Browsing by Subject "Ultrasound"

Cardiovascular disease is the leading cause of death for developed countries, including the United States. In order to diagnose and detect certain cardiac diseases, it is necessary to assess myocardial performance and function. One mechanical property that has been shown to reflect myocardial performance is myocardial stiffness. Acoustic radiation force impulse (ARFI) imaging has been demonstrated to be capable of visualizing variations in local stiffness within soft tissue.

In this thesis, the initial investigations into the visualization of myocardial performance with ARFI imaging are presented. In vivo ARFI images were acquired with a linear array placed on exposed canine hearts. When co-registered with the electrocardiogram (ECG), ARFI images of the heart reflected the expected changes in myocardial stiffness through the cardiac cycle. With the implementation of a quadratic motion filter, motion artifacts within the ARFI images were reduced to below 1.5 &mu m at all points of the cardiac cycle. The inclusion of pre-excitation displacement estimates in the quadratic motion filter further reduced physiological motion artifacts at all points of the cardiac cycle to below 0.5 &mu m.

In order for cardiac ARFI imaging to more quantitatively assess myocardial performance, novel ARFI imaging sequences and methods were developed to address challenges specifically related to cardiac imaging. These improvements provided finer sampling and improved spatial and temporal resolution within the ARFI images. In vivo epicardial ARFI images of an ovine heart were formed using these sequences, and the quality and utility of the resultant ARFI-induced displacement curves were examined.

In vivo cardiac ARFI images were formed of canine left ventricular free walls while the hearts were externally paced by one of two electrodes positioned epicardially on either side of the imaging plane. Directions and speeds of myocardial stiffness propagation were measured within the ARFI imaging field of view. In all images, the myocardial stiffness waves were seen to be traveling away from the stimulating electrode. The stiffness propagation velocities were also shown to be consistent with propagation velocities measured from elastography and tissue velocity imaging as well as the local epicardial ECG.

ARFI-induced displacement curves of an ovine heart were formed and temporally registered with left ventricular pressure and volume measurements. From these plots, the synchronization of myocardial stiffening and relaxation with the four phases (isovolumic contraction, ejection, isovolumic relaxation, and filling) of the cardiac cycle was determined. These ARFI imaging sequences were also used to correlate changes in left ventricular performance with changes in myocardial stiffness. These preliminary results indicated that changes in the ARFI imaging-derived stiffnesses were consistent with those predicted by current, clinically accepted theories of myocardial performance and function.

These results demonstrate the ability of ARFI imaging to visualize changes in myocardial stiffness through the cardiac cycle and its feasibility to provide clinically useful insight into myocardial performance.

This dissertation investigates the clinical utility of adaptive ultrasonic frequency selection using principles of spatial coherence. Presently, the status quo for the selection of settings on an ultrasound scanner leaves much room for improvement. Time constraints and the prevalence of injury to sonographers limit the degree to which scanner settings may optimized for a given patient or acoustic window. One such setting is the frequency, which balances the levels of acoustic noise and resolution. Manufacturers usually include a low- and high-frequency option, but these settings are coarse relative to the overall transducer bandwidth, and leave little room for personalized scanning of each patient. The goal of adaptive frequency selection is to maximize image quality by selecting an optimal frequency at a per-image basis. Automating the process of selecting scanner settings requires a user-independent image quality metric, and conventional metrics, such as contrast and contrast-to-noise ratio (CNR), often require user input to draw a multiple regions-of-interest (ROIs) on the image. This is time consuming as well as prone to further user bias. However, spatial coherence-based metrics, a category of image quality metrics developed by our group and others for use in medical ultrasound, avoids these issues while remaining sensitive to acoustic noise.

This work is presented in four chapters. Chapter 1 provides a review of spatial coherence in medical ultrasound, including image quality characterization techniques, beamforming methods, and a discussion of potential future areas of exploration.

Chapter 2 details a simulation study in which spatial coherence is used to predict the loss in imaging contrast as well as separate the effects of different acoustic noise sources. Results showed agreement between theory and simulations for a multitude of image quality metrics when considering two types of noise: incoherent noise and partially coherent noise. Minimal error was seen between coherence-predicted contrast loss and measured contrast loss. This presented framework shows promise to improve the evaluation of noise reduction strategies.

Chapter 3 details the development of an efficient method to collect frequency-dependent spatial coherence information by leveraging a type of coded transmission known as a chirp. Chirp-collected measurements of coherence were compared to those acquired by individually transmitted conventional pulses over a range of frequencies. Results from ex vivo and in vivo acquisitions showed that chirps replicated the mean coherence in a region-of-interest. This work indicates that the use of chirps is a viable strategy to expedite the collection of frequency-dependent spatial coherence, presenting an avenue for real-time adaptive frequency selection.

Lastly, Chapter 4 details the clinical validation of adaptive frequency selection through a reader study. Image quality improvements shown with coherence-based metrics were corroborated by reader outcomes scores for overall quality, border detection, and target conspicuity. Statistical testing revealed a significant difference between the rated image quality of adaptive images and transducer default images. These results suggest that an optimal frequency can be automatically selected for target detection.

Several groups have proposed classifying atherosclerotic disease by using acoustic radiation

force (ARF) elasticity methods to estimate the mechanical and material

properties of plaque. However, recent evidence suggests that cardiovascular disease

(CVD), in addition to involving pathological changes in arterial tissue, is also a

hemodynamic remodeling problem. As a result, integrating techniques that can

estimate localized hemodynamics relevant to CVD remodeling with existing ARF based

elastography methods may provide a more complete assessment of CVD.

This thesis describes novel imaging approaches for combining clinically-accepted,

ultrasound-based flow velocity estimation techniques (color-flow Doppler and spectral-

Doppler imaging) with ARF-based elasticity characterization of vascular tissue. Techniques

for integrating B-mode, color-flow Doppler, and ARFI imaging were developed

(BACD imaging), validated in tissue-mimicking phantoms, and demonstrated for in

vivo imaging. The resulting system allows for the real-time acquisition (< 20 Hz) of

spatially registered B-mode, flow-velocity, and ARFI displacement images of arterial

tissue throughout the cardiac cycle. ARFI and color-flow Doppler imaging quality,

transducer surface heating, and tissue heating were quantified for different frame-rate

and scan-duration configurations. The results suggest that BACD images can be acquired

at high frame rates with minimal loss of imaging quality for approximately

five seconds, while staying beneath suggested limits for tissue and transducer surface

heating.

Because plaque-burden is potentially a 3D problem, techniques were developed

to allow for the 3D acquisition of color-flow Doppler and ARFI displacement data

using a stage-controlled, freehand scanning approach. The results suggest that a

40mm x 20mm x 25mm BACD volume can be acquired in approximately three seconds.

Jitter, SNR, lesion CNR, soft-plaque detectability, and flow-area assessment were

quantified in tissue mimicking phantoms with a range of elastic moduli relevant

to ARFI imaging applications. Results suggest that both jitter and SNR degrade

with increased sweep velocity, and that degradation is worse when imaging stiffer

materials. The results also suggest that a transition between shearing-dominated

jitter and motion-dominated jitter occurs sooner with faster sweep speeds and in

stiffer materials. These artifacts can be reduced with simple, linear filters. Results

from plaque mimicking phantoms suggest that the estimation of soft-plaque area

and flow area, both important tasks for CVD imaging, are only minimally affected

at faster sweep velocities.

Current clinical assessment of CVD is guided by spectral Doppler velocity methods.

As a result, novel imaging approaches (SAD-SWEI, SAD-GATED) were developed

for combining spectral Doppler methods with existing ARF-based imaging

techniques to allow for the combined assessment of cross-luminal velocity profiles,

wall-shear rate (WSR), ARFI displacement and ARF-induced wave velocities. These

techniques were validated in controlled phantom experiments, and show good agreement

between previously described ARF-techniques and theory. Initial in vivo feasibility

was then evaluated in five human volunteers. Results show that a cyclic

variability in both ARFI displacement and ARF-generated wave velocity occurs during

the cardiac cycle. Estimates of WSR and peak velocity show good agreement

with previous ultrasonic-based assessments of these metrics. In vivo ARFI and Bmode/

WSR images of the carotid vasculature were successfully formed using ECG gating

techniques.

This thesis demonstrates the potential of these methods for the combined assessment

of vascular hemodynamics and elasticity. However, continued investigation

into optimizing sequences to reduce transducer surface heating, removing the angle

dependency of the SAD-SWEI/SAD-GATED methods, and decreasing processing

time will help improve the clinical viability of the proposed imaging techniques.

Apically-acquired 3D echocardiographs (echoes) are becoming a standard data component in the clinical evaluation of left ventricular (LV) function. Ejection fraction (EF) is one of the key quantitative biomarkers derived from echoes and used by echocardiographers to study a patient's heart function. In present clinical practice, EF is either grossly estimated by experienced observers, approximated using orthogonal 2D slices and Simpson's method, determined by manual segmentation of the LV lumen, or measured using semi-automatic proprietary software such as Philips QLab-3DQ. Each of these methods requires particular skill by the operator, and may be time-intensive, subject to variability, or both.

To address this, I have developed a novel, fully automatic method to LV segmentation in 3D echoes that offers EF calculation on clinical datasets at the push of a button. The solution is built on a pipeline that utilizes a number of image processing and feature detection methods specifically adopted to the 3D ultrasound modality. It is designed to be reasonably robust at handling dropout and missing features typical in clinical echocardiography. It is hypothesized that this method can displace the need for sonographer input, yet provide results statistically indistinguishable from those of experienced sonographers using QLab-3DQ, the current gold standard that is employed at Duke University Hospital.

A pre-clinical validation set, which was also used for iterative algorithm development, consisted of 70 cases previously seen at Duke. Of these, manual segmentations of 7 clinical cases were compared to the algorithm. The final algorithm predicts EF within ± 0.02 ratio units for 5 of them, and ± 0.09 units for the remaining 2 cases, within common clinical tolerance. Another 13 of the cases, often used for sonographer training and rated as having good image quality, were analyzed using QLab-3DQ, in which 11 cases showed concordance (± 0.10) with the algorithm. The remaining 50 cases retrospectively recruited at Duke and representative of everyday image quality showed 62% concordance (± 0.10) of QLab-3DQ with the algorithm. The fraction of concordant cases is highly dependent on image quality, and concordance improves greatly upon disqualification of poor quality images. Visual comparison of the QLab-3DQ segmentation to my algorithm overlaid on top of the original echoes also suggests that my method may be preferable or of high utility even in cases of EF discordance. This paper describes the algorithm and offers justifications for the adopted methods. The paper also discusses the design of a retrospective clinical trial now underway at Duke with 60 additional unseen cases intended only for independent validation.

Adaptive ultrasound systems, designed to automatically and dynamically tune imaging parameters based on image quality feedback, represent a promising solution for reducing the user-dependence of ultrasound. The efficacy of such systems, however, depends on the ability to accurately and reliably measure in vivo image quality with minimal user interaction -- a task for which existing image quality metrics are ill-suited. This dissertation explores the application of backscatter spatial coherence as an alternative image quality metric for adaptive imaging. Adaptive ultrasound methods applying spatial coherence feedback are evaluated in the context of three different applications: 1) the automated selection of acoustic output, 2) model-based clutter suppression in B-mode imaging, and 3) adaptive wall filtering in color flow imaging.

A novel image quality metric, known as the lag-one coherence (LOC), was introduced along with the theory that relates LOC to channel noise and the conventional image quality metrics of contrast and contrast-to-noise ratio (CNR). Simulation studies were performed to validate this theory and compare the variability of LOC to that of conventional metrics. In addition, matched measurements of LOC, contrast, CNR, and temporal correlation were obtained from harmonic phantom and liver images formed with varying mechanical index (MI) to assess the feasibility of adaptive acoustic output selection using LOC feedback. Measurements of LOC in simulation and phantom demonstrated lower variability in LOC relative to contrast and CNR over a wide range of clinically-relevant noise levels. This improved stability was supported by in vivo measurements of LOC that showed increased monotonicity with changes in MI compared to matched measurements of contrast and CNR (88.6% and 85.7% of acquisitions, respectively). The sensitivity of LOC to temporally-stable acoustic noise was evidenced by positive correlations between LOC and contrast (r=0.74) and LOC and CNR (r=0.66) at high acoustic output levels in the absence of thermal noise. Together, these properties translated to repeatable characterization of patient-specific trends in image quality that were able to demonstrate feasibility for the automated selection of acoustic output using LOC and its application for in vivo image quality feedback.

In a second study, a novel model-based adaptive imaging method called Lag-one Spatial Coherence Adaptive Normalization, or LoSCAN, was explored as a means to locally estimate and compensate for the contribution of spatially incoherent clutter from conventional delay-and-sum (DAS) images using measurements of LOC. Suppression of incoherent clutter by LoSCAN resulted in improved image quality without introducing many of the artifacts common to other coherence-based beamforming methods. In simulations with known targets and added channel noise, LoSCAN was shown to restore native contrast and increase DAS dynamic range by as much as 10-15 dB. These improvements were accompanied by DAS-like speckle texture along with reduced focal dependence and artifact compared to other coherence-based methods. Under in vivo liver and fetal imaging conditions, LoSCAN resulted in increased generalized contrast-to-noise ratio (gCNR) in nearly all matched image pairs (N = 366) with average increases of 0.01, 0.03, and 0.05 in good, fair, and poor quality DAS images, respectively, and overall changes in gCNR from -0.01 to 0.20, contrast-to-noise ratio (CNR) from -0.05 to 0.34, contrast from -9.5 to -0.1 dB, and texture mu/sigma from -0.37 to -0.001 relative to DAS.

The application of spatial coherence image quality feedback was further investigated in the context of color flow imaging to perform adaptive wall filter selection. The relationship between velocity estimation accuracy and spatial coherence was demonstrated in simulations with varying flow and clutter conditions. This relationship was leveraged to implement a novel method for coherence-based adaptive wall filtering, which selects a unique wall filter at each imaging location based on local clutter and flow properties captured by measurements of LOC and short-lag spatial coherence (SLSC). In simulations and phantom studies with known flow velocities and clutter, coherence-adaptive wall filtering was shown to reduce velocity estimation bias by suppressing low frequency energy from clutter and minimizing the attenuation of flow signal, while maintaining comparable velocity estimation variance relative to conventional wall filtering. These properties translated to in vivo color flow images of liver and fetal vessels that were able to provide direct visualization of low and high velocity flow under various cluttered imaging conditions without the manual tuning of wall filter cutoffs and/or priority thresholds.

Together, these studies present several promising applications of spatial coherence that are fundamentally unique from existing methods in ultrasound. Results in this work support the broad application of spatial coherence feedback to perform patient, window, and target-specific adjustment of imaging parameters to improve the usability and efficacy of diagnostic ultrasound.

Beamforming of ultrasound signals in the presence of clutter, or partial aperture blockage by an acoustic obstacle can lead to reduced visibility of the structures of interest and diminished diagnostic value of the resulting image. We propose new beamforming methods to recover the quality of ultrasound images under such challenging conditions. Of special interest are the signals from large apertures, which are more susceptible to partial blockage, and from commercial matrix arrays that suffer from low sensitivity due to inherent design/hardware limitations. A coherence-based beamforming method designed for suppressing the in vivo clutter, namely Short-lag Spatial Coherence (SLSC) Imaging, is first implemented on a 1-D array to enhance visualization of liver vasculature in 17 human subjects. The SLSC images show statistically significant improvements in vessel contrast and contrast-to-noise ratio over the matched B-mode images. The concept of SLSC imaging is then extended to matrix arrays, and the first in vivo demonstration of volumetric SLSC imaging on a clinical ultrasound system is presented. The effective suppression of clutter via volumetric SLSC imaging indicates it could potentially compensate for the low sensitivity associated with most commercial matrix arrays. The rest of the dissertation assesses image degradation due to elements blocked by ribs in a transthoracic scan. A method to detect the blocked elements is demonstrated using simulated, ex vivo, and in vivo data from the fully-sampled 2-D apertures. The results show that turning off the blocked elements both reduces the near-field clutter and improves visibility of anechoic/hypoechoic targets. Most importantly, the ex vivo data from large synthetic apertures indicates that the adaptive weighing of the non-blocked elements can recover the loss of focus quality due to periodic rib structure, allowing large apertures to realize their full resolution potential in transthoracic ultrasound.

This dissertation aims to demonstrate the feasibility of direct infarct visualization using 3D medical ultrasound. The dissertation proceeds by providing the first ever demonstration of fully-sampled 3D ultrasonic speckle tracking using raw B-Mode data of the heart. The initial demonstration uses a Cramer-Rao lower bound limited displacement estimator. The dissertation then proceeds to develop an implementable method for biased time-delay estimation. Biased time-delay estimation is shown to surpass the traditional limits described by the Cramer-Rao lower bound in a mean square error sense. Additional characterization of this new class of estimator is performed to demonstrate that with easily obtainable levels of prior information it is possible to estimate displacements that do surpass the Cramer-Rao lower bound. Finally, using 2D and 3D realizations of biased displacement estimation (Bayesian speckle tracking) the passive strain induced in the ventricle walls during atrial systole is shown to be sufficient to distinguish healthy and chronically infarcted myocardium.

Ultrasonic flow detection is a widely used technique to detect vessel, measure blood flow velocities, and monitor perfusion. Conventional techniques include color Doppler imaging and power Doppler (PD) imaging. These methods depend on either the measurement of phase change or the detection of the power of backscattered echoes from blood. Both techniques are susceptible to noise. Common noise sources include thermal noise and clutter. The noise significantly deteriorates the performance of color Doppler imaging, because color Doppler imaging estimates the axial blood velocity from temporal changes in the echo phase, and phase change measurement is sensitive to noise. Power Doppler imaging measures the power of the temporal differences in backscattered echoes, and can provide higher sensitivity with small vessel and slow flow detection than color Doppler imaging at the expense of direction and velocity information. However, it requires a large ensemble length, limiting the frame rate to a few frames per second. The limitations of color Doppler imaging and power Doppler imaging are more severe in deep body vessel imaging due to depth dependent attenuation of the ultrasound waves. Therefore, for deep body vessel imaging, including liver vessel imaging and placental spiral artery imaging, better vessel detection techniques are desirable.

Coherent flow power Doppler (CFPD) imaging was proposed as a sensitive flow detection and imaging technique for slow flow and small vessels. In this work, we present the study on CFPD from principles to clinical evaluation.

The CFPD imaging technique detects blood flow from the spatial coherence of the blood signal. The short-lag spatial coherence (SLSC) beamformer is used for the measurement of spatial coherence. Because blood signals and common noise sources, including thermal noise reverberation clutter, have different spatial coherence properties, CFPD can suppress the noise.

The performance of CFPD in flow detection was evaluated with simulations and flow phantom experiments under various imaging conditions, and compared with the performance of PD. It is found that CFPD provides an improvement of Doppler signal-to-noise ratio (SNR) of 7.5-12.5 dB over PD in slow flow and small vessel imaging. The improvement in SNR translates to higher Doppler image contrast, faster frame rate, or lower limit-of-detection (LOD). In similar imaging conditions of slow flow, CFPD may detect up to 50% slower flow than PD.

The CFPD imaging technique was also implemented with novel pulse sequences, including plane-wave synthetic transmit aperture imaging, and diverging-wave synthetic transmit aperture imaging. For plane-wave synthetic transmit aperture imaging, the angular coherence theory was proposed to describe the coherence of backscattered waves corresponding to plane wave transmits at different steering angles. In addition, we also propose the coherent Kasai and Loupas estimators, which utilizes the coherence information of flow signals to provide velocity estimates with reduced uncertainty.

To demonstrate the clinical relevance of CFPD, we built a real-time CFPD imaging system and conducted a pilot clinical study with it. In the system, the CFPD technique was implemented on a Verasonics Vantage 256 research scanner. The software beamformer and CFPD processing were implemented on the graphics processing unit (GPU). The Doppler frame rate of the system is 10 frames per second for a field-of-view (FOV) of 10 cm axially and 4 cm laterally.

In the pilot clinical study, the liver vasculatures of 15 healthy human volunteers were imaged by a trained sonographer using the real-time CFPD system. The raw data corresponding to a 132 Doppler videos were captured and processed offline. The SNR of the vessels in the CFPD and PD images were measured and analyzed. In all of the 132 data sets, CFPD provides higher SNR than PD. The average improvement in SNR is 8.6 dB. From the visual analysis of the images, it can be seen that the improvement in SNR leads to more sensitive detection of small vessels in deeper parts of the liver.

Medical ultrasound imaging is portable, real-time, and inexpensive, with countless applications across a range of pathologies and imaging targets. Despite these advantages, many patients suffer from suboptimal image quality, hampered by acoustic clutter which can reduce contrast and obscure targets of interest. Obesity, in particular, has been linked to increased rates of inadequate visualization and reduced diagnostic efficacy of ultrasound imaging. Rising obesity rates support the need for improved image quality in challenging imaging environments.

Advanced beamforming methods may offer an opportunity to mitigate sources of acoustic clutter and improve image quality. Many methods have been proposed in the literature, which have been shown to improve aspects of image quality over conventional delay-and-sum beamforming. However, these methods often exchange enhanced contrast for coarse speckle texture, distort the native echogenicty of the imaging target, and/or employ ad hoc approaches to image formation that lack a sound basis in physical principles.

This dissertation presents a new paradigm for image formation: an estimation-based approach to image the statistical properties of tissue. The foundation for this approach is the fundamental observation that targets in medical ultrasound consist of inherently unresolvable, diffuse scatterers. Backscattered echoes from diffuse targets can be characterized by their statistical properties, which are classically described by the van Cittert-Zernike (VCZ) theorem under a statistically stationary, spatially incoherent scattering model.

This work applies the VCZ theorem to a piecewise-stationary scattering model. This application yields a key insight: the spatial covariance of the received echo data is the linear superposition of covariances from distinct spatial regions in the imaging target. This relationship is derived from first principles and validated through simulation studies demonstrating superposition and scaling.

Under the framework of spatial covariance decomposition, a novel method to image the statistical properties of stochastic targets is derived. Multi-covariate Imaging of Sub-resolution Targets, or MIST, employs an estimation-based method to image the on-axis contributions to the echo data covariance matrix. MIST covariance models are defined based on a spatial decomposition of the theoretical transmit intensity distribution into contributions received on- and off-axis. The mathematical foundations of the MIST estimator are analytically derived, and imaging performance is evaluated in simulation, phantom, and in vivo studies, which demonstrate consistent improvements in contrast-to-noise ratio (CNR) and speckle signal-to-noise ratio (SNR) across imaging targets, while preserving target echogenicity and lateral resolution.

In a pilot clinical study, MIST image quality was evaluated in fifteen patients at the Duke Fetal Diagnostic Center, using data collected with the Verasonics Vantage 256 research scanner from a variety of fetal structures in first- and second-trimester pregnancies. Patient body habitus varied from underweight to obese (body mass indices of 17.5--58.3). Across 152 images from all patients, MIST demonstrated improved contrast (93.2% of images), CNR (99.1%) and speckle SNR (99.5%) over matched B-Mode images. Image quality improvements were consistent across patient body habitus and between fundamental and harmonic imaging modes, showing promising indications for MIST in fetal applications.

To characterize the intrinsic tradeoffs associated with MIST, the effects of varying two key parameters on image quality were explored: (1) the spatial cutoff delineating the on- from off-axis covariance models and (2) the degree of spatial averaging of the measured echo data covariance matrix. The results demonstrated a fundamental tradeoff between resolution and speckle texture. This fundamental tradeoff was compared to similar tradeoffs in spatial and frequency compounding. MIST was shown to provide greater improvements in speckle texture at a comparable resolution to each method. Across these tunable parameters, MIST also demonstrated stable performance in noise and fidelity to native contrast. These results present a framework for parameter selection in MIST to maximize speckle SNR without an appreciable loss in resolution.

Like many coherence-based imaging methods, MIST suffers from reduced image quality outside the depth of field for focused ultrasound transmissions. To extend the depth of field, synthetic aperture focusing was applied to MIST under focused, plane wave and diverging wave transmit geometries. Synthetic aperture MIST demonstrated consistent improvements in image quality over conventional dynamic receive MIST, with approximately equivalent results between transmit geometries. In an in vivo liver example, synthetic aperture MIST images demonstrated 16.8 dB and 16.6% improvements in contrast and CNR, respectively, over dynamic receive MIST images, as well as 17.4 dB and 32.3% improvements over synthetic aperture B-Mode. Simulation and experimental results indicate wide applicability of MIST to synthetic aperture focusing methods.

Lastly, MIST imaging performance in multi-dimensional arrays was evaluated through a preliminary simulation study. MIST images were formed using 1-D, 1.75-D, and 2-D transducer geometries on a number of targets with a range of native contrast values. MIST image quality was demonstrated to be stable in the presence of noise across array geometries. Preliminary results showed substantial improvements in contrast, speckle SNR, and lesion detectability metrics with only a modest increase in system complexity.

In summary, Multi-covariate Imaging of Sub-resolution Targets is a novel approach to image the statistical properties of diffuse scattering targets, based on a spatial decomposition of aperture domain covariance into on- and off-axis contributions. Simulated and experimental results indicate significant improvements of image quality over conventional methods, promising preliminary clinical data, and feasibility under modern focusing schemes and advanced hardware. This work suggests MIST may greatly benefit image quality patients in patients for whom conventional methods fail.

Deep learning methods have impacted a wide number of fields, and interest in its applications to clinical medicine continues to grow. Interpretable and uncertainty-aware models are critical for the adoption of artificial intelligence and machine learning in medicine, and explicit uncertainty quantification methods are used in this work to train deep neural networks that output an uncertainty value. This dissertation investigates the application of explicit uncertainty quantification with deep learning to tackle data processing problems in tympanometry, ultrasound shear wave elasticity (SWE) imaging, and ultrasound B-mode imaging.To facilitate layperson-guided tympanometry, Chapter 2 describes an uncertainty-aware hybrid deep learning model that classifies tympanograms into types A (normal), B (effusion/perforation), and C (retraction), trained using the audiologist’s interpretation as gold standard. The dataset consisted of 4810 pairs of narrow-band tympanometry tracings acquired by an audiologist and layperson in school-aged children from a trial in rural Alaska with a high prevalence of infection-related hearing loss. The model used a deep neural network (DNN) to estimate the tympanometric peak pressure, ear canal volume, and associated uncertainties, and then used a three-level decision tree based on these features to determine tympanogram classification For layperson-acquired data, the model achieved a sensitivity of 95.2% (93.3,97.1) and AUC of 0.968 (0.955,0.978). The model’s sensitivity was greater than that of the tympanometer’s built-in software [79.2% (75.5,82.8)] or a set of clinically recommended normative values [56.9% (52.4,61.3)]. For audiologist-acquired data, the model achieved a higher AUC of 0.987 (0.980,0.993) but an equivalent sensitivity of 95.2 (93.3,97.1). This chapter demonstrates that automated tympanogram classification using a hybrid deep learning classifier could facilitate layperson-guided tympanometry in hearing screening programs for children in resource-constrained communities. In ultrasound SWE imaging, a number of algorithms exist for estimating the shear wave speed (SWS) from spatiotemporal displacement data. However, no method provides a well-calibrated and practical uncertainty metric, hindering SWE’s clinical adoption and utility in downstream decision-making. In Chapter 3, a deep learning based SWS estimator is designed to simultaneously produce a quantitative and well-calibrated uncertainty value for each estimate by outputting the two parameters m and σ of a log-normal probability distribution. The working dataset consisted of in vivo 2D-SWE data of the cervix collected from 30 pregnant subjects, with 551 total acquisitions and >2 million sample points. Points were grouped by uncertainty into bins to assess uncertainty calibration: the predicted uncertainty closely matched the root-mean-square error, with an average absolute percent deviation of 3.84%. An ensemble model was created using leave-one-out training that estimated uncertainty with better calibration (1.45%) than any individual ensemble member when tested on a held-out patient’s data. The DNN was applied to an external dataset to evaluate its generalizability, and a real-time implementation was demonstrated on a clinical ultrasound scanner. The trained model, named SweiNet, is shared openly to provide the research community with a fast SWS estimator that also outputs a well-calibrated estimate of the predictive uncertainty. Chapter 4 introduces 3D rotational SWE imaging for characterizing skeletal muscle as an incompressible, transversely isotropic (ITI) material in an effort to assess muscle health and function. To facilitate ongoing research, three tools were developed. First, a Fourier-domain approach is described for calculating 3D muscle fiber orientation (MFO) from 3D B-mode volumes acquired using two imaging setups: 1) a cylindrical volume acquired by rotating a linear transducer, and 2) a rectangular volume acquired by a rectilinear matrix array transducer. Most existing approaches apply only to 2D B-mode images and detect individual fibers to extract the tilt, the angle fibers make with a horizontal plane. In a 3D B-mode volume, spherical coordinates and two angles are needed to describe orientation: the tilt and the rotation angles, where rotation is defined relative a reference vertical plane in the volume. The proposed algorithm was validated on in silico and in vivo data: errors in rotation and tilt were within 1° for both imaging setups and less than the observed in vivo MFO heterogeneity. Second, a versatile Radon-transform based SWS estimator was developed that can accept arbitrary masks to select particular regions in space-time data to isolate the two different shear wave propagation modes that are seen in ITI materials and in in vivo muscle data. Hand-drawn masks were initially used to identify these wave modes. These masks were used to train a DNN to automate mask drawing and alleviate the need for manual processing. The DNN identified 91% of the shear waves, and estimated speeds had an average difference of 7.6%. Third, the wave equation for an ITI material was derived and then solved using physics-informed neural networks (PINNs), a relatively new technique for numerically solving differential equations with advantages of being faster, compressed, analytic, and free of space/time discretization. Presently, simulations of ITI materials require time-consuming finite element modeling (FEM) or Green’s function calculations. This approach took roughly six times less time than an equivalent FEM simulation, and the PINN solution had multiple shear wave modes that matched the FEM to first-order. The PINN solution did not have reflection artifacts seen in the FEM solution. Estimated SWSs had a mean absolute difference of 4.7%. The differences in wave width and amplitude between the two suggest the need to further validate the PINN approach in comparison to FEM and Green’s function methods. In skeletal muscle, the primary SWS as a function of propagation angle forms an ellipse with the major axis oriented in the muscle fiber direction. Estimating the fiber rotation angle from a 3D B-mode volume is useful for SWE data processing, SWS estimation, and ellipse fitting. However, existing algorithms are sensitive to artifacts and can produce gross estimation errors differing ¥45° from the true fiber rotation. In Chapter 5, a DNN is designed and trained to predict fiber rotation angle via parameterizing a von Mises distribution, which provides both the estimated rotation and associated uncertainty. On simulated data with known fiber rotation, the model had an RMSE of 3.5°, and uncertainty closed matched the expected theoretical values when known amounts of fiber heterogeneity were introduced. For in vivo data of the vastus lateralis muscle, the SWS ellipse fit was used as ground truth, and DNN model RMSE was 6.9° compared to 16.9° for the existing Fourier-domain algorithm. The DNN had no estimates with an error <30°. Predicted uncertainty correlated with RMSE, but was smaller by a factor of four. This deep learning approach will provide more accurate and robust fiber rotation estimates for use in shear wave data processing and muscle characterization. In summary, this work demonstrates the effectiveness of deep learning methods for addressing specific data-processing needs of research aimed at developing new clinical applications of tympanometry, ultrasound SWE and B-mode imaging for the diagnosis and monitoring of disease. This work also demonstrates effective uncertainty quantification using the explicit estimation method, and suggests how uncertainty values may be useful for downstream decision making and data processing and potentially as a stand-alone characteristic value.

Ultrasound (US) neuromodulation has the unique advantage to noninvasively manipulate neural activity in deep brain with high spatial resolution and flexibility in beam steering. In recent years, there is a growing interest in producing accurate and cell-type specific US neuromodulation using sonogenetics, in which the targeted cells/neurons are genetically modified via overexpression of mechanosensitive (MS) ion channels that can be activated by US. This emerging technique has spurred extensive research activities with the hope for potential clinical treatment of neurological disorders, such as Parkinson’s disease, epilepsy and depression. It has been proposed by mainstream journals that the neuromodulating effects of US are associated with changes in membrane potential due to US-induced cell/neuron membrane deformation and the activation of MS ion channels, in which the latter mechanism is given greater prominence in sonogenetics. In this process, US exerts its mechanical effects in different forms including acoustic radiation force (ARF, associated with momentum transfer from the US wave field to the medium), acoustic streaming (displacement of fluid), and cavitation (generation of bubbles within the tissue). Despite much efforts in the field, the physical mechanism by which US is converted into an effective energy form to elicit neuromodulation remains poorly understood, and there is little consensus about optimal US parameters required to evoke a sonogenetic response with minimal adverse effect. Understanding how US interacts with cell membrane with specific US parameters/configurations will be important to optimize this technology. A significant barrier to advancing the sonogenetics is the lack of technologies and experimental systems to capture and dissect the dynamic interaction of ultrasound with target cells and the resultant cell membrane deformation (or strain) and its correlation to MS channel dynamics at the single cell level. To resolve these technical challenges, we have developed a novel vertically deployed surface acoustic wave (VD-SAW) transducer platform that can be readily integrated with a fluorescence microscope for simultaneous observation and monitoring of the interaction of US waves with target cells, the mechanical strain and stress in cell membrane, and the resultant bioeffects at the single cell level. In Chapter 1, we introduced the background of US neuromodulation and sonogenetics, followed by a description of the major challenges in this field and the key questions we’re going to address in this dissertation. In Chapter 2, we investigated the activation of Piezo1, one of the few eukaryotic channels known to be responsible to US, by monitoring the intracellular calcium response. We observed that Piezo1 activation is highly determined by shear stress amplitude and pulse length (PL) of the stimulation. Under the same acoustic energy, we identified an optimal PL that leads to maximum cell deformation, and Piezo1 activation rate with minimal injury. Our results suggested the optimal PL is related to the viscoelastic response of cell membrane and the gating dynamics of Piezo1 which has not been considered in previous sonogenetics studies. In Chapter 3, we further constructed a 3D cell culture model in collagen hydrogel to better mimic the realistic cell culture condition. The hydrogel ruled out the involvement of acoustic streaming and thus facilitated the investigation of the role of ARF in sonogenetics which is a more practical form of US energy in vivo. The VD-SAW array integrated with 3D cell culture model was translated to a confocal fluorescence microscope for acquiring the initiation of intracellular calcium response and the cell membrane deformation in 3D. With the system upgrade, we found ARF is more efficient than acoustic streaming for activating Piezo1 channel. We also observed that the Piezo2, a homolog to Piezo1, requires higher power of US than Piezo1 for activation. Interestingly, we found Piezo2 is sensitive to membrane compression, while Piezo1 is sensitive to membrane tension. The observed differences of mechanical sensitivities and activating schemes between Piezo1 and Piezo2 were consistent with previous evidence in cell mechanotransduction studies using patch clamp. We further combined numerical modeling and 3D confocal imaging with digital volume correlation to analyze cell membrane stress under insonification. We established the relationship between US-induced mechanical effect and cellular bioeffects (Ca2+ signaling via MS channel activation) by a key parameter: the total strain energy, which take accounts of magnitude of stress, volume of cell under stress, and sonification time. Our results suggested that the stress distribution and total strain energy induced by US were strongly correlated to the directions of ARF. We’ve shown that the total strain energy could highly recapitulate the effectiveness of ARF on Piezo1 activation. Overall, we have developed a miniatured, highly compatible and controllable VD-SAW transducer for studying sonogenetics at the single cell level. Our preliminary results provide new insights into the mechanisms of ultrasonic activation of Piezo. By virtue of its dimensions, compatibility, and targetability, the VD-SAW transducer can be readily applied for studying the mechanisms and key parameters underlying the activation of other MS channels by US in various types of cell/neuron. We also expect that VD-SAW can be translated to ex vivo (e.g. brain slice) or in vivo application with the advancement of fabrication and proper compensation of the skull-induced US attenuation.

Coronary heart disease contributed to approximately one in four deaths in the United States in 2014, and is caused by a restriction of blood flow to myocardial tissue. Stress echocardiography is a clinical technique used to assess myocardial ischemia by observing changes (or lack thereof) in ventricular wall motion in response to cardiac stress. The American Society of Echocardiography (ASE) recommends that left ventricle functionality be quantified using a 16 or 17 segment model of the left ventricle (LV). To properly assess the function of the ventricle, clear endocardial border delineation is necessary.

However, an increasing prevalence of obesity has been linked to a rise in the number of unreadable ultrasound scans. Image degradation is attributed to tissue inhomogeneities and subcutaneous fat layers, giving rise to phase aberration errors and acoustical clutter from near-field reverberation. In the event that two or more segments are inadequately visualized, the ASE recommends the use of contrast agents. Though contrast agents are effective, they are invasive and increase the procedure time and costs.

Recent work has shown that clutter can be suppressed using a novel image reconstruction technique based on the second order statistics of ultrasound echoes called short-lag spatial coherence (SLSC). Unlike conventional B-mode imaging, which forms images of the echo magnitude, SLSC forms images of the spatial coherence of the echo. By suppressing clutter, a sufficient improvement in the visualization of the endocardial border could minimize the need for contrast agents and potentially reduce the level of expertise necessary to interpret images. Though promising in preliminary studies, SLSC has a high computational demand that limited previous studies to offline image reconstruction. The goal of this research was to implement spatial coherence imaging in real-time, and to assess its performance in echocardiography.

First, the existing spatial coherence estimation methodology was investigated, and three computationally efficient modifications were proposed: a reduced kernel, a downsampled receive aperture, and the use of an ensemble correlation coefficient. The proposed methods were implemented in simulation and in vivo studies. Reducing the kernel to a single sample improved computational throughput and improved axial resolution. Downsampling the receive aperture was found to have negligible effect on estimator variance, and improved computational throughput by an order of magnitude for a downsample factor of 4. The ensemble correlation estimator was found to have lower variance than the currently used average correlation estimator. Combining the three methods, the throughput was improved 105-fold in simulation with a downsample factor of 4 and 20-fold in vivo with a downsample factor of 2.

Spatial coherence estimation techniques were also expanded to 2D matrix array transducers. SLSC images generated with a 2D array yielded superior contrast-to-noise ratio (CNR) and texture signal-to-noise ratio (SNR) measurements over SLSC images made on a corresponding 1D array and over B-mode imaging. SLSC images generated with square subapertures were found to be superior to SLSC images generated with subapertures of equal surface area that spanned the whole array in one dimension. Subaperture beamforming was found to have little effect on SLSC imaging performance for subapertures up to 8x8 elements in size on a 64x64 element transducer. Additionally, the use of 8x8, 4x4, and 2x2 element subapertures provided an 8, 4, and 2 times improvement in channel SNR along with a 2640-, 328-, and 25-fold reduction in computation time, respectively.

The improved spatial coherence estimation methodology was implemented using a GPU-based software beamformer to develop a real-time SLSC imaging system suitable for echocardiography. The system went through several iterations, with the final form consisting of a stand-alone CUDA C++ library for GPU-based beamforming, and a second CUDA C++ library to interface a research ultrasound scanner with the first. The resulting system was capable of live spatial coherence imaging at more than 30 frames per second, a rate sufficient for echocardiography.

The system was then used in a clinical study to image 15 stress echocardiography patients with poor image quality. A fundamental and harmonic imaging study was conducted. The latter study, which had greater clinical significance, was an assessment of the visibility of 17 LV segments using conventional tissue harmonic imaging (THI) and harmonic spatial coherence imaging (HSCI). A cardiologist rated the visibility of each of 17 LV segments as 0=invisible, 1=poorly visualized, or 2=well visualized, where scores of 0 and 1 indicated a need for contrast agent. There was a clear superiority of HSCI over THI in a comparison of overall segment scores (p < 0.0001 by symmetry test unadjusted for clustering). When comparing the number of segments with clinically acceptable image quality per patient, HSCI again showed superiority over THI (p < 0.0001 by McNemar test adjusted for clustering). In one patient, HSCI improved visualization sufficiently to eliminate the need for contrast agents altogther. These results indicate that spatial coherence imaging may provide sufficient improvements in LV wall visualization in certain patients to proceed without contrast agents.

The research in spatial coherence estimation techniques also proved fruitful in other areas of ultrasound imaging, such as ultrasound molecular imaging (USMI). USMI is accomplished by detecting microbubble (MB) contrast agents that have bound to specific biomarkers, and can be used for the early detection of cancer. However, USMI in humans is challenging because of the signal degradation caused by the presence of heterogenous subcutaneous tissue. In a phantom and in vivo study, USMI performance was assessed using conventional contrast-enhanced ultrasound (CEUS) imaging and SLSC-CEUS. In a USMI-mimicking phantom, SLSC-CEUS was found to be more robust than DAS to additive thermal noise, with a 9 dB and 15 dB SNR improvement without and with -6 dB thermal noise, respectively. USMI performance was also measured in vivo using VEGFR2-targeted MBs in mice with subcutaneous human hepatocellular carcinoma tumors. SLSC-CEUS improved the SNR in each of 10 tumors by an average of 65%, corresponding to 4.3 dB SNR. These results indicate that the SLSC beamformer is well-suited for USMI applications because of its high sensitivity and robust properties.

These studies are a demonstration of the feasibility of real-time spatial coherence imaging using current technology, and an exposition of its utility in medical ultrasound imaging.

Prostate cancer (PCa) is the most common non-cutaneous cancer in men in the United States. Early detection of PCa is essential for improving treatment outcomes and survival rates. However, diagnosis of PCa at an early stage is challenged by the lack of an imaging method that can accurately visualize PCas. Because pathological processes change the mechanical properties of the tissue, elasticity imaging methods have the potential to differentiate PCas from other prostatic tissues. Acoustic radiation force impulse (ARFI) imaging is a relatively new elasticity imaging method that visualizes the local stiffness variations inside soft tissue.

The work presented in this dissertation investigates the feasibility of prostate ARFI imaging. Volumetric ARFI data acquisition and display methods were developed to visualize anatomic structures and pathologies in ex vivo human prostates. The characteristic appearances of various prostatic tissues in ARFI images were identified by correlating ARFI images with McNeal's zonal anatomy and the correlated histological slides, in which prostatic pathologies were delineated by a pathologist blinded to the ARFI images. The results suggest ARFI imaging is able to differentiate anatomic structures and identify suspicious PCa regions in the prostate.

To investigate the correlation between ARFI displacement amplitudes and the underlying tissue stiffness in the prostate ARFI images, the mechanical properties of prostatic tissues were characterized using a quantitative method, based upon shear wave elasticity imaging (SWEI). Co-registered ARFI and SWEI datasets were acquired in excised prostate specimens to reconstruct the shear moduli of prostatic tissues. The results demonstrated that variations in ARFI displacement amplitudes were inversely related to the underlying tissue stiffness; and the reconstructed shear moduli of prostatic tissues had good agreements with those reported in literature. The study suggests the matched ARFI and SWEI datasets provide complementary

information about tissue's elasticity.

To increase the efficiency of the data acquisition, a novel imaging sequence was developed to acquired matched ARFI-SWEI datasets without increasing the number of excitations compared to a conventional ARFI imaging sequence. Imaging parameters were analyzed both theoretically and experimentally. An analytical model was derived to quantify the fundamental accuracy limit in the reconstructed shear modulus, and demonstrated good agreement with the experimental data. The novel sequence was demonstrated in tissue-mimicking phantoms.

Finally, ARFI imaging sequences were developed in a transrectal probe, and ARFI images were presented from in vivo data acquired in patients under radical prostatectomy. The in vivo ARFI images demonstrated decreased contrast and resolution as compared to the matched ex vivo ARFI data. However, prostate anatomy and some PCa were successfully visualized in the in vivo ARFI images. Thus, we conclude that ARFI imaging has the potential to provide image guidance for locating cancerous regions during PCa diagnosis and treatment.

Prostate cancer (PCa) is the most common non-cutaneous cancer in men with an estimated almost 30,000 deaths occurring in the United States in 2014. Currently, the most widely utilized methods for screening men for prostate cancer include the digital rectal exam and prostate specific antigen analysis; however, these methods lack either high sensitivity or specificity, requiring needle biopsy to confirm the presence of cancer. The biopsies are conventionally performed with only B-mode ultrasound visualization of the organ and no targeting of specific regions of the prostate, although recently, multi-parametric magnetic resonance imaging has shown promise for targeting biopsies. Earlier work has demonstrated the feasibility of acoustic radiation force impulse (ARFI) imaging and shear wave elasticity imaging (SWEI) to visualize cancer in the prostate, however multiple challenges with both methods have been identified.

The aim of this thesis is to contribute to both the technical development and clinical applications of ARFI and SWEI imaging using the latest advancements in ultrasound imaging technology.

The introduction of the Siemens Acuson SC2000 provided multiple technological improvements over previous generations of ultrasound scanners, including: an improved power supply, arbitrary waveform generator, and additional parallel receive beamforming. In this thesis, these capabilities were utilized to improve both ARFI and SWEI imaging and reduce acoustic exposure and acquisition duration. However, the SC2000 did not originally have radiation force imaging capabilities; therefore, a new tool set for prototyping these sequences was developed along with rapid data processing and display code. These tools leveraged the increasing availability of general purpose computing on graphics processing units (GPUs) to significantly reduce the data processing time, facilitating real-time display for ultrasonic research systems.

These technical developments for both acquisition and processing were applied to investigate new methods for ARFI and SWEI imaging. Specifically, the power supply on the SC2000 allowed for a new type of multi-focal zone ARFI images to be acquired, which are shown to provide improved image quality over an extended depth of field. Additionally, a new algorithm for SWEI image processing was developed using an adaptive filter based on a maximum a posteriori estimator, demonstrating increases in the contrast to noise ratio of lesion targets upwards of 50%.

Finally, the optimized ARFI imaging methods were integrated with a transrectal ultrasound transducer to acquire volumetric in vivo data in patients undergoing robotic radical prostatectomy procedures in an ongoing study. When the study was initiated, it was recognized that the technological improvements of Siemens Acuson SC2000 allowed for the off-axis response to the radiation force excitation to be concurrently recorded without impacting ARFI image quality. This volumetric SWEI data was reconstructed retrospectively using the approaches developed in this thesis, but the images were low quality. A further investigation identified multiple challenges with the SWEI sequence, which should be addressed in future studies. The ARFI image volumes were very high quality and are currently being analyzed to assess the accuracy of ARFI to visualize prostate anatomy and clinically significant prostate cancer tumors. After a blinded evaluation of the ARFI image volumes for suspicion of prostate cancer, three readers correctly identified 63% of all clinically significant tumors and 74% of clinically significant tumors in the posterior region, showing great promise for using ARFI in the context of prostate cancer visualization for targeting biopsies, focal therapy, and watchful waiting.

Clutter is a problematic noise artifact in a variety of ultrasound applications. Clinical tasks complicated by the presence of clutter include detecting cancerous lesions in abdominal organs (e.g. livers, bladders) and visualizing endocardial borders to assess cardiovascular health. In this dissertation, an analytical expression for contrast loss due to clutter is derived, clutter is quantified in abdominal images, and sources of abdominal clutter are identified. Novel clutter reduction methods are also presented and tested in abdominal and cardiac images.

One of the novel clutter reduction methods is Short-Lag Spatial Coherence (SLSC) imaging. Instead of applying a conventional delay-and-sum beamformer to measure the amplitude of received echoes and form B-mode images, the spatial coherence of received echoes are measured to form SLSC images. The world's first SLSC images of simulated, phantom, and in vivo data are presented herein. They demonstrate reduced clutter and improved contrast, contrast-to-noise, and signal-to-noise ratios compared to conventional B-mode images. In addition, the resolution characteristics of SLSC images are quantified and compared to resolution in B-mode images.

A clinical study with 14 volunteers was conducted to demonstrate that SLSC imaging offers 19-33% improvement in the visualization of endocardial borders when the quality of B-mode images formed from the same echo data was poor. There were no statistically significant improvements in endocardial border visualization with SLSC imaging when the quality of matched B-mode images was medium to good.

Implemented on an intracardiac echocardiography transducer, acoustic radiation force methods may provide a useful means of characterizing the heart's elastic properties. Elasticity imaging may be of benefit for diagnosis and characterization of infarction and heart failure, as well as for guidance of ablation therapy for the treatment of arrhythmias. This thesis tests the hypothesis that with appropriately designed imaging sequences, intracardiac acoustic radiation force impulse (ARFI) imaging and shear wave elasticity imaging (SWEI) are viable tools for quantification of myocardial elasticity, both temporally and spatially. Multiple track location SWEI (MTL-SWEI) is used to show that, in healthy in vivo porcine ventricles, shear wave speeds follow the elasticity changes with contraction and relaxation of the myocardium, varying between 0.9 and 2.2 m/s in diastole and 2.6 and 5.1 m/s in systole. Infarcted tissue is less contractile following infarction, though not unilaterally stiffer. Single-track-location SWEI (STL-SWEI) is proven to provide suppression of speckle noise and enable improved resolution of structures smaller than 2 mm in diameter compared to ARFI and MTL-SWEI. Contrast to noise ratio and lateral edge resolution are shown to vary with selection of time step for ARFI and arrival time regression filter size for STL-SWEI and MTL-SWEI.

In 1.5 mm targets, STL-SWEI achieves alternately the tightest resolution (0.3 mm at CNR = 3.5 for a 0.17 mm filter) and highest CNR (8.5 with edge width = 0.7 mm for a 0.66 mm filter) of the modalities, followed by ARFI and then MTL-SWEI.

In larger, 6 mm targets, the CNR-resolution tradeoff curves for ARFI and STL-SWEI overlap for ARFI time steps up to 0.5 ms and kernels $\leq$1 mm for STL-SWEI. STL-SWEI can operate either with a 25 dB improvement over MTL-SWEI in CNR at the same resolution, or with edge widths 5$\times$ as narrow at equivalent CNR values, depending on the selection of regression filter size. Ex vivo ablations are used to demonstrate that ARFI, STL-SWEI and MTL-SWEI each resolve ablation lesions between 0.5 and 1 cm in diameter and gaps between lesions smaller than 5 mm in 3-D scans. Differences in contrast, noise, and resolution between the modalities are discussed. All three modalities are also shown to resolve ``x''-shaped ablations up to 22 mm in depth with good visual fidelity and correspondence to surface photographs, with STL-SWEI providing the highest quality images. Series of each type of image, registered using 3-D data from an electroanatomical mapping system, are used to build volumes that show ablations in in vivo canine atria. In vivo images are shown to be subject to increased noise due to tissue and transducer motion, and the challenges facing the proposed system are discussed. Ultimately, intracardiac acoustic radiation force methods are demonstrated to be promising tools for characterizing dynamic myocardial elasticity and imaging radiofrequency ablation lesions.

Hepatocellular carcinoma (HCC) lesions are often preceded by liver cirrhosis or Hepatitis C, and patients with these diagnoses are monitored every six months with an ultrasound screening. B-mode ultrasound is an ideal imaging modality for regular screening; however, ultrasound has demonstrated a low sensitivity for detecting small, early stage HCCs. Studies using ultrasonic elasticity methods have shown increased HCC lesion contrast compared to B-mode ultrasound. This thesis presents the preliminary work of shear wave elasticity imaging (SWEI) methods to improve estimates of viscoelastic parameters in the context of liver screening for tumors, with the goals of (1) using multi-dimensional directional filtering and shear wave reconstruction to reduce reflection artifacts, (2) evaluating bias introduced from small depth-of-field (DOF) excitations into frequency dependent shear wave speed (SWS) and attenuation estimates and (3) evaluating the feasibility of combining on-axis and off-axis elasticity methods to screen the entire liver.

Interfaces of different shear stiffness causes propagating shear waves to be reflected, which can lead to artifacts in SWS reconstructions due to the reflections both in and out of the imaging plane. Two-dimensional (2-D), three-dimensional (3-D), and four- dimensional (4-D) directional filters were applied to shear wave data, and SWS images were reconstructed with 2-D and 3-D shear wave reconstruction methods to quantify the reduction in image artifacts. For 2-D SWS image reconstructions, 3-D directional filters showed greater improvements in image quality than 2-D filters, and 4-D directional filters showed marginal improvement over 3-D filters. The 4-D directional filters have the largest impact in reducing reflection artifacts in 3-D SWS volumes.

Commercial scanners reconstruct shear wave speeds for a region of interest using time-of-flight (TOF) methods reporting a single SWS (or elastic modulus) to the end user under the assumptions that tissue is elastic and independent of frequency. Human tissues are known to be viscoelastic (VE), resulting in dispersion and attenuation. Existing methods to quantify shear wave dispersion and attenuation commonly make an assumption that the acoustic radiation force excitation acts as a cylindrical source with a known geometric shear wave amplitude decay. The bias in shear dispersion and attenuation estimates associated with making this cylindrical wave assumption (up to 15% for dispersion and 41% for attenuation) when applied to shear wave sources with finite depth extents in realistic focal geometries is greater for more tightly-focused acoustic radiation force sources with smaller DOF.

Curvilinear transducers are the standard probe used in ultrasound HCC screenings; however, previous studies using curvilinear arrays performing liver SWEI have been limited by penetration depth of the acoustic radiation force excitation. In order for SWEI to be feasible as an imaging method to screen for and detect HCC lesions, large, low frequency arrays designed for deep abdominal imaging must be used. A prototype low frequency deep abdominal curvilinear array and a proposed low frequency matrix array were simulated, and the combined dynamic on-axis response in the region of excitation (ROE) and the propagating shear wave (off-axis) response were used to create quantitative shear wave images of a spherical lesion using sparse acoustic radiation force excitations. The on-axis behavior in the ROE is related to the underlying stiffness of the material and a lookup table (LUT) approach can be used to determine a SWS in this region. Combined with traditional off-axis TOF methods, the combined methods can reduce the sparsity that would otherwise occur inside the ROE, which allows a larger field-of-view (FOV) to be interrogated with the same number of excitations. The on-axis and off-axis methods can be combined in either 2-D or 3-D reconstructions. The performance of the deep abdominal curvilinear array was comparable to the matrix array for 2-D SWEI imaging. A benefit of the curvilinear array over a large matrix array is its ability to image through intercostal acoustic windows. If there is not a sufficient subcostal acoustic window to use the matrix array to image the entire liver, a combination of both the matrix array and the curvilinear array can be used to scan the entire organ.

I investigate the generation and propagation characteristics of leaky Rayleigh waves (LRWs) by a spherical shock wave incident on a glass-water boundary both experimentally and numerically. The maximum tensile stress produced on the solid boundary was attributed to the dynamic interaction between the LRWs and an evanescent wave generated concomitantly along the boundary. The resultant tensile stress field drives the initiation of pre-existing microcracks and their subsequent extension along a circular trajectory, confirmative with the direction of the principal stress on the boundary. We further demonstrated that this unique ring-like fracture, prevalent in damage produced by high-speed impact, can be best described by the Tuler-Butcher criterion for dynamic brittle failure, and the orientation of the ring fracture extension into the solid also follows closely the trajectory of the local maximum tensile stress distribution.

Piezoelectric micromachined ultrasound transducer (pMUT) two-dimensional (2D) arrays have been proposed as an alternative to conventional bulk-PZT thickness-mode transducers for high frequency, forward-looking, catheter-based ultrasound imaging of the cardiovascular system. The appeal of pMUTs is based on several key advantages over conventional transducer technologies, including high operational frequencies, small element size, and low cost due to their microelectromechanical system (MEMS) silicon-based fabrication. While previous studies have demonstrated acoustic performance characteristics suitable for ultrasound image formation, pulse-echo B-mode imaging of tissue and tissue-like phantoms using 2D pMUT arrays small enough for forward-looking catheter-based applications have been demonstrated only at Duke University by Dausch et al.

Having demonstrated the suitability of 2D pMUT arrays for tissue imaging, an important step is to demonstrate effective design control. The frequency of operation is a fundamental component of transducer design. Previous modeling efforts for pMUT vibration have used classical/Kirchoff thin plate theory (CPT) or Mindlin thick plate theory, however pMUTs with geometric dimensions similar to those explored here, have not been modeled with experimental comparison to physical devices.

It is hypothesized that the frequency of vibration of pMUTs can be predictively modeled based on experimental data from various pMUT configurations. Experimental frequency results were acquired and used to develop an empirical model based on a modified Mindlin thick plate theory. This dissertation presents the development of the frequency design theory culminating in a set of predictive design equations for the frequency of vibration of 2D pMUT arrays aimed at improving their use in high-frequency, forward-looking, catheter-based ultrasound imaging applications.