W 875 W Average Power 5 W 8 W 16 W 2 2 2 Isata 53 W/cm 83 W/cm 170 W/cm 2 2 2 Isata in situ * 85 W-MHz/cm 133 W-MHz/cm 272 W-MHz/cm Frequency Dead Elements 0 3 47

Table 4.2: Temperature rise, thermal dose, and damage associated with HIFU sequences. All ultrasound pulses are at 4.44 MHz. The intensity and power measurements are found by linear extrapolation from low system voltages. The intensity, power, and simulated temperature rises assumed no dead elements and were overestimates after damage occurred.

97 4.4.3 Transducer Damage

During the testing of aggressive sequences on standard, diagnostic transducers, several sonication regimes were found to cause damage to several elements. Table 4.3 provides a listing of these sequences. It was found that one of these sequences (CH6- 2, 4.44 MHz, 10 cycles, 55% voltage, 2 cm, F/1, 7.9 kHz PRF, 3 minutes) resulted in transducer damage with the setup described for the HIFU testing in this chapter; however, no damage was seen after repeated use of the sequence when the same timing (duration and PRF) and voltage were used with a focus of 5 cm with an F/1.5 conﬁguration and room temperature water surrounding the bottom centimeter of the transducer.

Transducer Frequency Cycles System Focus F/# PRF Duration (MHz) Voltage (%) (cm) (kHz) (s) CH6-2 4.44 40 55 5 1.5 7.9 0.38 *CH6-2 4.44 10 55 2 1.0 7.9 180 CH6-2 4.44 10 50 2 1.0 14.9 8 CH6-2 4.44 10 98 2 1.0 7.9 9 PH4-1 2.67 120 55 3.75 1.5 0.3 16.7 75L40 6.67 200 40 1.5 1.5 10.5 3 75L40 5.71 30 70 1.9 1.5 10.5 1 75L40 5.71 15 70 1.9 1.5 12.5 0.8

Table 4.3: M-mode sequences that are known to cause damage to standard, diagnostic transducers (see Chapter 3 for custom transducer damage) and should not be used without cooling measures. Note: The * sequence can be used if the bottom cm of the transducer is completely surrounded by room temperature (or colder) water.

4.5 Discussion

Several pulsed insoniﬁcation regimes, which are more conducive for implementation on diagnostic transducers and systems than the traditional, continuous-wave HIFU regime, were evaluated in this chapter. The eﬀect of sequence duration on

98 temperature rise was evaluated for the system presented herein to better understand the trade-offs between increased thermal dose using longer insonification times and other effects, such as transducer heating and tissue thermal conductivity. The optimum sequence duration was dictated by differences in the temporal heating profiles

of the transducer and the tissue. These experiments were done through a water path, using porcine muscle. A cooling time constant of 0.26 0.02 s was determined for ± the porcine muscle. As expected, a longer time constant (3.9 1.7 s) for transducer ± face cooling was observed. This observation would hold even if the focal heating was measured with no water path because of the higher thermal diffusivity of the lens material over that of the tissue (refer to equation 1.5). Additionally, the slopes of the linear regions of the temperature versus time curves were 12.96 ◦C/s and 6.98 ◦C/s at the focus and transducer face, respectively. These observations support the use of shorter insonification times to reduce the effects of conduction at the focus and to deposit the focal energy before transducer heating becomes dominant. Shorter insonification times are also desirable from a clinical standpoint in order to keep the total procedure time at a resonable level; most reports of HIFU durations are less than 20 seconds.

However, faster delivery of energy to the focus does not always guarantee a better trade-off between focal heating and transducer heating. Only the heating at the outer surface of the lens was evaluated for the CH6-2 in this chapter. Yet, as indicated by simultaneous backing layer and face heating measurements with the Duke-2 transducer (see Chapter 3), total insonification time can have varying effects at different positions in the acoustic stack, such as less face heating relative to heating in the backing layer as the insonification time increases. This is clearly dependent on the thermal conductivities within the stack, boundary conditions, and conductive paths away from different layers in the acoustic stack for a given array.

99 Furthermore, the external thermal conditions (e.g., a cold water bag) surrounding the acoustic stack can affect this tradeoff, such that longer sequences with a low duty cycle can be transmitted without risk to the transducer while shorter ones could deliver heat to the focus faster but result in transducer damage. If the average power delivered to the transducer (increasing for faster heat delivery to the focus) leads to more internal heating, then the passive cooling performed by the water bag may not conduct the heat away fast enough to prevent damage. The most significant case where faster energy delivery results in more transducer heating than focal heating is when thermal runaway occurs. This occurs when the transducer heats up leading to lower electrical impedance. As a result, increased power delivery to the array in addition to increasing losses in the piezo material occur. These effects lead to even more heating [132]. Clearly, the rate of heat removal from the array must be high enough to avoid this scenario for the sonication regime of interest.

System (transducer, ultrasound scanner, and propagation medium) nonlinearities were investigated by looking at the temperature rise for different system transmit voltages. Neglecting any nonlinear losses, a linear relationship would be expected over the full range of system voltages because power is proportional to intensity, which is also proportional to heating (equations 1.5 and 1.6). Nonlinearities in the transducer and ultrasound scanner were considered negligible because the transducer face heated linearly with power as expected. The plateau at the focus at higher powers, as shown in Figure 4.3, is attributed to saturation in the waterpath, leading to appreciable nearfield loss and a plateau in energy reaching the focus [36]. Saturation effects in tissue are expected to be less severe than in water, due to higher absorption in tissue [22]. Thus, there is utility in using a short to non-existent waterpath for high-power sequences. The fit to the focal temperature data (slope=1.05e-3 ◦C/V2) from 93 to 1605 V2 (10 to 40% system voltage) yielded a different slope than the

100 transducer face data (2.78e-4 ◦C/V2). Therefore, for lower system powers, the focal gain of the transducer aided in a higher temperature rise at the focus than internally. In this linear region, these slopes would be six times less at the focus if the propagation path were entirely porcine muscle instead of water; however, in the nonlinear region, the slope would be eight times more as calculated by equations 4.4 and 4.5. Because the aperture and timing of the sequence remained the same throughout this study, increasing nonlinearities introduced in the ultrasonic propagation path through water to the focus contribute to the decreased ratio between focal heating and face heating at higher system powers. As the source pressure of the transducer increases, the shock parameter increases [36], and it has been shown by Sinilo et al. that HIFU heat deposition increases in the presence of shocks in the waveform in tissue [117]. By accounting for the attenuation in the propagation path and added absorption at the focus due to increasing shock, weak shock absorption theory pre-

dicted the observed focal trend well. Finally, the effect of duty cycle on temperature rise was considered, because most diagnostic regimes are designed with a low duty cycle, pulsed-wave scenario while traditional HIFU ablation uses continuous-wave or a very high duty cycle. As expected, doubling the duty cycle, or PRF, yielded twice the temperature rise at the transducer face and focus. The slopes of the temperature versus PRF data at the face and focus differed, however. Assuming a standard epoxy as the lens material for the CH6-2 transducer used in this study, the thermal diffusivity of the transducer face (κ=2.25e-7 m2/s) was greater than that of porcine muscle (κ=1.25e-7 m2/s) by

1.8 times, which approximates the diﬀerence in slopes observed in Figure 4.4 (focus - 0.8 ◦C/kHz, face - 0.4 ◦C/kHz). Traditional HIFU is generally performed with continuous wave (CW) sonication, or 100% duty cycle. As the PRF, or duty cycle, of the sonication sequence presented herein increases the focus will heat by an increasing

101 factor more than the transducer face, as indicated by the higher temperature versus PRF slope at the focus. This relationship will not hold true if thermal runaway occurs. The aggressive therapy sequences used for HIFU testing resulted in transducer damage and established the transducer as the current limiting factor in using a diagnostic system for ablation. Transducer performance can be affected by sudden and gradual temperature changes as well as humidity. Thermal shock (transient temperature gradients yielding rapid rates of thermal stress application resulting from different coefficients of expansion in a composite material [76]) in the matching layers was considered as a potential problem during aggressive ultrasound sequences. When circulating ice water around the perimeter of the lens, the majority of the acoustic stack was still submerged in the heated waterbath for the liver. As a result, the transducer was effectively preheated but with an outlet for heat dissipation along the edges of the lens; yet, damage to the array still occurred. These results, in addition to the fact that changes in performance during thermal cycling tests of transducers (performed by transducer manufacturers) have not been associated with matching layer delamination [132], indicate that thermal shock is not the cause of transducer damage observed herein. A high absolute temperature was most likely the cause of transducer damage in our case. A possible solution to this problem is internal active cooling. Although the lens, a main source of heat generation [110], can be well cooled from the outside (as seen when a waterbag surrounded the transducer), internal active cooling would need to be integrated into the transducer design to effectively cool the piezo elements because of the thermal impedance of the intervening lens and matching layers. Passive cooling in the form of high thermally conductive paths away from the elements is not believed to be a suitable alternative to active cooling for these purposes due to the damage seen in the Duke-2 transducer

102 in Chapter 3. The necessary improvements to this diagnostic system in order to successfully ablate liver tissue can be considered in a few ways. For an insoniﬁcation time of 8 seconds, the diagnostic system only raised the liver temperature by 8.6◦C, thus its associated thermal dose was 0.81 thermal equivalent minutes (relative to 43◦C) using the high PRF sequence. This value is two orders of magnitude less than the thermal dose (45-60 minutes) referenced by Damianou et al. to achieve 100% necrosis using radiofrequency ablation [29]; it is four orders of magnitude less than the lowest thermal dose found in the literature for HIFU of liver tissue (see Table 4.1). Because of the discrepancy between the the thermal dose and HIFU exposures used in the literature, the thermal dose is believed to greatly underestimate the actual energy required to ablate tissue, even when considering that many HIFU regimes overexpose tissue to ensure lesion visibility with B-mode monitoring.

As a result, another quantity was developed to evaluate the reported rate of energy deposition necessary to form a lesion in a reasonable amount of time (< 20 seconds) in a given tissue: the product of the in situ intensity at depth and the transmit frequency. This product should be near 200 W-MHz/cm2 in liver when using the spatial-average temporal-average intensity (see Table 4.1). This quantity, which will be refered to as the thermal predictive value, was derived from the rate of volumetric heat production for ultrasound (equation 1.6) which is proportional to this value for a given tissue attenuation. This value was exceeded in the high power case; however, extensive damage was seen after 9 seconds. Other sequences with this transducer (e.g., 40 cycles, 55% system voltage, 5 cm, F/1.5, 7.9 kHz, shown in Table 4.3) have also exceeded this value, but have resulted in transducer damage in an even shorter period of time. The thermal predictive value for the extended duration peaking sequence (10 cyc, 55% system voltage, 2 cm, F/1, 7.9

103 kHz PRF, 9 sec), which was the only one not resulting in transducer damage, was 85 W-MHz/cm2. Based on this metric, the transducer would need to deliver 2.4 times the power without thermal damage to the acoustic stack before a lesion would form in the absence of perfusion. To relate this back to thermal dose, the temperature rise would increase to 16.7◦C resulting in a thermal dose of approximately 173 thermal equivalent minutes (relative to 43◦C). The high power sequence should have achieved this thermal predictive value (there is a loss in efficiency that can result from a variation in transducer properties with temperature [13]) but resulted in transducer damage; therefore, we conclude that tissue ablation with this diagnostic system is transducer limited. The system used herein was limited to an average power of 40W with the ability to reach higher transient powers, which should be sufficient for delivering the energy necessary for ablation using the thermal predictive value. However, the total acoustic power reported in the HIFU literature varies from 35 to 371 W. The lower end of this spectrum is feasible with the current system’s power supply and simple modifications (e.g., heat sinking some inductors in the power supply or adding fans for cooling) may help raise this limit.

4.6 Conclusions

In summary, transducer capabilities are the current limiting factor for use of diagnostic transducers for ablation. Several design implementations for decreasing transducer heating to allow for increased drive levels and duty cycles were tested in Chapter 3, with one being promising (PZT-4 composite multilayer) but damaged during testing. Several standard, diagnostic transducers were tested in Chapter 2, and the probe with the highest potential for successfully performing tissue ablation

104 was selected. In this chapter, several aggressive sequences were developed based on parametric studies and empirical observations from other damaged arrays. None of these sequences created a lesion and the additional energy requirements are on the order of 2.4 times those that can be currently produced without transducer damage in a clinically relevant amount of time (10-20 seconds per spot). In order to perform a complete analysis of the prominent methods for reducing transducer heating, active cooling would need to be evaluated. Furthermore, a probe with a lossless lens (or no lens), thermally tolerant matching, backing, and adhesive layers, PZT-4 multilayer composite, and passive or active cooling (whichever is determined to have the best cost-beneﬁt ratio) could be evaluated to provide the best chance for maintaining a non-damaging internal temperature. These studies did suggest that unmodiﬁed, diagnostic transducers could be used therapeutically in hyperthermia applications. One such application, liposomal drug delivery, is investigated in the following chapter.

4.7 Acknowledgments

Thanks to Mike Zipparo for his helpful discussions.

105 Chapter 5

Liposomal Drug Delivery

In the previous chapter, it was shown that a standard, diagnostic tranducer could not create ablation lesions, but temperatures and durations necessary for hyperthermia applications were possible. As a result, we investigated the production of temperatures for the release of thermally sensitive liposomes using a diagnostic transducer in this chapter. The required temperatures and time periods necessary for release were achieved.

5.1 Introduction

Chemotherapy is mainly used as an adjuvant to surgery and radiation for cancer instead of a primary treatment because of the inability of drugs to reach the DNA and RNA of the cancer and associated cells [88]. Therefore, a drug carrier, such as a liposome, which can load and retain the drug, evade the body’s defenses, and so target (passively and specifically) the interstitial tissue of tumors, is necessary to make chemotherapy more effective and less toxic [88]. Several methods for initiating the release of the encapsulated drug have been explored. Lipid-based drug carrier systems can load and retain drugs by passive encapsulation and remote pH loading, evade the body’s defenses with a polyethylene glycol-derivatized (PEG) coating, and target the interstitial tissue of tumors through passive or specific means [96]. How-

106 ever, the drug release from these liposomes is passive and slow [96]. As a result, temperature-sensitive liposomes for site-speciﬁc release of drugs (and magnetic resonance contrast agents) have been investigated by several groups [34,88,107,122,123]. The permeability of these liposomes peaks around the melting point of the lipid shell,

which can be judiciously chosen such that the drugs are released during mild, localized hyperthermia [88]. Traditional thermosensitive liposomes are triggered in the range of 42 to 45◦C to release their drug over approximately 30 minutes, while low tempetuare-sensitive liposomes (LTSLs) release their payload within seconds in temperature ranges of 39 to 40◦C [34,67].

A study was performed at the National Institutes of Health (NIH) to determine if pulsed high-intensity focused ultrasound (HIFU) could eﬀectively activate LTSLs to deliver doxorubicin in tumors [34]. This study showed that LTSLs exposed to pulsed HIFU released doxorubicin more rapidly and at a signiﬁcantly higher concentration

than LTSLs or non-thermosensitive liposomes alone. These 1-MHz pulsed-HIFU

2 exposures had an Isata of 1300 W/cm and consisted of 120 100-ms pulses at a pulse- repetition frequency (PRF) of 1 Hz, resulting in temperature elevations in murine tumors (located in the ﬂank) on the order of 4 to 5◦C [34,43]. These results implied

that if a diagnostic scanner could locally raise the body temperature by at least 4◦C for 2 minutes, then liposomal release could be achieved, and imaging could be performed concurrently. The purpose of the work in this chapter is to demonstrate that a diagnostic transducer and scanner is capable of producing enough heat to trigger liposomal drug release. Speciﬁcally, we would like to achieve an absolute temperature of at least 41.3◦C for 1 minute (longer time durations necessary for higher temperatures). With this thermal dose, temperature-sensitive liposomes (developed at Duke) containing doxorubicin have been shown to release 100% of their contents [96]. The

107 shell of these liposomes (and others made at Duke) contain phopholipids (e.g., di- palmitoyl phosphatidylcholine (DPPC) and distearoyl phosphatidylcholine (DSPC))

◦ to retain encapsulated drugs until reaching a melting point (Tm) of 41 C, polymers (e.g., PEG) to evade the body’s defenses, and lysolipids (e.g., monostearoyl phosphatidylcholine (MSPC)) to increase the amount of doxorubicin release by forming stable pores at Tm [96].

5.2 Background

◦ Doxorubicin-loaded liposomes preferentially release at 41.3 C (Tm), as shown in Figure 5.1. Full release can occur at temperatures of 40◦C and higher, but a longer time period for heating is required [86]. Because maintaining precisely 41.3◦C is very challenging with ultrasonic heating and with any in vivo scenario, a temperature at

or above this Tm is acceptable. While most of the release will occur in the ﬁrst minute at higher temperatures, full release can take upwards of 30 minutes.

5.3 Methods

5.3.1 Intensity Measurements

Focal intensity measurements were performed as described in Section 3.3.2 using a 10 cycle, 4.44 MHz, unapodized, F/1.5 conﬁguration on a CH6-2 transducer and

a calibrated membrane hydrophone (Sonic Technologies (Wyndmoor, PA) 0.6 mm spot size).

108 Figure 5.1: Moles of Doxorubicin released versus time in minutes from DPPC:MSPC(10%):DSPE-PEG(2000) (4%) vesicles at 30, 37, 39, 40, 41.3, 42 and 45◦C. The release of 3.86e−9 moles of drug corresponds to 100% release of contents. All data points represent the mean of three separate experiments. Open circles represent release at temperatures above the transition temperature (41.3◦C). This graph was reproduced from [86].

5.3.2 Ex Vivo Heating

To show the temperature rises necessary to release the liposomes could be achieved, ultrasonic heating was measured in ex vivo chicken breast muscle. The muscle was placed in room temperature deionized water. Because a waterbag around the transducer was necessary in the release assay experiments, a waterbag was placed around

the transducer for the temperature measurements and was used to make contact with the water containing the muscle. A holder for the needle and syringe used to inject the liposomes was attached to the transducer holder such that the needle tip was in the center of the transducer in the lateral and elevation dimensions and at

5 cm in depth. A 33 gauge, hypodermic needle, type-T thermocouple (Omega En- gineering, Inc., Stamford, CT) was ﬁxed to the needle used for injections such that the thermocouple and needle tip were aligned. With the thermocouple and needle in place, the temperature rise from a 4.44 MHz M-mode sequence transmitting 10

109 cycles at a pulse repetition frequency of 7.1 kHz with a F/1.5 focal configuration at 5 cm was measured for different insonification times and system voltages. The data were analyzed with the same method described in section 2.3.4.

Frequency (MHz) Cycles PRF (kHz) F/# Focus (cm) 4.44 10 7.1 1.5 5

Table 5.1: Heating sequence used for ex vivo and in vivo studies.

5.3.3 In Vivo Heating

To show the temperature rises necessary to release the liposomes could be achieved in vivo, ultrasonic heating in the mouse brain was measured. (This speciﬁc organ

was chosen to provide requested data for a neurosurgeon while simultaneously showing the required in vivo temperature rise for our purposes.) Although heating with a diagnostic transducer would not be transferrable to the human brain with the skull intact, this experiment was hypothesized to demonstrate that the necessary heating for release of liposomes could be achieved in a highly perfused, in vivo scenario with high loss (18% pressure loss through the mouse skull). A waterbag coupled with gel to the skull of a male C57BL/6 mouse was used to provide a stand-oﬀ for the transducer. While the animal was under isoﬂurane anesthesia, the in vivo temperature rise from the sequence (see Table 5.1) was measured using a 33 gauge, hypodermic needle, type-T thermocouple (Omega Engineering, Inc.) that had been inserted into the mouse brain with the tip 3 mm below the tissue surface. The timing (> 30 minutes) and voltage (50% and 55% system voltage) of this sequence was similar to the non-damaging HIFU sequence tested in Chapter 4; however, a waterpath was used for this sequence to keep the transducer cool over the long duration used for peaking.

110 A Visualsonics stereotaxic stage (Toronto, Canada) was used to find the location of the maximum temperature rise manually in three dimensions. A temperature versus time cooling profile was recorded at this location. In a second mouse, the procedure was repeated, and the thermocouple was translated with a micro-manipulator to record a temperature versus lateral position profile of the heating. (The insertion of the thermocouple into the brain resulted in some tearing of the brain tissue.)

5.4 Results

5.4.1 Acoustic Output

The focal intensities associated with the heating sequences herein are shown in Table 5.2. The ﬁrst column represents the intensity in water that has been linearly extrapolated from low system voltages. The second column uses equation 4.3 to calculate the nonlinear loss through 4 cm of water and an attenuation of 0.5 dB/cm/MHz to calculate the additional loss through 1 cm chicken breast muscle. Note that the intensity at the focus in the chicken breast muscle varied based on the exact propagation distance through water and muscle. These depth variations were a result of variable muscle thickness between the entry point of the needle and the ﬁnal position of the tip during the ex vivo heating and release experiments. The third column calculates the nonlinear loss through 4.7 cm of water, mouse skull (0.13 Np/MHz over skull thickness [21]), and 3 mm of brain (α=0.6 db/cm/MHz).

111 System Underated Muscle Brain Voltage (%) Intensity (W/cm2) Intensity (W/cm2) Intensity (W/cm2) 20 2811 663 – 25 4405 699 – 30 6353 751 – 35 8655 808 – 50 17688 – 166 55 21407 – 169

Table 5.2: Intensities associated with 4.44 MHz, F/1.5, unapodized, 10 cycle pulses focused at 5 cm. Intensities in each column are using linear extrapolation in water, assuming a 4-cm water path (with nonlinear losses) and 1 cm of chicken breast muscle (α=0.5 dB/cm/MHz), and assuming a 47-mm water path (with nonlinear losses), mouse skull (0.13 Np/MHz over skull thickness [21]), and 3 mm of brain (α=0.6 dB/cm/MHz), respectively. The in situ intensities are only given for the heating scenarios evaluated herein. These measurements were not repeaked, but the standard repeatability is 8%.

5.4.2 Ex Vivo Heating

The temperature rises associated with the sequence at variable durations and voltages are shown in Table 5.3 at a depth of 3 mm in chicken breast muscle. The depth of propagation through chicken was expected to range from 3 mm to 1 cm; thus, the 10.2◦C rise using a 90-second sequence at 35% system voltage could be as low as 7.1◦C (44.1◦C absolute temperature when starting from body temperature)

with a 1-cm tissue path. As seen in Figure 5.2, a 4.3◦C rise (41.3◦C when starting at body temperature, which is considered the optimal temperature for release of these LTSLs) was reached in approximately 19 seconds with a 3-mm muscle path; this increases to approximately 32 seconds with a 1-cm tissue path.

5.4.3 In Vivo Heating

The thermocouple was inserted into the brain of mouse 1 as shown in the x-ray in Figure 5.3a. After the point of peak heating was found to be 42.7◦C (5.3◦C rise

112 Duration System Temperature (s) Voltage (%) Rise (◦C) 5 20 1.2 25 20 2.9 60 20 3.8 60 25 5.4 25 30 6.6 90 30 9.0 90 35 10.2

Table 5.3: Temperature rises associated with M-mode sequences transmitting 4.44 MHz, F/1.5, unapodized, 10 cycle pulses focused at 5 cm with a pulse repetition frequency of 7.1 kHz. These values were measured at a depth of 3 mm in chicken breast muscle through a waterbag interface in the nearﬁeld. The thermocouple was positioned using the needle holder/alignment apparatus at a depth of 3 mm below the muscle surface. Temperature deviations due to system variation are negligible.

12 3 mm 10 mm 10

2 Temperature Rise (deg C) 0

−2 0 20 40 60 80 100 120 Time (s)

Figure 5.2: Temperature rise versus time for the M-mode sequence transmitting 4.44 MHz, F/1.5, unapodized, 10 cycle pulses focused at 5 cm with a pulse repetition frequency of 7.1 kHz at 35% system voltage for 90 seconds through 3 mm and 10 mm of chicken breast muscle. Thermocouple measurements have an error 2%. ≤ above 37.4◦C baseline), the ultrasound was turned oﬀ, and the cooling proﬁle was recorded as shown in Figure 5.3b. The ultrasound sequence was run continuously for 37 minutes in this case. (Note: No transducer damage was noted.) The thermocouple was inserted into the mouse cerebellum of mouse 2 as shown in Figure 5.4a. Temperature versus lateral position were determined as shown in

113 43

40 Temperature (deg C) 39

38 0 50 100 150 200 Time (s) (a) (b)

Figure 5.3: a) X-ray of mouse 1 showing the placement of the thermocouple in the brain. b) Cooling proﬁle in mouse 1 after reaching a peak temperature of 42.7◦C. Baseline temeprature was 37.4◦C.

Figure 5.4b, where the baseline temperature before the initiation of the ultrasound regime was 37.8◦C. The line connecting the points indicates the order in which the

data were taken over the range of -0.65 mm to 0.45 mm. The continuity of this proﬁle and the gradual temperature decay in Figure 5.3 also indicate that negligible viscous heating impacted these results.

5.5 Discussion

The lipid shell of LTSLs (using the Duke formuation) is designed to have the fastest release at 41.3◦C, while higher temperatures still achieve 100% release but over a long time period. As indicated in Figure 5.1, 60% release is expected after 1 minute at or above 41.3◦C, after which the additional release at the higher temperatures slows down considerably. Furthermore, pulsed-HIFU exposures resulting

in temperatures between 41 and 42◦C for a 2-minute time period have shown a signiﬁcant diﬀerence in release between untreated and heated liposomes [34]. This

114 41.5

40.5

40 Temperature (deg C)

39.5

−0.6 −0.4 −0.2 0 0.2 0.4 Lateral Position (cm) (a) (b)

Figure 5.4: a) X-ray of mouse 2 showing the placement of the thermocouple in the brain. b) Lateral proﬁle of the heating in the cerebellum of mouse 2. The line indicates the order in which the data points were taken starting at the circle and ending at the square. Baseline temperature was 37.8◦C.

duration can also be achieved without transducer damaged, as evidenced by insoni- ﬁcation times up to 37 minutes in the in vivo studies. Ex vivo and in vivo heating

measurements demonstrated that these temperatures could be achieved for anywhere from 60 seconds to 37 minutes without damage. Based upon our analysis, ex vivo experiments were designed to demonstrate diagnostic ultrasonic image-guided targeting and heating of LTSLs. However, these

experiments were inconclusive, mainly because of the ex vivo implementation and challenges with freezing and sectioning modifying the liposomes. In order to demonstrate release, in vivo experiments are, therefore, suggested due to complicating factors that have been observed ex vivo (see Appendix A). In a study at the Na- tional Institute of Health, increased release was demonstrated by administering the doxorubicin-loaded liposomes through a tail vein injection in mice and then applying a pulsed-HIFU exposure to a tumor. In order to measure this release, the mouse was euthanized and perfused with saline to clear the vasculature of unreleased lipo-

115 somes, and the tumor was excised. The tumor was weighed, placed in acidic ethanol, homogenized, rotated overnight in the dark at 4◦C, centrifuged, and then placed in a well plate to be read by a ﬂuorimeter. (Doxorubicin emits at 590 nm with a 480 nm excitation.) The ﬂuorescence readings were compared to a standard curve of

serial dilutions of doxorubicin [34]. Another possible assay would be to monitor tumor growth after treatment with either LTSLs or LTSLs plus ultrasonic heating. In vivo experiments have the added beneﬁt of virtually eliminating the possibility that cavitation could be aiding in the liposome release, which is a possibility when injecting a bolus of liposomes into ex vivo tissue.

5.6 Conclusions

Ex vivo and in vivo temperature results demonstrated the use of the CH6-2 transducer in a hyperthermia application capable of producing temperatures for the duration necessary to release chemotherapeutic agents from thermally-activated liposomes without damage to the transducer. Future work should include demonstrating the release of a ﬂuorescent agent from these liposomes in vivo. The following chapter investigates another method of delivering chemotherapeutic agents to a speciﬁc location by using focal ultrasound and contrast agent to open the blood-brain barrier with a mechanical mechanism.

5.7 Acknowledgments

The author would like to thank Gabriel Howles-Banerji for his assistance with the mice. The author would also like to thank Drs. Gerald Grant, Mark Dewhirst, and David Needham for providing the liposomes and motivation for this study and

116 Siemens Medical Solutions USA, Inc. Ultrasound Division for their technical assistance.

117 Chapter 6

Blood-Brain Barrier Disruption

In the previous chapter, we investigated using a diagnostic transducer for hyperthermia with the goal of releasing chemotherapeutic agents from thermally sensitive liposomes. In this chapter, we look at a method of delivering small molecule agents (e.g. chemotherapeutic agents and MR contrast) across the blood-brain barrier using ultrasound and contrast agent with a mechanical mechanism. Several parameters are investigated and a clinically available sequence for opening the blood-brain barrier is determined using a diagnostic system. This work is under review in Ultrasound in Medicine and Biology.

6.1 Introduction

The blood-brain barrier (BBB) is a diﬀusion barrier present in all areas of the brain consisting of endothelial cells, astrocyte end-feet, and pericytes [11]; however, in some brain structures (e.g. subfornical organ, area postrema, organum vasculosum of the lamina terminalis), the BBB is incomplete or leaky [48, 84, 97]. BBB endothelial cells have no fenestrations, more extensive tight junctions than the rest of the body, and sparse pinocytic vesicular transport [11]. This specialized endothelial barrier, therefore, limits the passage of substances (including many therapeutic agents) into the brain [20].

118 Several methods, some involving ultrasound, have been investigated to increase the permeability of the BBB to drugs, antibodies, and gene transfer [66,115,116]. In cell monolayers, mild hyperthermia (41◦C) using ultrasound (20 min, 0.4 W/cm2) has been shown to reversibly enhance passive diffusion of hydrophobic drugs and allow them to bypass efflux transporters [20]. Evidence has also been presented supporting the use of high-intensity focused ultrasound (HIFU) to selectively disrupt the BBB in rats and rabbits; however, this method is often associated with brain tissue damage [81,85]. Low frequency ultrasound (300 kHz) at low intensities has been shown to open the BBB in humans [102]. An alternative method of localized BBB disruption is to use low-pressure ultrasound (e.g. P− (in situ)=0.4-1.5 MPa), at low diagnostic ultrasound frequencies (0.26-2.04 MHz), in conjunction with microbubble contrast agent (such as Optison or Definity) [78, 79]. This has been shown to open the BBB to allow molecules, such as gadolinium for MR contrast, imaging flu- orophores for molecular imaging, and immunotherapeutics for Alzheimer’s disease, to enter the brain of mice and rabbits [21,57,101]. Although ultrasound contrast agents may cause hemolysis and platelet aggrega- tion [59], BBB disruption with ultrasound and contrast agent has been performed with minimal extravasated erythrocytes and no damage to neurons [56, 66]. Stud- ies where opening occurred with a temperature rise of only 0.025◦C suggest that thermal effects are not required in this method of BBB opening [65]. Furthermore, this BBB disruption has occurred without the detection of wide-band emissions, the signature for inertial cavitation; this type of BBB disruption also showed no blood cell extravasation [65, 77]. Consequently, the factors likely to be responsible for such BBB disruption are the oscillation of microbubbles as occurs with stable cavitation, acoustic streaming around the microbubbles, and radiation force on the microbubbles [31,65,77,100,115]. These phenomena may cause mechanical stretch-

119 ing of vessels that leads to opening of tight junctions or the triggering of biochemical reactions to open the BBB [115]. This stretch theory is further supported by work showing that microbubbles can induce a mechanical stretch, activating BKCa channels and leading to rapid hyperpolarization of the cell membrane potential that is in direct contact with the bubbles [119]. To the best of our knowledge, a diagnostic imaging transducer with contrast has not been previously used for the intentional disruption of the BBB. However, Hyny- nen et al. have shown BBB disruption with a 1.5 MHz piston with exposure levels near 6.3 MPa using common diagnostic ultrasound pulse lengths (10 µs) and pulse repetition frequencies (PRF, 1 kHz); yet, these ultrasound exposures sometimes resulted in vascular and neuronal damage [55]. Typical diagnostic transducers operate at 2-10 MHz, however the highest ultrasound frequency currently reported in the literature for BBB opening is 2.04 MHz in rabbits [79]. Diagnostic transcranial color- coded sonography at 3.5 MHz with microbubble contrast did not result in detectable BBB opening when transmitted through the skull in humans [111]. However, if a diagnostic transducer could be used for BBB disruption, it would have the advantage of providing both image guidance of the brain and therapeutic ultrasound delivery

(automatically co-registered) without the need for additional devices. Furthermore, diagnostic scanners are more readily available to biomedical researchers. The primary goal of this work was to investigate the potential use of a diagnostic scanner for locally increasing BBB permeability in animal models. In addition to demonstrating the feasibility of using a diagnostic probe with contrast agent for image-guided BBB disruption in mice, secondary goals were to examine the eﬀect of ultrasonic pressure, pulse duration, frequency, and the delay of ultrasound initiation after microbubble injection as well as contrast agent dose. Furthermore, because Deﬁnity has not been studied in the context of BBB disruption to the extent of Optison [78], this study

120 also serves to add to the knowledge base concerning the only ultrasonic contrast agent currently on the market in the United States.

6.2 Methods

6.2.1 Animal Setup

All animal procedures were approved by the Duke Institutional Animal Care and Use Committee. Twenty-four C57BL/6J mice (22-27 g) were used in this study. Each mouse was anesthetized with isoﬂurane and the scalp depilated. An IV tail catheter

R for perﬂutren lipid microspheres (Deﬁnity , Bristol-Myers Squibb Medical Imaging, N. Billerica, MA, USA) injection and an IP catheter for gadopentetate dimeglumine

R (Magnevist , Bayer Schering Pharma, Berlin, Germany) injection were put in place. A thin plastic bag containing a 17 mm water path was coupled to the scalp with ultrasound gel. A hemicylindrical plastic shell was placed over the thorax of the mouse to prevent the weight of the water from adversely aﬀecting breathing. A Visualsonics stereotaxic positioning system (Vevo Integrated Rail System, Toronto, Canada) was used to center the B-mode image in the transverse plane through the eyes.

6.2.2 Ultrasound Application

A Siemens SonolineTM Antares diagnostic scanner and VF10-5 transducer (Siemens Medical Solutions USA, Inc., Issaquah, WA) were used to insonify the mouse brain approximately 3 mm deep to the dorsal surface of the skull using a transducer focal depth of 2 cm (for both electronic focusing in azimuth and the lens focus in elevation; a water path was used as a standoﬀ to this depth). All acoustic pressure

121 measurements were made with a Sonora SN S4-251 hydrophone with a 0.4-mm spot size membrane (Sonora Medical Systems, Inc., Longmont, CO) and are reported in water (no derating). A baseline sequence with 20 ms, 2.72 0.03 MPa (peak-to-peak) ± pulses repeated at 10 Hz for 30 seconds was implemented based on [21]. Figure 6.1 shows typical pulses used for this study. Modulation of the ultrasonic sequence as well as the dosage and timing of the Deﬁnity injection was performed. Ultrasonic parameters were investigated by varying pulse durations between 0.35 µs and 20 ms, peak-to-peak pressures between 1.05 0.06 and 6.16 0.02 MPa, and frequencies ± ± between 5 and 8 MHz. Deﬁnity doses between 10 and 60 µL (400-2400 µL/kg) and

Definity injection times were also examined, ranging from start of insonification to 2 minutes prior to insonification. Table 6.1 summarizes the parameters investigated and the number of insonified spots investigated for each.

6.2.3 BBB Opening Procedure

For opening the BBB, two different ultrasound sequences (as shown in Table 6.1) were tested on each animal - one on each side of the brain - to reduce the number of animals sacrificed for these experiments. For each sequence, three different locations were serially insonified 2 mm apart in the rostral-caudal direction (see Figure 6.2). Using B-mode, ultrasound image guidance and the stereotaxic positioning system, the transducer was moved to the first location: 3 mm posterior to the eyes and 1.5 mm to the left of the midline as shown in Figure 6.2. At the onset of a 30 second ultrasound sequence, Magnevist (6.3 mmol/kg IP) and Definity (30 µL IV) were injected. (We have found that this dose of Magnevist produces a consistent level of enhancement in mice.) After the 30-second sequence was finished, the transducer was then translated such that two more focal spots were insonified 2 and 4 mm posterior to the first spot at 1 and 2 minutes after the Definity injection (only one

122 Figure 6.1: Example waveforms (a,c) and power spectra (b,d) of pulses with peak-to- peak pressures of 2.72 MPa (a,b) and 6.16 MPa (c,d). At these pressures, the waveforms demonstrate some nonlinearity. The corresponding MI (P−.3/√f) are 0.33 and 0.65, respectively, assuming propagation through 2 cm of tissue. injection per side), respectively. Prior to administering the second sequence (right side of brain), an IV saline flush was given, and the Definity was allowed to clear over 5 minutes. The half-life for Definity in blood is reported to be only a 1.3 minutes [5,120], which is consistent with qualitative observations in this work. The same procedure was then repeated with a different sequence 1.5 mm to the right of the midline but without reinjection of Magnevist, which clears slowly with a half-life of 1.6 hours [6].

Because Magnevist is normally excluded by the BBB, our assay for BBB disruption was to monitor the signal enhancement in MR images. After insoniﬁcation of all six locations, the animal was placed in a quadrature, 300.5 MHz birdcage coil (M2M

123 Number of Deﬁnity Delay after Frequency Pressure Pulse Sonications Dosage (µL) Deﬁnity (MHz) (MPa) Duration (ms) injection (s) 2 10 0 5.7 6.16 20 2 30 0 5.7 6.16 20 2 60 0 5.7 6.16 20 2 30 7 5.7 6.16 20 2 30 60 5.7 6.16 20 1 30 120 5.7 6.16 20 2 30 0 5.0 2.27 20 4 30 0 5.7 2.72 20 2 30 0 6.7 2.75 20 2 30 0 8.0 2.26 20 1 30 0 5.7 1.05 20 1 30 0 5.7 1.60 20 1 30 0 5.7 3.80 20 1 30 0 5.7 2.72 3.5e-4 1 30 0 5.7 2.72 2.0e-3 2 30 0 5.7 2.72 7.0e-2 *1 30 0 5.7 2.72 7.0e-3

Table 6.1: This table summarizes the exposure parameters investigated in this study along with the number of insonifications evaluated for each set of parameters. Each location was insonified for 30 seconds with a PRF of 10 Hz and an unapodized, F/1.5 configuration except the PW Doppler sequence (*) which used an 100 Hz PRF and an apodized, F/4 configuration.

Imaging, Cleveland, OH, USA) tunable for mice (20 - 30 grams) and imaged in a 7T

MRI system interfaced to a GE EXCITE console. A 3D spoiled gradient recalled echo (SPGR) sequence was used to acquire T1-weighted images approximately 30 minutes after insoniﬁcation of the ﬁrst spot. Because Magnevist is normally excluded by the BBB, regions of brain enhancement in the T1-weighted images were interpreted as regions of BBB disruption.

124 (a) (b)

Figure 6.2: (a) Anatomical sketch of a coronal slice of the brain with the insoniﬁcation spots. Only the two most rostral spot positions were analyzed in the MR images. (b) Setup and transducer orientation relative to the mouse. Note: The water bag is not shown here.

6.2.4 Image Analysis

Image registration between the ultrasound and MR images was performed by aligning a control point deﬁned at the top of the skull directly above the center of the BBB opening from left to right in the MR images and along the beam index used for BBB opening in the ultrasound image. The hyper-intense structures in the ultrasound image, corresponding to bones in this study, were then overlaid onto the corresponding MR images to evaluate the eﬀectiveness of the ultrasonic guidance.

The degree of opening was evaluated by semi-automatic segmentation of the volumes of enhanced brain tissue in the MR images. By inverting the gray-scale values in the MR image, applying a 3-D watershed algorithm (The MathWorks, Inc., Natick, MA), and ignoring any voxels originally below the background level, the contrast-enhanced volumes associated with BBB opening were segmented. The

125 full-width half-maximum (FWHM) contours in each slice of a volume were then used to calculate the mean gray-level as well as the dimensions and total volume for each opened region, or spot, in the brain. If the contralateral region of the brain to the region of interest was not insonified, the unopened BBB level was calculated as the mean gray-level in the opposite hemisphere; otherwise, the mean level in an unopened region of equivalent size and shape a few millimeters lateral and caudal to the opened region was used. The contrast-to-noise ratio (CNR) was calculated as the difference in mean gray-levels of the FWHM-defined volumes for the opened and unopened BBB regions divided by the standard deviation in an empty region of the

MR image (no tissue present); therefore, a higher CNR is indicative of more BBB opening (or Magnevist in the brain tissue) and a CNR of 0 indicates no discernible opening.

6.2.5 Histology

The brains of selected mice (one insonified with the greatest pressure and one with the greatest duty cycle) were processed for histology. The tissue of these mice was fixed using transcardiac formalin (10%) perfusion. Coronal sections of the excised brains were taken at 0.5 mm intervals (over 40 slices per animal) and stained with hematoxylin-eosin. These sections were examined for evidence of red blood cell extravasation into the brain parenchyma, which has been reported to be the first sign of tissue damage [15,56].

6.3 Results

Confounding eﬀects from the ventricles and other anatomical variations are among the challenges with these experiments. The blood-choroid barrier of the ventricles

126 opened more readily than the blood-brain barrier. As shown in Figure 6.3, only one of the three insoniﬁcation spots is visible in brain tissue, but the ventricles are clearly visible. Furthermore, because the ventricles are interconnected, an opening of the blood-choroid barrier in one part of the brain caused enhancement througout the

ventricular network. As a result, quantitative measurements are only reported for the most rostral spot (spot 1) for each 3 spot parameter set, because the ventricles were not present within this region (Figure 6.2).

Figure 6.3: BBB opening with PW Doppler. 5.7-MHz, 7-µs ultrasound pulses repeated at 100 Hz with an apodized F/4 conﬁguration yielding 2.72 MPapp were transmitted for 30 seconds immediately after a 30-µL Deﬁnity injection. The white arrow points to an opened spot in the brain. The black arrow points to the ventricle.

The stereotaxic stage in conjunction with ultrasound image guidance prior to

injection of Definity made repeatable localization within the brain efficient. The midline between the eyes was easily visible in B-mode images and the stereotaxic positioning system could be moved such that the BBB opening insonification accurately occurred 3 mm caudal and 1.5 mm lateral to this point. As demonstrated in Figure 6.4, the maximum contrast in the BBB opening was accurate in the medial-

lateral and ventral-dorsal axes.

127 (a) (b) (c)

Figure 6.4: Images showing a) B-mode ultrasound only (5.7 MHz), b) MR only, and c) structures seen in ultrasound (found by thresholding) overlaid in red on the MR image. The yellow + shows the intended center of the ultrasound focus based on the B-mode image. The white region surrounding the + on the right side of the MR image is indicative of T1 enhancement from Magnevist crossing the BBB. BBB opening was achieved using 5.7-MHz, 20-ms ultrasound pulses repeated at 10 Hz with an F/1.5 configuration, yielding pressures of 6.16 MPapp, in a 30-second insonification immediately after a 30-µL Definity injection.

The impact of varying the amount of Definity present during insonification was evaluated in two ways: (1) increasing the dose and (2) changing the time in circulation before insonification. An insignificant (p > 0.05) change in CNR was seen with an increasing dose of Definity from 10 to 60 µL, as seen in Figure 6.5.

30 CNR

0 0 10 20 30 40 50 60 Definity Dosage (µ L)

Figure 6.5: BBB opening as a function of Definity dose. 5.7-MHz, 20-ms ultrasound pulses repeated at 10 Hz with an F/1.5 configuration, yielding pressures of 6.16 MPapp, were transmitted for 30 seconds immediately after Definity injection. Each * represents one animal and the dashed line connects the mean at each dose.

The impact of varying delays between Deﬁnity injection and ultrasound initiation

128 were studied over a range of times. For some experimental configurations, it may be hard to have concurrent injection and insonification initiation; therefore, a fast, but reasonable, range of delays between 0 and 2 minutes were evaluated. No significant (p > .05) change in CNR was noted over these delays as shown in Figure 6.6.

30 CNR

0 0 20 40 60 80 100 120 Time of ultrasound initiation after Definity injection (s)

Figure 6.6: Effect of delay between Definity injection and start of insonification on BBB opening from 0 to 120 s. 5.7-MHz, 20-ms ultrasound pulses repeated at 10 Hz with an F/1.5 configuration, yielding pressures of 6.16 MPapp, were transmitted for 30 seconds after a 30-µL Definity injection. Each * represents one animal and the dashed line connects the mean at each delay. CNR differences were insignificant (p>0.05) for the delays studied.

Previous BBB opening studies have looked at frequencies below 2.04 MHz, but

the bandwidths of diagnostic transducers are usually centered around higher frequencies. Therefore, frequencies of 5, 5.7, 6.7 and 8 MHz were tested. Among these frequencies, 5.7 MHz was found to yield the highest CNR (p < 0.05) in the MR images when applying equivalent input voltages to the transducer (note the diﬀerences

in pressure), as shown in Figure 6.7. Among these frequencies, BBB disruption was generated at 6.7 MHz and below. BBB opening was not observed at 8 MHz. The acoustic output for the frequencies tested are also shown. The values measured in water and derated by the attenuation of the skull and intervening brain tissue (attenuation values reported in [21,35]), as well the MI (peak negative pressure derated

by 0.3 dB/cm/MHz over the square root of frequency) [4] and estimated MIin situ (peak negative in situ pressure over the square root of frequency) values [79] are

129 reported. These diﬀerences in acoustic output are a function of the bandwidth of the transducer.

Frequency 5.0 5.7 6.7 8.0 (MHz) Ppp (water) 2.27 2.72 2.75 2.26 (MPa) Ppp (in situ) 0.99 1.21 1.08 0.69 (MPa) P− (water) 1.02 1.18 1.15 0.94 (MPa) P− (in situ) 0.44 0.52 0.45 0.29 (MPa) MI 0.32 0.33 0.28 0.19 MIinsitu 0.20 0.22 0.17 0.10

Figure 6.7: BBB opening for ultrasonic transmission frequencies from 5 to 8 MHz for the same system input voltage. 20-ms ultrasound pulses repeated at 10 Hz with an F/1.5 focal conﬁguration were transmitted for 30 seconds immediately after a 30-µL Deﬁnity injection. At least two animals were tested per frequency. Non-derated and derated

pressures as well as MI (P−.3/√f) and MIin situ (P−in situ /√f, [79]) for each frequency are listed. The standard deviation of these pressure measurements are 1%. ≤

Regardless of the mechanism, most acoustic bioeﬀects are related to the en-

ergy delivered to the region of interest and duration of insonification. Therefore, we evaluated the effects of changing pressure and pulse duration on the degree of BBB opening. While maintaining a constant frequency (5.7 MHz) and changing the pressure, visible opening was shown to require a peak-to-peak pressure exceeding a threshold between 1.05 0.01 MPa and 1.60 0.01 MPa, as shown in Figure 6.8. ± ± Above 2.72 0.01 MPa, the increase in contrast was insignificant (p > .05). ± In order to show the feasibility of BBB opening with a clinical scanner, a range

130 50

30 CNR

0 0 2 4 6 8 P (MPa) pp

Figure 6.8: Effect of ultrasonic pressures from 1.05 to 6.16 MPapp (non-derated) on BBB opening. 5.7-MHz, 20-ms ultrasound pulses repeated at 10 Hz with an F/1.5 configuration were transmitted for 30 seconds immediately after a 30-µL Definity injection. Each * represents one animal and the dashed line connects the mean at each pressure. of pulse durations corresponding to B-mode, Doppler, and acoustic radiation force impulse (ARFI) imaging were evaluated (see Table 6.1) and compared with a 20- ms pulse previously shown to open the BBB [21], all at a pulse repetition frequency

(PRF) of 10 Hz and a total insoniﬁcation time of 30 seconds. A similar threshold-like behavior to that of the pressure amplitude was noted for increasing pulse duration and total number of cycles, as shown in Figure 6.9 (semi-log plots in x). For this conﬁguration, the threshold for uniform, well visible (CNR > 10) opening is a pulse duration between 2 and 70 µs, repeated such that the total number of cycles exceeds

105. ∼ In the preliminary work for this study, standard B-mode insonification (MI=1.5, as defined by [2]) with Definity present was found to result in some blood cell extravasation, as shown in Figure 6.10. Therefore, for all other data, B-mode images were only acquired prior to Definity injection. The histologic data from the most aggressive, experimental ultrasound regime (MI=0.65, 5.7 MHz transmit frequency, 6.17-MPa peak-to-peak pressure (in water), F/1.5, 20 ms pulse duration, 3.42e7 total cycles, and a 10-Hz PRF) did not show evidence of blood cell extravasation or

131 45 45

40 40 35 35

30 30 25 25 CNR CNR 20 20 15 15

10 10 5 5

0 0 −2 0 2 2 4 6 8 10 10 10 10 10 10 10 Pulse Duration (ms) Total Number of Cycles (a) (b)

Figure 6.9: a) Effect of pulse durations of 0.35 µs (B-mode), 2 µs (Color Doppler), 70 µs (Acoustic Radiation Force Impulse Imaging), and 20 ms on BBB opening. 5.7-MHz ultrasound pulses repeated at 10 Hz with an F/1.5 configuration yielding 2.72 MPapp were transmitted for 30 seconds immediately after a 30-µL Definity injection. Each * represents one animal and the dashed line connects the mean at each pulse duration. b) The same data as in a) presented as a function of the total number of cycles in the insonification sequence. Note these are semi-log plots in x.

(a) (b)

Figure 6.10: H & E stained histology of a) blood cell extravasation caused by standard B-mode (MI=1.5, 0.35 µs, 5.7 MHz, 34.60 MPapp insonifying for five 30 second periods at a 36 Hz frame rate with 30-µL Definity) and b) no damage with the most aggressive experimental ultrasound exposure used for this study (MI=0.65, 5.7-MHz transmit frequency, 6.17-MPapp pressure (in water), F/1.5, and 20-ms pulse duration with 30-µL Definity.) neuronal damage. Based on the range of pulse durations and pressures that resulted in obvious opening (CNR>10) presented here, it became evident that a pulsed Doppler sequence

132 could be utilized for BBB opening in the mouse. As a proof of concept, the VF10-5 transducer was placed in the standard, clinical, pulsed wave (PW) Doppler mode on the Antares system with a frequency of 5.7 MHz, gate size of 15 mm, PRF of 100 Hz, and a system power of 15%, as indicated on the scanner monitor, for 30

seconds (see Figure 6.11. These settings resulted in a pulse duration of 7 µs and 1.2x105 total cycles with an apodized F/4 focal configuration. The MI and peak-to- peak pressure of this configuration were equal to one of the standard configurations tested in this study (MI=0.33, 2.72 0.01 MPa ). As shown in Figure 6.3, this ± pp sequence easily opened the BBB (CNR = 34), while histology showed no evidence

of blood extravasation or neuronal damage.

Figure 6.11: Example of image guidance and system settings for PW Doppler mode BBB opening.

6.4 Discussion

Visualization of the skull, zygomatic arches, and eyes in B-mode images made 3-D localization with the stereotaxic positioning system simple, fast, and repeatable. As evidenced by the registration of B-mode to MR images (Figure 6.4), the location of

133 peak opening was close to the focal point shown on B-mode. The center of the visible opening was not centered around this focus in the ventral-dorsal direction because the focus of the ultrasound beam was closer to the top of the skull. Furthermore, our studies indicated a change in the depth of the opening (center and dorsal-ventral

extent) with anatomical position in the brain (rostral-caudal and left-right). Variable thickness in the skull and confounding effects from the ventricles (where the blood- choroid barrier is easier to open) may explain the variations with position. With a higher attenuation and speed of sound than tissue, variable thicknesses in the skull lead to changes in the pressure delivered in vivo due to increased attenuation and phase aberration [118]. This study suggests that doses of Definity exceeding the manufacturer’s clinical recommendations (10 µL/kg [5]) can be given without histologic signs of damage to the mouse brain (although more detailed histology is necessary for confirmation).

Because it is diﬃcult to administer the clinical doses for the small body weight of a mouse, the doses used in this study were in the range of 400 to 2400 µL/kg (bolus injection). BBB opening was achieved at all studied doses with CNR independent of dose. These results are consistent with those of another group using focused,

ultrasound (0.69 MHz in rabbits) with Optison at lower doses (50-250 µL/kg) [80]. Previous work in mice demonstrated the need for increased pressure (near 3-fold) to observe BBB opening when there was a 15-minute versus a 1-minute delay between contrast agent (Optison) injection and insonification (intact skull, 1.5 MHz, 20-ms pulses at 10 Hz for 30 seconds, 400 µL/kg of Optison) [21]. Figure 6.6 shows minimal variation in opening for up to a 2 minute delay between injection and the start of insonification. However, these data do suggest (without statistical significance) that starting the ultrasound insonification at exactly the same time as Definity injection may be optimal. To ensure that the most Definity possible is insonified before it is

134 cleared or degraded by the system, it may be optimal to initiate insonification prior to injection. A midrange subset of typical diagnostic frequencies were evaluated in this study to show the potential for using diagnostic scanners for BBB opening. A couple of factors, the in situ pressure and the resonance frequency of Definity, could influence the BBB opening observed at a given frequency for a constant pulse duration and insonification time. Of these two factors, the in situ pressure was directly evaluated and had an interesting impact on the BBB opening observed. At 5.7 MHz, the frequency with the highest observed degree of BBB opening, there was a significant

(p < .05) change in CNR between 1.05 and 2.72 MPapp and an insigniﬁcant change

between 2.72 and 6.16 MPapp. By assuming linear attenuation and accounting for acoustic loss through the skull (as reported by [21] at 1.5 MHz) and brain [35], the in situ pressures shown in Figure 6.7 result. These pressures are indicative of the trade-

off between the efficiency, or acoustic output, of the transducer at each frequency and the estimated attenuation in the acoustic path. Distortions of the beam due to phase aberration effects have been shown to increase with frequency [91] and, therefore, may have further reduced the actual in situ pressure due to defocusing

of the beam. Nevertheless, the average CNR between the mice evaluated at each of these frequencies follows a similar trend to the estimated in situ peak-to-peak pressure. The second factor to consider is the resonance frequency of the Deﬁnity microbubbles. The mean bubble diameter of Deﬁnity, as described by the manufacturer, is

between 1.1 and 3.3 µm [5]. According to Goertz et al., a lipid encapsulated bubble of those dimensions will have resonance frequencies ranging from about 13 to 3 MHz, respectively. A 2.2 µm diameter bubble (median of Deﬁnity diameters) with minimal damping should resonate around 4 to 5 MHz according to simulations [46]; however,

135 as bubbles travel through the vasculature, this frequency decreases in vessels of smaller radii (e.g. capillaries) and further decreases near the center (lengthwise) of these smaller vessels [108, 109]. Therefore, the bubbles themselves may bias the degree of BBB opening toward the lower end of the transducer bandwidth used in

this study. The impact of frequency on bubble dynamics is also included in the mechanical index, which indicates lower frequency insoniﬁcations result in an increased likelihood for cavitation [8]. McDannold et al. reported recently that the threshold for

BBB disruption is constant with a variant of mechanical index (MIin situ), which they defined as the peak negative pressure in situ normalized by the square root of frequency [79]. In our data, an MIin situ of 0.10 at 8 MHz did not open the BBB, while an MIin situ of 0.17 at 6.7 MHz did. This is lower than McDannold’s threshold of 0.46 in rabbits with 50 µL/kg of Optison injected 10 seconds prior to insonification, though those experiments were performed post-craniotomy to eliminate skull aber- rations [79]. The order-of-magnitude higher concentration and/or type of ultrasonic contrast agent (Definity instead of Optison) may explain the lower MIin situ threshold required for BBB disruption reported here. Other possible differences might include exacerbated acoustic losses and phase aberration at 8 MHz, as well as different sonication conditions and animal models from McDannold’s work. In addition, the total number of cycles may also play a role in this threshold difference (3.42e7 for this study and 4.08e5 for McDannold’s study), as described below. Possible mechanisms for BBB opening can be hypothesized based on the pulse duration studies. As with ultrasonic pressure, there appears to be a threshold for pulse duration (at a given pulse repetition frequency (PRF) of 10 Hz and total insonification time of 30 sec) that must be exceeded in order to observe appreciable BBB opening (CNR>10, Figure 6.9) for the low pressures used in the majority of

136 this study ( 6.16 MPa in water, MI<0.65) which do not lead to tissue damage. ≤ pp At these low pressures, a pulse length of 2 µs, which is typical for diagnostic Color Doppler pulses, resulted in some opening with a very low CNR (2.4). However, when B-mode (0.35-µs) pulses with a high MI (1.5) were used, easily visible opening was seen but it was associated with some blood cell extravasation. Therefore, longer pulses with lower pressures were found to be preferred for BBB opening without blood cell extravasation. The fact that low pressures are effective is consistent with the hypothesis that inertial cavitation is not necessary for BBB opening [42,77]. These pulse duration studies also suggest that acoustic radiation force may be involved in the mechanism for BBB opening. Primary acoustic radiation force is proportional to acoustic temporal-average intensity [32]. In this study, a significant (p < 0.05) increase in visible opening was observed between 2 µs (Ispta=1.1 2 2 mW/cm ) and 70 µs (Ispta=4.0 mW/cm ), low pressure (2.72 MPapp) pulses at the same PRF and total insonification time, supporting the hypothesis that increased primary radiation force results in more BBB opening [100]. Although not monitored herein, these increased pulse durations would also be providing a longer time period for driving stable cavitation (i.e. bubble resonance without violent rupture) to open the BBB, as described in the literature for thrombolysis [31]. Given the data presented here, the total number of cycles deposited at the focus for a given frequency may be a good indicator of the degree of BBB opening (CNR) observed. Two sequences with different pulse lengths and repetition frequencies, but the same total number of cycles, had similar CNRs, whereas the same pulse length but fewer cycles did not. Specifically, the PW Doppler sequence with 7-µs

2 pulses at a 100 Hz PRF and a total number of cycles of 1.2e5 (Ispta=4.0 mW/cm ) had a CNR of 34. Based upon the Color Doppler (2-µs pulses, 3.4e3 total cycles, I =1.1 mW/cm2, CNR 2) and ARFI pulse lengths (70-µs pulses, 1.2e5 total cy- spta ≈

137 cles, I =4.0 mW/cm2, CNR 30) investigated using a PRF of 10 Hz (Figure 6.9), spta ≈ the PW Doppler sequence was expected to have a CNR of approximately 4. How- ever, the diﬀerence in CNR between two sequences with the same total number of cycles and pressure at the focus, the PW Doppler sequence and acoustic radiation

force pulses at 10 Hz, was insigniﬁcant (p>0.05). Conversely, previous studies by McDannold et al. showed no signiﬁcant change in MR signal intensity while increasing the PRF from 0.5 to 5 Hz which resulted in an increase in the total number of cycles between 6.9e4 and 6.9e5 [80]. These comparisons warrant further investigation into the relationship between the total number of cycles, pulse duration, and

PRF in order to fully understand their impact on the degree of BBB opening. Transcranial opening of the BBB with a diagnostic system was proven feasible in mice; however, signiﬁcant barriers exist for extending this to humans. The mouse skull is very thin resulting in minimal acoustic loss due to attenuation and phase

aberration (18% [21]). Increased attenuation and defocusing due to thicker skulls in larger animals would require more acoustic output from the transducer to achieve the necessary in situ pressures. Although diagnostic systems may be capable of the necessary output, it may lead to excessive skull heating unless aberration correction or other techniques are employed [9, 24]. However, there could be intra-operative situations in which a craniotomy has already been performed and similar methods to those presented herein could be applied directly to the brain, but with the use of a more clinically relevant Deﬁnity dosage. A post-operative situation in which an acoustically transparent window has been implanted might also be feasible.

138 6.5 Conclusion

The results of this study demonstrate the feasibility of BBB opening in the mouse with a commercial, diagnostic system and ultrasound contrast agent. Ultrasound at a frequency capable of imaging relevant anatomical landmarks in the mouse skull was successfully utilized both to image the mouse brain and to open the BBB in the presence of ultrasound contrast agent. Longer duration pulses (> 2 µs over a

30-second insoniﬁcation time, at PRFs of 10-100Hz, for a total number of cycles from 105 to 108) with low pressure amplitudes (1.6 to 6.2 MPa , MI 0.65) were found ∼ pp ≤ to allow MR contrast agent to enter the brain without blood cell extravasation. B-mode also opened the BBB but resulted in blood cell extravasation. However, by using standard, system settings with a low MI (e.g., PW Doppler, 15% power, maximum gate size), the BBB was successfully opened without damage. The results of this study can be used to gauge the potential of other custom sequences or existing diagnostic regimes for studies to locally deliver drugs or other therapeutic agents through the BBB.

6.6 Acknowledgments

This work was supported by NIH grants 2R01 EB-002132 and 1R01 CA-114075, as well as an NSF Graduate Research Fellowship. All work was performed at the Duke Center for In Vivo Microscopy, an NIH/NCRR National Biomedical Technol- ogy Resource Center (P41 RR005959) and an NCI Small Animal Imaging Resource Program (U24 CA092656). The authors would also like to thank Dr. Thomas Cum- mings for his assistance in histologic evaluation and Siemens Medical Solutions USA, Inc. Ultrasound Division for their technical assistance.

139 Chapter 7

Conclusions and Future Work

7.1 Conclusions

A diagnostic system has been shown herein to successfully localize and perform focal blood-brain barrier (BBB) disruption using a mechanical mechanism. Moder- ate sustained local temperature increases for use with low temperature-sensitive liposome drug delivery were achieved. However, the system could not generate sustained acoustic output for tissue ablation. The fundamental limitation was determined to be the transducer. Several design modiﬁcations to reduce internal heating in custom, diagnostic transducers while maintaining diagnostic image quality were evaluated, including using a PZT-4 multilayer composite, a lossless lens, passive cooling, and cMUT technology. These were shown to be insuﬃcient to prevent damage. Based on these results, a passively and actively cooled, PZT-4 multilayer composite transducer is the most promising diagnostic transducer design for therapeutic purposes. Further- more, several diagnostic transducers were evaluated in terms of their potential for performing therapeutic applications and one was selected for use in ablation. Abla- tion was not feasible without transducer damage. Internal transducer heating was determined to be the key limitation. Technologies showing promise for maintaining reasonable, but inferior, image quality and generating the required output for small

140 ablation lesions from the HIFU side include piezocomposites with air backing [13], the use of synthetic aperture and single-transmit focus imaging [37], and the use of thermally tolerant matching layers with an annular array design [61]. Using a diagnostic system, two therapeutic techniques were shown to be feasible using a diagnostic array. The necessary temperature rise for the release of the contents from thermally activated liposomes with a standard diagnostic transducer was achieved. However, an improved release assay must be developed to conclusively demonstrate this method. For applications in the brain, ultrasound, in conjunction with ultrasonic contrast agent, was shown to eﬀectively open selected regions of the

BBB using a diagnostic system with a non-thermal mechanism.

7.2 Future Work

7.2.1 Transducer Design

The analysis presented in this dissertation provides evidence and suggestions for diagnostic transducer design modifications to perform therapeutic applications. HIFU transducer designs modified to perform imaging functions are currently under development as dual-mode ultrasound arrays [37]. However, a modified diagnostic array would provide better image quality and possibly be easier to transfer to the clinic. Although actively cooled, PZT-4 multilayer composite designs may be beneficial for increasing the robustness of diagnostic arrays for hyperthermia applications, we conclude that dual-mode arrays for ablation would best be designed from the

HIFU perspective.

141 7.2.2 Hyperthermia Applications

The work presented herein shows the successful production of heat required for thermal release of liposomal contents for chemotherapeutic applications. Another potential application to explore for a similar heating regime would be gene therapy. Ultrasonic heating can be used to control transgene expression when the gene is under the control of a heat-sensitive promoter, such as heat shock protein (hsp) 70 [69, 105]. The up-regulation of hsps may play a critical role in inducing an anti-tumor immunity [129]. An absolute temperature between 42 and 50◦C has been shown to increase hsp70 expression [69], although higher temperatures may be used [71, 72]. The work in this dissertation has shown that a diagnostic system is capable of achieving temperatures within this range in vivo.

7.2.3 Blood-Brain Barrier Applications

BBB disruption was achieved using a diagnostic system and ultrasonic contrast agent herein. This method of disruption could also be used for eﬃcient gene transfer into the brain [116]. Although phase aberration and skull attenuation are possibly prohibitive to the use of a diagnostic system for BBB disruption through the human skull, a catheter probe may be introduced through a small burr hole and used for localized opening of the BBB. A Siemens AcunavTM transducer has already been shown to be capable of excellent localization of structures in the brain through such a burr hole, as shown in Figure 7.1. Future work could include testing the BBB disruption capabilities of this transducer. The Acunav can reach a non-derated pressure of 2.72 MPapp (55% system voltage, 5.33 MHz, 2 cm focus), which was repeatedly used by the VF10-5 (4% system voltage, 5.71 MHz, 2 cm focus) to open the BBB with duty cycles as low as 0.07%. This resulting in situ pressure through 2

142 cm of brain with the skull removed would exceed the in situ pressure of 1.21 MPapp delivered to the mouse brain (through skull) by the VF10-5. The ﬁrst question to investigate would be whether the Acunav is capable of maintaining this pressure at the prescribed duty cycles and total insoniﬁcation time used in chapter 6 without damage to itself.

(a) (b)

Figure 7.1: Images of a canine brain using an Acunav transducer for a) B-mode imaging of gyri and sulci and b) Color doppler imaging of the internal carotid artery.

143 Appendix A

Liposomal Release Assay

This appendix details the methods and results for an ex vivo liposomal release assay. The complications with this assay are discussed, and suggestions for future work are given.

A.1 Methods

A.1.1 Liposome Evaluation

For this study, the lipid shell of each liposome contained the same components found in the current Duke formulation for doxorubicin liposomes. Rhodamine B (excitation=554 nm, emission=627 nm) was then conjugated to the lipids to la- bel the liposome shell. The rhodamine labeling was designed to show the presence and amount of liposomes in a given region, regardless of whether release had occurred. The liposomes were loaded with carboxyfluorescein (excitation=492 nm, emission=517 nm). Carboxyfluorescein was chosen because it is self-quenching (non- fluorescing due to high concentrations) when contained within the liposome but not once it is released.

Two control experiments were performed before testing the release of the liposomes with ultrasound. The ﬁrst experiment was to test the release kinetics, or

144 percent release from exposure to different temperatures through time, of the liposomes in plasma. The purpose of this experiment was to ensure that there was a quantifiable difference between the release at 37◦C (body temperature) and 41.3◦C (minimum quoted temperature for ultrasonic heating release of liposomes). The plasma was placed in a glass tube to reach equilibrium with the waterbath temperature. At the start of the experiment, liposomes (10 mg/mL) were added to the plasma to yield a concentration of 5 µL/mL of liposome solution in plasma. Sixty microliters of the liposome-plasma mixture were removed from the heat at 2 minute intervals (or more often if the liposomes released quickly at the given temperature) for 30 minutes and placed in an Ependorf tube containing 1 mL of Hepes buffer (ph=7.4) at 4◦C for 20 seconds. After 20 seconds, the Ependorf tube was placed at room temperature. Three minutes after removing a sample from the heated waterbath, it was placed in a cuvette, and the fluorescence of the carboxyfluorescein was measured with an RF-1501, spectrofluorophotometer (Shimadzu Scientific In- struments, Columbia, MD). After measuring the release from all the time points for a given temperature, 5 µL of TritonTM X (Dow Chemical Company, Cary, NC) was added to each sample and allowed to incubate for 10 minutes. After 10 minutes, each sample was vortexed for a few seconds, and the fluorescence was measured in the spectrofluorophotometer. Because Triton acts as an emulsifier, these measurements served as the fluorescence level for 100% release. This process was repeated at 30, 37, 39, 40, 41.3, 42, and 45◦C. The second experiment was to test the release of the liposomes due to freezing to determine whether cryosectioning for histologic evaluation of release was a viable option. Ten microliters of 10 mg/mL liposomes were diluted in 500 µL of saline and exposed to one of the following conditions: 4◦C (normal storage temperature), -20◦C freezer for 1 hour, -80◦C for 1 hour (dry ice, typical freezing temperature prior

145 to cryosectioning), or 50◦C for 20 minutes (full release condition). To measure the fluorescence, a sample from each condition was diluted by 9x to measure fluorescence values within the dynamic range of the spectrofluorophotometer.

A.1.2 Ex Vivo Release Assay

After the liposome controls were completed, the ex vivo release of the carboxyfluorescein from the liposomes using ultrasonic heating was evaluated. A fixed needle holder and alignment apparatus was used to guide the 27 gauge, 3.5 inch spinal needle to the ultrasound focus each time (5 cm depth, centered in elevation and azimuth). The insertion point on the top surface of the muscle was marked with a surgical marker, and the muscle was placed in 37◦C phosphate-buffered saline (PBS) and allowed to thermally equilibriate. The transducer and needle holder apparatus were translated using a 3-D stepper motor controlled translation stage (Newport, Irvine, CA) to align the needle guide with the marked spot of the surface of the muscle. The needle/syringe was then inserted into the tissue and a low power B- mode image was acquired. Fifty microliters of liposomes (25 mg/mL) were injected into the muscle, and a 90-second ultrasound sequence (F/1.5, 4.44 MHz, 10 cycle, unapodized, M-mode, 7.1 kHz PRF) was transmitted at 35% system voltage. This sequence resulted in 58 seconds or longer of heat at or above 41.3◦C, as shown in

Figure 5.2. After the sequence terminated, another low power B-mode sequence was acquired, the needle was retracted from the muscle, and the muscle was removed from the PBS solution. An approximate 2 cm (elevation dimension) by 5 cm (lateral dimension) by muscle thickness (near 2 cm) section was cut around the injection

site, placed ﬂat on a petri dish, and frozen at -20◦C. This experimental process was repeated for six muscle samples. Six controls were also performed in the same exact manner except that the ultrasound sequence was not transmitted during the 90

146 seconds alotted for heating. Each muscle sample was cut into 20-µm thick sections in the lateral-axial plane at 0.5 mm intervals in elevation over the range where the orange color of the liposome injectate solution was visible to the naked eye. These sections were stored at -20◦C for 36 hours prior to fluorescence microscopy evaluation. Images were taken of each section at 1.6X magnification while exciting the rhodamine and then the carboxyfluorescein wavelengths.

A.2 Results

A.2.1 Liposome Evaluation

Release kinetics studies on the carboxyfluorescein-containing, rhodamine-labeled liposomes showed that the liposomes were fully released at temperatures of 40◦C and above after 30 minutes of heating. After this same time period at 37◦C, the liposomes only released 16% of their contents. Similarly, for 60 seconds of heating, approximately 50% of the carboxyfluorescein was released at 45◦C, while only 2% was released at 37◦C. At temperatures at or above 41.3◦C, the fastest rate of release occurred in the first one to two minutes. The release of the liposomes in response to freezing conditions were evaluated.

As shown in Table A.1, liposomal release due to freezing at -80◦C is similar to that of heating for 100% release; however, -20◦C resulted in only 27% release. The saline (which was not strictly pH-balanced (7.4)) used for dilution in this test may be the cause of the 24% release at 4◦C (normal liposome storage temperature). The

conclusion from this test was that samples at -20 and 4◦C had similar release levels.

147 Temperature (◦C) Fluorescence % Release -80◦C 397 86 -20◦C 124 27 4◦C 109 24 50◦C 459 100

Table A.1: Carboxyfluorescein fluorescence and release at different temperatures as measured by a spectrofluorophotometer. The liposomes were exposed to each temperature for 1 hour except the full release condition at 50◦C, which only required 20 minutes. The precision of the spectrofluorophotometer was 0.6. One sample was evaluated per tem- ± perature.

A.2.2 Ex Vivo Release

The needle alignment device was used to inject the liposomes at the same location relative to the transducer for each trial (5 cm focal depth centered in lateral and elevation). Figure A.1 shows a B-mode image acquired before injection of the needle and its guide along with a subtraction image (after injection - before injection), where the liposome injectate can clearly be seen. The injectate was a sphere with a diameter of approxiately 3 mm according to the subtraction image. The amount of carboxyfluorescein released relative to the amount of liposome present for each trial is shown in Table A.2. Although a significant difference was not observed between the control and heated conditions overall (p=0.11), the heated values are slightly higher for each quantity evaluated. The heated liposomes had a 1.4 times increase in the absolute maximum seen across all sections and trials. Ex- ample fluorescence images of the sections with the absolute minimum and maximum amount of release for the control and heated conditions are shown in Figure A.2.

148 (a)

(b)

Figure A.1: B-mode before injection of liposomes (a) and subtraction image showing lipsome injectate (b) in chicken breast muscle. Dashed yellow line shows the M-mode line used for ultrasonic heating.

Control Heated p-value Mean 1.27 0.45 1.36 0.53 0.11 ± ± Mean Min 0.71 0.24 1.21 0.42 0.03 ± ± Mean Max 2.06 0.20 2.26 0.61 0.46 ± ± Abs. Min 0.44 0.55 – Abs. Max 2.34 3.28 –

Table A.2: The ratio of carboxyﬂuorescein to rhodamine (release to amount of liposomes present) is given for the controls and ultrasonically heated liposomes. The mean, mean of the minimums (per sample), mean of the maximums (per sample), absolute minimum, and absolute maximum were taken across all sections in all samples. The p-value from a two-sample t-test is shown.

149 Control Heated

(a) (b)

Figure A.2: Example images at 1.6x magnification (height=4.1 mm, width=5.4 mm) of liposomal release from control (a & c) and ultrasonically heated (b & d) conditions. Green shows carboxyfluorescein release, and red shows presence of lipid shell of liposomes. The ratio of carboxyfluorescein to rhodamine in these images were a) 0.44, b) 0.55, c) 2.34, and d) 3.28, and they represent the minimum (a & b) and maximum (c & d) ratios seen over all trials for the control and heated groups.

A.3 Discussion

The lipid shell of the liposomes used in these experiments is designed to have the fastest release at 41.3◦C, while higher temperatures still achieve 100% release but over a long time period. The release kinetics curves in plasma indicate that about 25 times more ﬂuorescence should be seen at the carboxyﬂuorescein wavelengths at 41.3◦C for 60 seconds compared to 37◦C. Ex vivo and in vivo heating measurements ≥ (see Chapter 5) demonstrated that these temperatures could be achieved for the

150 speciﬁed amount of time (and longer) without damage to the transducer. However, the release assay did not provide conclusive evidence that ultrasonic heating resulted in increased release. Although the heated liposomes did show a slightly greater release than the controls overall, it was not signiﬁcant (p=0.11).

There are several reasons that could explain the failure of the assay. Freezing at -20◦C for one hour did not show release; however, the time period necessary for cryosectioning all of the samples is on the order of a day. Therefore, prolonged freezing may result in release, which was seen in each of the controls. Second, ex vivo tissue contains enzymes and other released cellular components that could be

breaking down the lipid shell of the liposomes. These two reasons could explain why we saw a signiﬁcant amount of carboxyﬂuorescein release in the control images, but the release kinetics suggest that a maximum of 8% should be released for the 5 to 7 minutes that each chicken breast muscle sample was in the 37◦C PBS. Finally, in

order to visually see a more distinctive diﬀerence in release, it may be necessary to heat for longer and over a larger area. Heating for as long as 30 minutes should ensure full release, although the fastest rate of release occurs within the ﬁrst one to two minutes.

A few ﬁnal tests are recommended for this release assay. A control where the liposomes are diluted in PBS (to eliminate the potential pH problem) and frozen at -20◦C for 1, 12, 24, 36, and 48 hours should be done and compared to 4◦C and full release conditions. These time points will show whether freezing for a reasonable amount of time to perform cryosectioning and evaluating the sections using

microscopy causes signiﬁcant release. If release occurs at the later time points, then an in vivo assay is suggested. If the results are favorable for the use of cryosectioning, the same ex vivo experiment should be performed but using a heated waterbath to achieve the desired temperature at the focus without ultrasound heating (41.3◦C)

151 for 60 seconds and 30 minutes. Additional test cases should be added where 1) liposomes are injected into chicken breast muscle at room temperature (no PBS) and 2) the liposomes are injected into muscle in a 34◦C PBS bath (no release expected and 41.3◦C case can still be achieved). If the room temperature sample releases

a significant amount compared to the 30 minute heated case, then it can be con- cluded that the enzymes, etc. in the dead tissue are releasing the liposomes. If the 37◦C case results in significantly more release than the room temperature case, the combination of enzymes and body temperature are enough to cause release. If the 60-second heated case results in significantly more release than the 37◦C case, then

the registration between the liposome injection and the ultrasonically heated region was not aligned well enough to achieve the proper heating of the liposomes in the alotted time and should be adjusted. Note that the beam is aligned with the right portion of the injectate in Figure A.1. Scanning the M-mode line over a 3 mm x 3

mm area (size of injectate herein) will aid in ensuring full heating coverage of the liposomes. If the 30-minute heated case results in significantly more release than the 60 second heated case, then a longer heating sequence is required. If a significant difference in release is not observed between the 30 minute heated case and the 34◦C case, then this assay will be proven inadequate for showing liposome release with ultrasound or any other heating method. If a significant difference in release is not observed, then another assay should be developed, and it is believed that an in vivo assay based on working methods in the literature would be best. In a study at the National Institutes of Health,

increased release was demonstrated by administering the liposomes through a tail vein injection in mice and then applying a pulsed-HIFU exposure to a tumor. In order to measure this release, the mouse was euthanized and perfused with saline to clear the vasculature of unreleased liposomes, and the tumor was excised. The

152 tumor was weighed, placed in acidic ethanol, homogenized, rotated overnight in the dark at 4◦C, centrifuged, and then placed in a well plate to be read by a ﬂuorimeter. The ﬂuorescence readings were compared to a standard curve of serial dilutions of doxorubicin [34]. Another possible assay would be to monitor tumor growth after

treatment with either LTSLs or LTSLs plus ultrasonic heating. In vivo experiments would have the added beneﬁt of virtually eliminating the possibility that cavitation could be aiding in the liposome release, which is a possibility when injecting a pool of liposomes into ex vivo tissue.

A.4 Acknowledgments

The author would like to thank Pavel Yarmolenko and Ji-Young Park for their assistance with the liposomes and Laura Dyer for her assistance with histology. The author would also like to thank Drs. Gerald Grant, Mark Dewhirst, and David Needham for providing the liposomes and motivation for this study and Siemens Medical Solutions USA, Inc. Ultrasound Division for their technical assistance.

153 Bibliography

[1] Acoustic Output Labeling Standard for Diagnostic Ultrasound Equipment. American Institute of Ultrasound in Medicine, Laurel, MD, 1992.

[2] Acoustic Output Measurement and Labeling Standard for Diagnostic Ultrasound Equipment. American Institute of Ultrasound in Medicine, Rockville, MD, 1992.

[3] Acoustic Output Measurement Standard for Diagnostic Ultrasound Equipment. National Electrical Manufacturers Association, Washington, D.C., 1998.

[4] Exposure Criteria for Medical Diagnostic Ultrasound: II. Criteria Based on All Known Mechanisms. National Council on Radiation Protection and Measurements, Bethesda, MD, 2002.

R [5] Deﬁnity Vial for (Perﬂutren Lipid Microsphere) Injectable Suspension. Package insert. Bristol-Myers Squibb Medical Imaging, North Billerica, MA, September 2004.

R [6] Magnevist (brand of gadopentetate dimeglumine) Injection. Package insert. Berlex Imaging, Montville, New Jersey, 2008.

[7] AIUM. How to interpret the ultrasound output display standard for higher acoustic output diagnostic ultrasound devices. Ultrasound Med. Biol., 17(8):535–538, 1998.

[8] R.E. Apfel and C.K. Holland. Gauging the likelihood of cavitation from short-pulse, low-duty cycle diagnostic ultrasound. Ultrasound Med. Biol., 17:179–185, 1991.

[9] J.F. Aubry, M. Tanter, M. Pernot, J.L. Thomas, and M. Fink. Experimental demonstration of noninvasive transskull adaptive focusing based on prior computed tomography scans. J. Acoust. Soc. Am., 113(1):84–93, 2003.

[10] M.R. Bailey, V.A. Khokhlova, O.A. Sapozhnikov, S.G. Kargl, and L.A. Crum. Physical mechanisms of the therapeutic eﬀect of ultrasound (a review). Acoustical Physics, 49(4):369–388, 2003.

154 [11] P. Ballabh, A. Braun, and M. Nedergaard. The blood-brain barrier: an overview; structure, regulation, and clinical implications. Neurobiology of Disease, 16:1–13, 2004.

[12] R. Berriet and G. Fleury. Design of a piezocomposite matric transducer conﬁguration for multi-mode operation in HIFU applications. IEEE Ultrasonics Symposium, pages 2377–2380, 2007.

[13] G. Bouchoux, C. Lafon, R. Berriet, J. Chapelon, G. Fleury, and D. Cathignol. Dual-mode transducer for image-guided interstitial thermal therapy. Ultrasound Med. Biol., 34(4):607–616, 2008.

[14] S. Burgess, V. Zderic, and S. Vaezy. Image-guided acoustic hemostasis for hemorrhage in the posterior liver. Ultrasound Med. Biol., 33(1):113–0, 2007.

[15] H.G. Burkitt, A. Stevens, J.S. Lowe, and B. Young. Wheater’s Basic Histopathology: A Colour Atlas and Text. Churchill Livingstone, New York, 3rd edition, 1996.

[16] E.L. Carstensen, D. Dalecki, S.M. Gracewski, and T. Christopher. Nonlinear propagation and the output indices. J Ultrasound Med, 18:69–80, 1999.

[17] L. Chen, G. ter Haar, C.R. Hill, M. Dworkin, P. Carnochan, H. Young, and J.P.M. Bensted. Eﬀect of blood perfusion on the ablation of liver parenchyma with high-intensity focused ultrasound. Phys. Med. Biol., 38:1661–1673, 1993.

[18] S-Q. Cheng, X-D. Zhou, Z-Y. Tang, Y. Yu, S-S. Bao, and D-C. Dian. Iodized oil enhances the thermal eﬀect of high-intensity focused ultrasound on ablating experimental liver cancer. J Cancer Res Clin Oncol, 123:639–644, 1997.

[19] S-Q. Cheng, X-D. Zhou, Z-Y. Tang, Y. Yu, H-Z. Wang, S-S. Bao, and D-C. Qian. High-intensity focused ultrasound in the treatment of experimental liver tumour. J Cancer Res Clin Oncol, 123:219–223, 1997.

[20] C-W. Cho, Y. Liu, W.N. Cobb, T.K. Henthorn, K. Lillehei, U. Christians, and K-Y. Ng. Ultrasound-induced mild hyperthermia as a novel approach to increase drug uptake in brain microvessel endothelial cells. Pharmaceutical Research, 19(8):1123–1129, 2002.

[21] J.J. Choi, M. Pernot, S.A. Small, and E.E. Konofagou. Noninvasive, transcranial and localized opening of the blood-brain barrier using focused ultrasound in mice. Ultrasound Med. Biol., 33(1):95–104, 2007.

155 [22] T. Christopher and E.L. Carstensen. Finite amplitude distortion and its relationship to linear derating formulae for diagnostic ultrasound systems. Ultrasound Med. Biol., 22(8):1103–1116, 1996.

[23] R.L. Clarke and G.R. Ter Haar. Temperature rise recorded during lesion formation by high-intensity focused ultrasound. Ultrasound Med. Biol., 23(2):299–306, 1997.

[24] G.T. Clement, P.J. White, R.L. King, N. McDannold, and K. Hynynen. A magnetic resonance imaging-compatible, large-scale array for trans-skull ultrasound surgery and therapy. J Ultrasound Med, 24:1117–1125, 2005.

[25] L. Curiel, R. Souchon, O. Rouviere, A. Gelet, and J.Y. Chapelon. Elastography for the follow-up of high-intensity focused ultrasound prostate cancer treatment: initial comparison with MRI. Ultrasound Med. Biol., 31(11):1461–8, 2005.

[26] C. Daft, S. Calmes, D. da Graca, K. Patel, P. Wagner, and I. Ladabaum. Microfabricated ultrasonic transducers monolithically integrated with high voltage electronics. IEEE Ultrasonics Symposium, pages 493–496, 2004.

[27] C. Daft, P. Wagner, S. Panda, and I. Ladabaum. Elevation beam proﬁle control with bias polarity patterns applied to microfabricated ultrasound transducers. IEEE Ultrasonics Symposium, pages 1578–1581.

[28] D. Dalecki, E.L. Carstensen, K.J. Parker, and D.R. Bacon. Absorption of ﬁnite amplitude focused ultrasound. J. Acoust. Soc. Am., 89(5):2435–2447, 1991.

[29] C.A. Damianou, K. Hynynen, and X. Fan. Evaluation of accuracy of a theoretical model for predicting the necrosed tissue volume during focused ultrasound surgery. IEEE Trans. Ultrason., Ferroelec., Freq. Contr., 42(2):182–187, 1995.

[30] C.A. Damianou, N.T. Sanghvi, F.J. Fry, and R. Maass-Moreno. Dependence of ultrasonic attenuation and absorption in dog soft tissues on temperature and thermal dose. J. Acoust. Soc. Am., 102(1):628–634, 1997.

[31] S. Datta, C-C. Coussios, A.Y. Ammi, D. Mast, G.M. De Courten-Myers, and R C.K. Holland. Ultrasound-enhanced thrombolysis using Deﬁnity as a cavitation nucleation agent. Ultrasound Med. Biol., 34(9):1421–1433, 2008.

156 [32] P.A. Dayton, K.E. Morgan, A.L. Klibanov, G. Brandenburger, K.R. Nightingale, and K.W. Ferrara. A preliminary evaluation of the eﬀects of primary and secondary radiation forces on acoustic contrast agents. IEEE Trans. Ultrason., Ferroelec., Freq. Contr., 44(6):1264–1277, 1997.

[33] D.L. Deardorﬀ and C.J. Diederich. Ultrasound applicators with internal water-cooling for high-powered interstitial thermal therapy. IEEE Transactions on Biomedical Engineering, 47(10):1356–1365, 2000.

[34] S. Dromi, V. Frenkel, A. Luk, B. Traughber, M. Angstadt, M. Bur, J. Poﬀ, J. Xie, S.K. Libutti, K.C.P. Li, and B.J. Wood. Pulsed-high intensity focused ultrasound and low temperature-sensitive liposomes for enhanced targeted drug delivery and antitumor eﬀect. Clin Cancer Res, 13(9):2722–2727, 2007.

[35] F.A. Duck. Physical Properties of Tissue. Harcourt Brace Jovanovich, San Diego, 1990.

[36] F.A. Duck. Nonlinear acoustics in diagnostic ultrasound. Ultrasound Med. Biol., 28(1):1–18, 2002.

[37] E.S. Ebbini, H. Yao, and A. Shrestha. Dual-mode ultrasound phased arrays for image-guided surgery. Ultrasonic Imaging, 28(2):65–82, 2006.

[38] BJ Fahey, SJ Hsu, PD Wolf, RC Nelson, and GE Trahey. Liver ablation guidance with acoustic radiation force impulse imaging: Challenges and opportunities. Phys Med Biol, 51:3785–3808, 2006.

[39] BJ Fahey, KR Nightingale, RC Nelson, ML Palmeri, and GE Trahey. Acoustic radiation force impulse imaging of the abdomen: Demonstration of feasibility and utility. Ultrasound in Medicine and Biology, 31:1185–1198, 2005.

[40] B.J. Fahey and G.E. Trahey. Abdominal acoustic radiation force impulse imaging. IEEE Ultrasonics Symposium, 1:736–739, 2004.

[41] G. Fleury, R. Berriet, J.Y. Chapelon, G. ter Haar, C. Lafon, O. Le Baron, L. Chupin, F. Pichonnat, and J. Lenormand. Safety issues for HIFU transducer design. International Symposium on Therapeutic Ultrasound 2005 Proceedings, 2005.

[42] J.B. Fowlkes, J.S. Abramowicz, C.C. Church, C.K. Holland, D.L. Miller, W.D. O’Brien, N.T. Sanghvi, M.E. Stratmeyer, and J.F. Zachary. American institute of ultrasound in medicine consensus report on potential bioeﬀects of diagnostic ultrsound: Executive summary. Journal of Ultrasound in Medicine, 27:,503–515, 2008.

157 [43] V. Frenkel, A. Etherington, M. Greene, J. Quijano, J. Xie, F. Hunter, S. Dromi, and K.C.P. Li. Delivery of liposomal doxorubicin (doxil) in a breast cancer tumor model: Investigation of potential enhancement by pulsed-high intensity focused ultrasound exposure. Academic Radiology, 13(4):469–479, 2006. [44] W.J. Fry and R.B. Fry. Determination of absolute sound levels and acoustic absorption coefficients by thermocouple probes - experiment. The Journal of the Acoustical Society of America, 26(3):311–317, 1954. [45] K.L. Gentry. Integrated Catheters for 3-D Intracardiac Echocardiography and Ultrasound Ablation. PhD thesis, Duke University, 2005. [46] D.E. Goertz, M. Frinlink, A. Bouakaz, C.T. Chin, N. de Jong, and A.W.F. van der Steen. The effect of bubble size on nonlinear scattering from microbubbles at high frequencies. 2003 IEEE Ultrasonics Symposium, pages 1503–1506, 2003. [47] R.L. Goldberg, M.L. Jurgens, C. Henriquez, D. Vaughan, D.M. Mills, and S.W. Smith. Modeling of piezoelectric multilayer ceramics using finite element analysis. IEEE Trans. Ultrason., Ferroelec., Freq. Contr., 44:1204–1211, 1997. [48] M. Hashimoto, T. Ueno, and M. Iriki. What roles does the organum vasculosum laminae terminalis play in fever in rabbits? Pflugers¨ Arch Eur J Physiol, 429:50–57, 1994. [49] J. Heikkila and Hynynen. Simulations of lesion detection using a combined phased array LHMI-technique. Ultrasonics, In Press, 2008. [50] C.D. Herrickhoff, E.D. Light, K.F. Bing, S. Mukundan, G.A. Grant, P.D. Wolf, and S.W. Smith. Intracranial catheter for integrated 3D ultrasound imaging and hyperthermia: Feasibility study. Ultrasound Med. Biol., In submission. [51] C.R. Hill. Optimum acoustic frequency for focused ultrasound surgery. Ultrasound Med. Biol., 20(3):271–277, 1994. [52] M.W. Hooker. Properties of PZT-based Piezoelectric Ceramics Between -150 and 250◦C. NASA/CR, Hampton, VA, 1998. [53] J.W. Hunt, M. Arditi, and F.S. Foster. Ultrasound transducers for pulse-echo medical imaging. IEEE Transactions on Biomedical Engineering, 30(8):453–480, 1983. [54] K. Hynynen and D.K. Edwards. Temperature measurements during ultrasound hyperthermia. Med. Phys., 16(4):618–626, 1989.

158 [55] K. Hynynen, N. McDannold, H. Martin, F.A. Jolesz, and N. Vykhodtseva. The threshold for brain damage in rabbits induced by bursts of ultrasound in the presence of an ultrasound contrast agent (Optison). Ultrasound Med. Biol., 29(3):473–481, 2003. [56] K. Hynynen, N. McDannold, N.A. Sheikov, F.A. Jolesz, and N. Vykhodtseva. Local and reversible blood-brain barrier disruption by noninvasive focused ultrasound at frequencies suitable for trans-skull sonications. Neuroimage, 24:12–20, 2005. [57] K. Hynynen, N. McDannold, N. Vykhodtseva, and F.A. Jolesz. Noninvasive MR imaging-guided focal opening of the blood-brain barrier in rabbits. Radiology, 220:640–646, 2001. [58] National Cancer Institute. Liver cancer. http://www.cancer.gov/cancertopics/types/liver/. [59] J.A. Jakobsen, R. Oyen, H.S. Thomsen, and S.K. Morcos. Safety of ultrasound contrast agents. Eur Radiol, 15:941–945, 2005. [60] J.A. Jensen and N.B. Svendsen. Calculation of pressure ﬁelds from arbitrarily shaped, apodized, and excited ultrasound transducers. IEEE Trans. Ultrason., Ferroelec., Freq. Contr., 39:262–267, 1992. [61] P. Kaczkowski, B. Cunitz, and G. Keilman. An eight-element annular array for image-guided high intensity focused ultrasound. Acoustics 2008, 2008. [62] J.E. Kennedy. High-intensity focused ultrasound in the treatment of solid tumours. Nat Rev Cancer, 5(4):321–7, 2005. [63] J.E. Kennedy, F. Wu, G.R. ter Haar, F.V. Gleeson, R.R. Phillips, M.R. Middleton, and D. Cranston. High-intensity focused ultrasound for the treatment of liver tumours. Ultrasonics, 42:931–935, 2004. [64] G.S. Kino. Acoustic Waves: Devices, Imaging, and Analog Signal Processing. Prentice-Hall, Inc., Englewood Cliﬀs, NJ, 1987. [65] M. Kinoshita, N. McDannold, F.A. Jolesz, and K. Hynynen. Noninvasive localized delivery of herceptin to the mouse brain by MRI-guided focused ultrasound-induced blood-brain barrier disruption. Proceedings of the National Academy of Sciences, 103(31):11719–11723, 2006. [66] M. Kinoshita, N. McDannold, F.A. Jolesz, and K. Hynynen. Targeted delivery of antibodies through the blood-brain barrier in MRI-guided focused ultrasound. Biochemical and Biophysical Research Communications, 340:1085–1090, 2006.

159 [67] G. Kong, G. Anyarambhatla, W.P. Petros, R.D. Braun, O.M. Colvin, D. Needham, and M.W. Dewhirst. Efficacy of liposomes and hyperthermia in a human tumor xenograft model: Importance of triggered drug release. Cancer Research, 60:6950–6957, 2000. [68] D. Kopelman, Y. Inbar, A. Hanannel, D. Freundlich, D. Castel, A. Perel, A. Greenfeld, T. Salamon, M. Sareli, A. Valeanu, and M. Papa. Magnetic resonance-guided focused ultrasound surgery (MRgFUS): Ablation of liver tissue in a porcine model. European Journal of Radiology, 59:157–162, 2006. [69] D.E. Kruse, M.A. Mackanos, C.E. O’Connel-Rodwell, C.H. Contag, and K.W. Ferrara. Short-duration-focused ultrasound stimulation of hsp70 expression in vivo. Phys. Med. Biol., 53(13):3641–360, 2008. [70] G. Kun and M. Wan. Effects of fascia lata on HIFU lesioning in vitro. Ultrasound Med. Biol., 30(7):991–998, 2004. [71] Y. Liu, T. Kon, C. Li, and P. Zhong. High intensity focused ultrasound-induced gene activation in sublethally injured tumor cells in vitro. J. Acoust. Soc. Am., 118(5):3329–3336, 2005. [72] Y. Liu, T. Kon, C. Li, and P. Zhong. High intensity focused ultrasound-induced gene activation in solid tumors. J. Acoust. Soc. Am., 120(1):492–501, 2006. [73] F.L. Lizzi, R. Muratore, C.X. Deng, J.A. Ketterling, S.K. Alam, S. Mikaelian, and A. Kalisz. Radiation-force technique to monitor lesions during ultrasonic therapy. Ultrasound Med. Biol., 29(11):1593–1605, 2003. [74] R.A. Lyon, J.S. Plugge, D.R. Mullen, M.G. Curley, and J.W. Sliwa. Active thermal control of ultrasound transducers. Patent, 1996. [75] I.R.S. Makin, T.D. Mast, W. Faidi, M. Runk, P.G. Barthe, and M.H. Slayton. Miniaturized ultrasound arrays for interstitial ablation and imaging. Ultrasound Med. Biol., 31(11):1539, 2005. [76] S.S. Manson. Report 1170: Behavior of materials under conditions of thermal stress. National Advisory Committee for Aeronautics, 1954. [77] N. McDannold, N. Vykhodtseva, and K. Hynynen. Targeted disruption of the blood-brain barrier with focused ultrasound: association with cavitation activity. Phys. Med. Biol., 51:793–807, 2006. [78] N. McDannold, N. Vykhodtseva, and K. Hynynen. Use of ultrasound pulses combined with Definity for targeted blood-brain barrier disruption: A feasibility study. Ultrasound in Med. & Biol., 33(4):584–590, 2007.

160 [79] N. McDannold, N. Vykhodtseva, and K. Hynynen. Blood-brain barrier disruption induced by focused ultrasound and circulating preformed microbubbles appears to be characterized by the mechanical index. Ultrasound Med. Biol., 34(5):834–840, 2008.

[80] N. McDannold, N. Vykhodtseva, and K. Hynynen. Eﬀects of acoustic parameters and ultrasound contrast agent dose on focused-ultrasound induced blood-brain barrier disruption. Ultrasound Med. Biol., 34(6):930–937, 2008.

[81] N. McDannold, N. Vykhodtseva, F.A. Jolesz, and K. Hynynen. MRI investigation of the threshold for thermally induced blood-brain barrier disruption and brain tissue damage in the rabbit brain. Magnetic Resonance in Medicine, 51:913–923, 2004.

[82] R.B. McIntosh, P.E. Mauger, and S.R. Patterson. Capacitive transducers with curved electrodes. IEEE Sensors Journal, 6(1):125–138, 2006.

[83] R.E. McKeighen. Design guidelines for medical ultrasonic arrays. SPIE International Symposium on Medical Imaging, 1998.

[84] M.J. McKinley, A.M. Allen, P. Burns, L.M. Colvill, and B.J. Oldﬁeld. Interaction of circulating hormones with the brain: The roles of the subfornical organ and the organum vasculosum of the lamina terminalis. Clinical and Experimental Pharmacology and Physiology, 25(S1):S61–S67, 1998.

[85] A.H. Mesiwala, L. Farrel, H.J. Wenzel, D.L. Silbergeld, L.A. Crum, H.R. Winn, and P.D. Mourad. High-intensity focused ultrasound selectively disrupts the blood-brain barrier in vivo. Ultrasound Med. Biol., 28:389–400, 2002.

[86] J.K. Mills and D. Needham. Lysolipid incorporation in dipalmitoylphosphatidylcholine bilayer membranes enhances the ion permeability and drug release rates at the membrane phase transition. Biochimica et Biophysica Acta (BBA) - Biomembranes, 1716(2):77–96, 2005.

[87] R. Mirzalou, N. Oliver, and T. Petersen. Thermal control of lens of ultrasound transducers. Patent, 2006.

[88] D. Needham and M.W. Dewhirst. The development and testing of a new temperature-sensitive drug delivery system for the treatment of solid tumors. Advanced Drug Delivery Reviews, 53:285–305, 2001.

161 [89] K.R. Nightingale, M.S. Soo, M.L. Palmeri, A.N. Congdon, K.D. Frinkley, and G.E. Trahey. Imaging tissue mechanical properties using impulsive acoustic radiation force. 2nd IEEE International Symposium on Biomedical Imaging: Macro to Nano, 1:41–44, 2004.

[90] A. Nikoozadeh, B. Bayram, G.G. Yaralioglu, and B.T. Khuri-Yakub. Analytical calculation of collapse voltage of cmut membrane. IEEE Ultrasonics Symposium, pages 256–259, 2004.

[91] L. Nock, G. Trahey, and S.W. Smith. Phase aberration correction in medical ultrasound using speckle brightness as a quality factor. J. Acoust. Soc. Am., 85(5):1819–1833, 1989.

[92] W.L. Nyborg. Physical Acoustics, volume II.B, chapter 11, pages 265–331. Academic Press, Inc., New York, 1965.

[93] W.L. Nyborg. Solutions of the bio-heat transfer equation. Phys. Med. Biol., 33(7):785–792, 1988.

[94] M.L. Palmeri, K.D. Frinkley, and K.R. Nightingale. Experimental studies of the thermal eﬀects associated with radiation force imaging of soft tissue. Ultrasonic Imaging, 26:29–40, 2004.

[95] M.L. Palmeri, S.A. McAleavey, G.E. Trahey, and K.R. Nightingale. Ultrasonic tracking of acoustic radiation force-induced displacements in homogeneous media. IEEE Trans. Ultrason., Ferroelec., Freq. Contr., 53(7):1300–1313, 2006.

[96] A.M. Ponce, Z. Vujaskovic, F. Yuan, D. Needham, and M.W. Dewhirst. Hyperthermia mediated liposomal drug delivery. International Journal of Hyperthermia, 22(3):205–213, 2006.

[97] C.J. Price, T.D. Hoyda, and A.V. Ferguson. The area postrema: A brain monitor and integrator of systemic autonomic state. The Neuroscientist, 14(2):182–194, 2008.

[98] T. Proulx. Teleconference with author., 2005.

[99] H. Rauhut. New Types of Microelectronic Epoxy Compounds. Technical paper. Dexter, 1994.

[100] S.B. Raymond, J. Skoch, K. Hynynen, and B.J. Bacskai. Multiphoton imaging of ultrasound/Optison mediated cerebrovascular eﬀects in vivo. Journal of Cerebral Blood Flow & Metabolism, 27:393–403, 2007.

162 [101] S.B. Raymond, L.H. Treat, J.D. Dewey, N.J. McDannold, K. Hynynen, and B.J. Bacskai. Ultrasound enhanced delivery of molecular imaging and therapeutic agents in Alzheimer’s disease mouse models. PLoS One, 3(5):1–7, 2008.

[102] M. Reinhard, A. Hetzel, S. Kruger, S. Kretzer, J. Talazko, S. Ziyeh, J. Weber, and T. Els. Blood-brain barrier disruption by low-frequency ultrasound. Stroke, 37:1546–1548, 2006.

[103] C. Richard, H.S. Lee, and D. Guyomar. Thermo-mechanical stress eﬀect on 1-3 piezocomposite power transducer performance. Ultrasonics, 42:417–424, 2004.

[104] R. Righetti, F. Kallel, R.J. Staﬀord, R.E. Price, T.A. Krouskop, J.D. Hazle, and J. Ophir. Elastographic characterization of hifu-induced lesions in canine livers. Ultrasound Med. Biol., 25(7):1099–1113, 1999.

[105] C. Rome, R. Deckers, and C.T. Moonen. Handb Exp Pharmacol, volume 185. Springer Berlin Heidelberg, 2008.

[106] K.Y. Saleh and N.B. Smith. Two-dimensional ultrasound phased array design for tissue ablation for treatment of benign prostatic hyperplasia. Int. J. Hyperthermia, 20(1):7–31, 2004.

[107] R. Salomir, J. Palussiere, S.L. Fossheim, A. Rogstad, U.N. Wiggen, N. Grenier, and C.T.W. Moonen. Local delivery of magnetic resonance (MR) contrast agent in kidney using thermosensitive liposomes and MR imaging-guided local hyperthermia: A feasibility study in vivo. Journal of Magnetic Resonance Imaging, 22:534–540, 2005.

[108] E. Sassaroli and K. Hynynen. Forced linear oscillations of microbubbles in blood capillaries. J. Acoust. Soc. Am., 115(6):3235–3243, 2004.

[109] E. Sassaroli and K. Hynynen. Resonance frequency of microbubbles in small blood vessels: a numerical study. Phys. Med. Biol., 50:5293–5305, 2005.

[110] O. Saunders, S. Clift, and F. Duck. Ultrasound transducer self heating: development of 3-d ﬁnite-element models. Journal of Physics: Conference Series Advanced Metrology for Ultrasound in Medicine, 1:72–77, 2004.

[111] F. Schlachetzki, T. Holscher, H.J. Koch, B. Draganski, A. May, G. Schuierer, and U. Bogdahn. Observation on the integrity of the blood-brain barrier after microbubble destruction by diagnostic transcranial color-coded sonography. J Ultrasound Med, 21:419–429, 2002.

163 [112] R. Seip, W. Chen, J. Tavakkoli, L.A. Frizzell, and N.T. Sanghvi. High-intensity focused ultrasound (HIFU) phased arrays: Recent developments in transrectal transducers and driving electronics design. Third International Symposium on Therapeutic Ultrasound, 3003.

[113] A. Shaw, N.M. Pay, R.C. Preston, and A.D. Bond. Proposed standard thermal test object for medical ultrasound. Ultrasound Med. Biol., 25(1):121–132, 1999.

[114] A. Shaw and G. ter Haar. Requirements for measurement standards in high intensity focused ultrasound (HIFU) ﬁelds. National Physical Laboratory Report, DQL AC 015, 2006.

[115] N. Sheikov, N. McDannold, N. Vykhodtseva, F. Jolesz, and K. Hynynen. Cellular mechanisms of the blood-brain barrier opening induced by ultrasound in presence of microbubbles. Ultrasound in Med. & Biol., 30(7):979–989, 2004.

[116] M. Shimamura, N. Sato, Y. Taniyama, S. Yamamoto, M. Endoh, H. Kurinami, M. Aoki, T. Ogihara, Y. Kaneda, and R. Morishita. Development of eﬃcient plasmid DNA transfer into adult rat central nervous system using microbubble-enhanced ultrasound. Gene Therapy, 11:1532–1539, 2004.

[117] T.V. Sinilo, V.A. Khokhlova, and O.A. Sapozhnikov. Experimental veriﬁcation of enhancement of HIFU-inducing heating of tissue mimicking phantoms due to acoustic nonlinearity. J. Acoust. Soc. Am., 115(5):2449, 2004.

[118] M. Tanter, J. Thomas, and M. Fink. Focusing and steering through absorbing and aberrating layers: Application to ultrasonic propagation through the skull. Journal of the Acoustical Society of America, 103(5):2403–2410, 1998.

[119] T.A. Tran, J.Y. Le Guennec, P. Bougnoux, F. Tranquart, and A. Bouakaz. Characterization of cell membrane response to ultrasound activated microbubbles. IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control, 55(1):44–49, 2008.

[120] E.C. Unger, T. Porter, W. Culp, R. Labell, T. Matsunaga, and R. Zutshi. Therapeutic applications of lipid-coated microbubbles. Advanced Drug Delivery Reviews, 56(9):1291–1314, 2004.

[121] W. Walters. Email with author., 2006.

164 [122] J.N. Weinstein, R.L. Magin, R.L. Cysyk, and D.S. Zaharko. Treatment of solid l1210 murine tumors with local hyperthermia and temperature-sensitive liposomes containing methotrexate. Cancer Research, 40:1388–1395, 1980.

[123] J. Wells, A. Sen, and S.W. Hui. Localized delivery to ct-26 tumors in mice using thermosensitive liposomes. International Journal of Pharmaceutics, 261:105–113, 2003.

[124] I.P. Wharton, I.H. Rivens, G.R. Ter Haar, D.J. Gilderdale, D.J. Collins, J.W. Hand, P.D. Abel, and N.M. Desouza. Design and development of a prototype endocavitary probe for high-intensity focused ultrasound delivery with integrated magnetic resonance imaging. J Magn Reson Imaging, 25(3):548–56, 2007.

[125] S.H. Wong, G.C. Scott, S.M. Conolly, G. Narayan, and D.H. Liang. Feasibility of noncontact intracardiac ultrasound ablation and imaging catheter for treatment of atrial ﬁbrillation. IEEE Trans. Ultrason., Ferroelec., Freq. Contr., 53(12):2394–2405, 2006.

[126] S.H. Wong, I.O. Wygant, D.T. Yeh, X. Zhuang, B. Bayram, M. Kupnik, O. Oralkan, A.S. Ergun, G.G. Yaralioglu, and B.T. Khuri-Yakub. Capacitive micromachined ultrasonic transducer arrays for integrated diagnostic/therapeutic catheters. AIP Conference Proceedings, 829:395–9, 2006.

[127] F. Wu, W-Z. Chen, J. Bai, J-Z. Zou, Z-L. Wang, H. Zhu, and Z-B. Wang. Tumor vessel destruction resulting from high-intensity focused ultrasound in patients with solid malignancies. Ultrasound Med. Biol., 28(4):535–542, 2002.

[128] F. Wu, W-Z. Chen, J. Bai, J-Z. Zou, Z-L. Wang, H. Zhu, and Z-H. Wang. Pathological changes in human malignant carcinoma treted with high-intensity focused ultrasound. Ultrasound Med. Biol., 27(8):1099–1106, 2001.

[129] F. Wu, Z-B. Wang, W-Z. Chen, J-Z. Zou, J. Bai, H. Zhu, K-Q. Li, F-L. Xie, C-B. Jin, H-B. Su, and G-W. Gao. Extracorporeal focused ultrasound surgery for treatment of human solid carcinomas: Early chinese clinical experience. Ultrasound Med. Biol., 30(2):245–260, 2004.

[130] G.G. Yaralioglu, A.S. Ergun, B. Bayram, E. Haeggstrom, and B.T. Khuri-Yakub. Calculation and measurement of electromechanical coupling coeﬃcient of capacitive micromachined ultrasonic transducers. IEEE Trans. Ultrason., Ferroelec., Freq. Contr., 50(4):449–456, 2003.

165 [131] X. Zhen, M. Raghavan, T.L. Hall, C. Ching-Wei, M.-A. Mycek, J.B. Fowlkes, and C.A. Cain. High speed imaging of bubble clouds generated in pulse ultrasound cavitational therapy - histotripsy. IEEE Trans. Ultrason., Ferroelec., Freq. Contr., 54(10):2091–2101, 2007.

[132] M. Zipparo. Email correspondence with author., 2008.

[133] M.J. Zipparo. Mid- to high-power ultrasound imaging arrays - from arﬁ to HIFU. Proceedings of the 2003 IEEE Ultrasonics Symposium, 1:684–688, 2003.

166 Biography

Kristin Frinkley Bing

Date of Birth: July 17, 1981 Place of Birth: Orlando, FL

Education:

Duke University, Pratt School of Engineering, Durham, NC Ph.D. in Biomedical Engineering, November 2008 Case Western Reserve University, Case School of Engineering, Cleveland, OH B.S.E. in Biomedical Engineering, Minor in Artiﬁcial Intelligence, May 2003

Select Publications:

K.F. Bing, G.P. Howles-Banerji, Y. Qi, and K.R. Nightingale. Blood-brain barrier (BBB) disruption using a diagnostic scanner and Deﬁnity in mice. Ultrasound in Medicine and Biology, In Review.

K.D. Frinkley, S.J. Rosenzweig, and K.R. Nightingale. Therapeutic potential metric for diagnostic transducers. Proceedings of the 2007 IEEE Ultrasonics Symposium, pages 16-119, 2007.

K.D. Frinkley, M.L. Palmeri, and K.R. Nightingale. Controlled spatio-temporal heating patterns using a commercial, diagnostic ultrasound system. Proceedings of the 2005 IEEE Ultrasonics Symposium, 2: 1130-1134, 2005.

M.L. Palmeri, K.D. Frinkley, R.W. Nightingale, Gregg E. Trahey, Kathryn R. Nightingale. On the balance of acoustic energy in soft tissue during diagnostic and therapeutic ultrasound imaging. in preparation.

C.D. Herickhoﬀ, E.D. Light, K.F. Bing, S. Mukundan, G.A. Grant, P.D. Wolf, and S.W. Smith. Intracranial catheter for integrated 3D ultrasound imaging & hyperthermia: Feasibility study. Ultrasound in Medicine and Biology, In Review.

167 M.L. Palmeri, J.J. Dahl, M.H. Wang, K.D. Frinkley, and K.R. Nightingale. Quantifying hepatic shear modulus in vivo using acoustic radiation force. Ultrasound in Medicine and Biology, 34(4): 546-58, 2008.

M.L. Palmeri, K.D. Frinkley, K.G. Oldenburg, and K.R. Nightingale. Characterizing acoustic attenuation of homogeneous media using focused impulsive acoustic radiation force. Ultrasonic Imaging, 28(2): 114-128, 2006.

M.L. Palmeri, K.D. Frinkley, L. Zhai, M. Gottfried, R.C. Bentley, K. Ludwig, and K.R. Nightingale. Acoustic radiation force impulse (ARFI) imaging of the gastrointestinal tract. Ultrasonic Imaging, 27: 75-88, 2005.

M.L. Palmeri, K.D. Frinkley, and K.R. Nightingale. Experimental studies of the thermal eﬀects associated with radiation force imaging of soft tissue. Ultrasonic Imaging, 26: 29-40, 2004.

Patents/Invention Disclosures:

G.A. Johnson, G.P. Howles-Banerji, K.D. Frinkley, K.R. Nightingale, M.L. Palmeri. ”Active staining for in vivo MR imaging in the brain.” Provisional Patent. Submitted October 2008.

M.L. Palmeri, K.R. Nightingale, G.E. Trahey, K.D. Frinkley. ”Shear wave velocity estimation and shear modulus reconstruction using acoustic radiation force impulse imaging.” Patent. Submitted February 2008.

M.L. Palmeri, K.D. Frinkley, G.E. Trahey, K.R. Nightingale. ”Characterizing ultrasonic attenuation using impulsive acoustic radiation force.” Invention Disclosure. Submitted March 2006.

Awards and Honors:

National Science Foundation Graduate Research Fellowship (2003) National Defense Science and Engineering Graduate Fellowship (2003) Summa Cum Laude (2003) Biomedical Engineering Society’s Rita Schaﬀer Undergraduate Award (2003) Robert J. Adler Undergraduate Engineering Student Award (2003) YWCA of Greater Cleveland’s Dr. Jennie S. Hwang Award (2002) Tau Beta Pi National Engineering Society (2001)

168