Browsing by Subject "Ultrasound Imaging"

In this thesis, we develop methods for addressing the deficiencies of array processing with linear thinned arrays. Our methods are designed for array systems mounted on moving platforms and exploit synthetic aperture processing techniques. In particular, we use array motion to decrease the sidelobe levels and increase the degrees of freedom available from thinned array systems. In this work, we consider two application areas 1) passive SONAR and 2) ultrasound imaging.

Synthetic aperture processing is a methodology for exploiting array motion and has been successfully used in practice to increase array resolution. By spatially sampling along the path of the array virtual sensors can be realized and coherently fused to the existing array. The novel contribution of this work is our application of synthetic aperture processing. Here our goal is not to increase array resolution, instead we propose to use the synthetic aperture process to expand the spatial covariance and spatial frequency sensing capabilities of thinned array system.

In the passive sensing case, we use a class of thinned arrays know as co-prime linear sensor arrays for source localization. The class of co-prime arrays provides roughly half the aperture worth of spatial covariances and with modest array motion can be extended to the full aperture of the array. The amount of motion required to produce a full set of spatial covariances is shown to be a function of the co-prime array parameters and is only a fraction of the total aperture of the array. The full set of spatial covariances can be used to form a spatial covariance matrix with dimension equal to that of a uniform array. With a spatial covariance matrix in hand one can perform signal processing tasks as if the array were fully populated. Three methods for spatial covariance matrix estimation are compared in different source localization scenarios. In the work presented here, we demonstrate the benefits of our approach for achieving reduced sidelobe levels and extending the source localization capabilities above the limits of the static co-prime array.

In the active sensing case, we develop a framework for incorporating motion using thinned arrays for ultrasound imaging. In this setting, array motion is used to augment the spatial frequency sensing capabilities of the thinned array system. Here we develop an augmentation strategy based on using quarter-wavelength array translations to fill-in missing spatial frequencies not measured by the static thinned array. The quarter-wavelength translation enables the thinned array system to sample missing spatial frequencies and increase the redundancy of other spatial frequencies sampled by the array. We compare the level of redundancy in sampling the spatial frequencies achieved by the thinned arrays post translation to different levels of sample redundancy derived from pruning the transmit/receive events of a uniform array. In this manner, we are able to examine how the level of spatial frequency redundancy afforded by different thinned arrays compare over the full redundancy range of the uniform array. While artificially pruning the uniform array does not necessarily create realizable arrays, it provides the means to compare image quality at different spatial frequency redundancy levels. In this work, we are able to conclude that images formed from thinned arrays using the translated synthetic aperture process are capable of approximating images formed from the corresponding uniform array. In particular, the systems considered in this work have approximately one-third of the active sensors when compared to the uniform array.

In both application areas, the use of thinned arrays offers a reduction in the cost to deploy and maintain a given array system. The feature that makes it possible to overcome the spatial sampling deficiencies of thinned array systems is motion and it is at the core of the performance gains in these applications.

Heart failure is one of the most common cardiac disorders and is projected to increase in prevalence over the next few decades. It can arise from a wide variety of root causes such as coronary artery disease, hypertension, cardiomyopathy, or cardiotoxicity and can manifest as systolic and/or diastolic dysfunction. Traditionally, its diagnosis has been based on monitoring qualitative changes in cardiac structure, such as chamber geometry and wall motion patterns, or quantitative changes in indices of cardiac function, such as the blood flow velocities and ventricular ejection fraction. These parameters are assessed in clinical settings using medical imaging modalities like ultrasound and magnetic resonance imaging. Recent research into cardiac pathophysiology has indicated that the progression of cardiac disease is often accompanied by changes in the mechanical properties of cardiac muscle. Interrogation of these changes could be used to gain useful diagnostic insight into the etiology of heart failure.

Acoustic radiation force (ARF)-based techniques, such as acoustic radiation force impulse (ARFI) imaging and shear wave elasticity imaging (SWEI), provide the means to measure mechanical properties of soft tissues using ultrasound. They operate on the principle that ultrasound can be used to remotely generate as well as track micron-level vibrations in the body and thus derive mechanical properties such as tissue stiffness. ARFI and SWEI have previously been shown to capture dynamic changes in myocardial stiffness in Langendorff set-ups, open-chest experiments, and intracardiac settings. This dissertation explores the challenges and opportunities of implementing acoustic radiation force-based methods for noninvasive applications via transthoracic imaging windows.

Transthoracic imaging of the heart using ultrasound can be challenging for a number of reasons. The two main sources of signal degradation that were hypothesized to impact ARFI and SWEI in this environment are acoustic clutter and intrinsic tissue motion. Acoustic clutter refers to incorrectly localized echoes which lead to the degradation of target conspicuity, border delineation, and image quality. Intrinsic tissue motion, on the other hand, impedes the ability to accurately measure the ARF-induced motion and consequently affects the estimation of tissue stiffness. The work presented herein focuses on quantifying the level of both sources of signal degradation under \textit{in vivo} imaging conditions and evaluating the effectiveness of strategies to minimize their impact. Lastly, the feasibility of tracking dynamic myocardial stiffness through the cardiac cycle via transthoracic imaging windows on human volunteers was investigated.

Harmonic imaging is often used to suppress acoustic clutter in clinical settings. Clutter levels are also closely tied to the choice of beamforming configuration used. Quantifying the impact of harmonic imaging and transmit beamforming (focused versus plane wave) on acoustic clutter, under \textit{in vivo} transthoracic imaging conditions is therefore important. Clutter level, for a given imaging scenario, was quantified using contrast between the cardiac chambers and the interventricular septum. Substantial variations in clutter levels were observed across as well as within volunteers. Harmonic imaging had a measurable impact in suppressing clutter under both the plane wave (2.97$\,$dB) and focused (6.1$\,$dB) configurations. However, even in the optimal configuration (harmonic-focused), clutter levels varied over a broad range (4$\,$-$\,$22$\,$dB). These results suggest that acoustic clutter, while consistently lowered through the use of harmonic imaging, is still likely to be a major detriment to transthoracic measurement of myocardial stiffness.

The heart exhibits complex and rapid three-dimensional motion; this could be a dominant confounder when attempting to measure micron-level ARF-induced displacements. Intrinsic cardiac motion of the interventricular septum, as observed through the parasternal long- and short-axis views, was analyzed in both the time- and frequency-domain. Two types of motion filters, frequency-based (high-pass filters) and recovery-based (polynomial filters) were compared to assess their ability to separate the axial component of cardiac motion from the ARF-induced motion. The effect of non-axial cardiac motion on speckle decorrelation was quantified using temporal coherence and related to the uncertainty of axial displacement estimation or jitter. High-pass filters with cutoffs $>$75$\,$Hz and quadratic polynomial filters were found to be equally effective at compensating for axial tissue motion. While high-pass filters are independent of a recovery-time assumption, they introduce a downward bias to measured ARF-induced motion; this bias increases with cutoff frequency. Temporal coherence was empirically related to measured displacement estimation jitter. At end-diastole, temporal coherence was high and jitter was low (0.5$\,$-$\,$2.5$\,\mu$m). In other phases of the cardiac cycle, however, jitter was found to increase dramatically with the span of the temporal window over which it was computed. Jitter for short spans, 2$\,$ms, was found to be in the range of 2$\,$-$\,$8$\,\mu$m, However, for spans of 10$\,$ms, it could be as high as 10$\,$-$\,$20$\,\mu$m. These results indicate that the noise-floor for micron-level axial displacement estimation in the myocardium via transthoracic imaging windows can be fairly high (compared to the magnitude of ARF-induced displacements) and can vary considerably over the cardiac cycle.

In the final study, M-mode ARFI imaging was performed on twelve healthy volunteers to track stiffness changes within the interventricular septum in the parasternal long- and short-axis views. Myocardial stiffness dynamics over the cardiac cycle were quantified using five indices: stiffness ratio, rates of relaxation and contraction, and time constants of relaxation and contraction. Yield of ARFI acquisitions was evaluated based on metrics of signal strength and tracking fidelity such as displacement signal-to-noise, signal-to-clutter level, temporal coherence of speckle, and spatial similarity within the region-of-excitation. These were quantified using the mean ARF-induced displacements over the cardiac cycle, the contrast between the myocardium and the cardiac chambers, the minimum correlation coefficients of RF signals (over a 2$\,$ms window), and the correlation between displacement traces across simultaneously-acquired azimuthal beams, respectively. Forty-one percent of ARFI acquisitions were determined to be \say{successful} using a mean ARF-induced displacement threshold of 1.5 $\mu$m. \say{Successful} acquisitions were found to have higher i) signal-to-clutter levels, ii) temporal coherence, and iii) spatial similarity compared to \say{unsuccessful} acquisitions. Median values of these three metrics, between the two groups, were measured to be 13.42$\,$dB vs. 5.42$\,$dB, 0.988 vs. 0.976, and 0.984 vs. 0.849, respectively. Signal-to-clutter level, temporal coherence, and spatial similarity were also found to correlate with each other. Across the cohort of healthy volunteers, stiffness ratio was measured to be 2.74$\,\pm\,$0.86; rate of relaxation was 7.82$\,\pm\,$4.69$\,$/s and contraction was -7.31$\,\pm\,$3.79$\,$/s; time constant of relaxation was 35.90$\,\pm\,$20.04$\,$ms, and contraction was 37.24$\,\pm\,$19.85$\,$ms. ARFI-derived indices of myocardial stiffness were found to be similar in both views.

In summary, despite the many challenges that are inherent to the transthoracic imaging environment, acoustic radiation force-based techniques were found to capture the dynamic trends of myocardial stiffness when appropriate conditions are met. Future work to improve the strength of ARF-excitations, better characterize or circumvent the influence of noise sources such as acoustic clutter and tissue motion, and explore the association between ARFI/SWEI-derived myocardial stiffness and traditional indices of cardiac function will be critical to realizing the diagnostic potential of acoustic radiation force-based ultrasound methods in clinical cardiology.