Elasticity imaging using conventional and high-frame rate ultrasound imaging: experimental study

High-frame rate ultrasound imaging is necessary to track fast deformation in ultrasound elasticity imaging, but the image quality may be degraded. Previously, we investigated the performance of strain imaging using numerical models of conventional and ultrafast ultrasound imaging techniques. In this paper, we performed experimental studies to quantitatively evaluate the strain images and elasticity maps obtained using conventional and high frame rate ultrasound imaging methods. The experiments were carried out using point target and tissue mimicking phantoms. The experimental results were compared with the results of numerical simulation. Our experimental studies confirm that the signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR), and axial/lateral resolution of the displacement and strain images acquired using high-frame rate ultrasound imaging are slightly lower but comparable with those obtained using conventional imaging. Furthermore, the quality of elasticity images also exhibits similar trends. Thus, high-frame rate ultrasound imaging can be used reliably for static elasticity imaging to capture the internal tissue motion if the frame rate is critical.     

Strain imaging using conventional and ultrafast ultrasound imaging: numerical analysis

In elasticity imaging, the ultrasound frames acquired during tissue deformation are analyzed to estimate the internal displacements and strains. If the deformation rate is high, high-frame-rate imaging techniques are required to avoid the severe decorrelation between the neighboring ultrasound images. In these high-frame-rate techniques, however, the broader and less focused ultrasound beam is transmitted and, hence, the image quality is degraded. We quantitatively compared strain images obtained using conventional and ultrafast ultrasound imaging methods. The performance of the elasticity imaging was evaluated using custom-designed, numerical simulations. Our results demonstrate that signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR) and spatial resolutions in displacement and strain images acquired using conventional and ultrafast ultrasound imaging are comparable. This study suggests that the high-frame-rate ultrasound imaging can be reliably used in elasticity imaging if frame rate is critical     

An autocorrelation-based method for improvement of sub-pixel displacement estimation in ultrasound strain imaging

In ultrasound strain and elasticity imaging, an accurate and cost-effective sub-pixel displacement estimator is required because strain/elasticity imaging quality relies on the displacement SNR, which can often be higher if more computational resources are provided. In this paper, we introduce an autocorrelation-based method to cost-effectively improve subpixel displacement estimation quality. To quantitatively evaluate the performance of the autocorrelation method, simulated and tissue-mimicking phantom experiments were performed. The computational cost of the autocorrelation method is also discussed. The results of our study suggest the autocorrelation method can be used for a real-time elasticity imaging system.     

The combined effect of signal decorrelation and random noise on the variance of time delay estimation

Tissue motion and elasticity imaging techniques commonly use time delay estimation (TDE) for the assessment of tissue displacement. The performance of these techniques is limited because the signals are corrupted by various factors including electronic noise, quantization, and speckle decorrelation. Speckle decorrelation is caused by changes in the coherent interference among scatterers when the tissue moves relative to the ultrasound beam. In time delay estimation, the effect of noise is usually addressed through the signal-to-noise ratio (SNR) term. Decorrelation, often a significant source of error in medical ultrasound, is commonly described in terms of the correlation coefficient. A relationship between the correlation coefficient and the SNR was previously derived in the literature, for identical signals corrupted by uncorrelated random noise. In this paper, we derive the relationship between the peak of the correlation coefficient function and the SNR for two jointly stationary signals when a delay is present between the signals. Recently, an expression for the Cramer-Rao lower bound (CRLB) has been derived in the literature for partially decorrelated signals in terms of the SNR and the correlation coefficient. Since the applicability of the CRLB is determined not only by the SNR, but also by the correlation coefficient, it is important to unify the expression for the CRLB for partially correlated signals. In this paper, we derive an expression for the CRLB in term of an equivalent SNR converted from the correlation coefficient using an SNR-p relationship, and show this expression to be equivalent to the expression for CRLB. We also corroborate the validity of the SNR-p expression with a simulation. Using this formulation, correlation measurements can be converted to SNR to obtain a composite SNR. The use of this composite SNR in lieu of those in the CRLB expression in the literature allows the extension of the literature results to the solution of the common TDE problems that involve signal decorrelation.     

Uniform precision ultrasound strain imaging

Ultrasound strain imaging is becoming increasingly popular as a way to measure stiffness variation in soft tissue. Almost all techniques involve the estimation of a field of relative displacements between measurements of tissue undergoing different deformations. These estimates are often high resolution, but some form of smoothing is required to increase the precision, either by direct filtering or as part of the gradient estimation process. Such methods generate uniform resolution images, but strain quality typically varies considerably within each image, hence a trade-off is necessary between increasing precision in the low-quality regions and reducing resolution in the high-quality regions. We introduce a smoothing technique, developed from the nonparametric regression literature, which can avoid this trade-off by generating uniform precision images. In such an image, high resolution is retained in areas of high strain quality but sacrificed for the sake of increased precision in low-quality areas. We contrast the algorithm with other methods on simulated, phantom, and clinical data, for both 2-D and 3-D strain imaging. We also show how the technique can be efficiently implemented at real-time rates with realistic parameters on modest hardware. Uniform precision nonparametric regression promises to be a useful tool in ultrasound strain imaging.     

A theoretical framework for performance characterization of elastography: the strain filter

This paper presents a theoretical framework for performance characterization in strain estimation, which includes the effect of signal decorrelation, quantization errors due to the finite temporal sampling rate, and electronic noise. An upper bound on the performance of the strain estimator in elastography is obtained from a strain filter constructed using these limits. The strain filter is a term used to describe the nonlinear filtering process in the strain domain (due to the ultrasound system and signal processing parameters) that allows the elastographic depiction of a limited range of strains from the compressed tissue. The strain filter predicts the elastogram quality by specifying the elastographic signal-to-noise ratio (SNR(e)), sensitivity, and the strain dynamic range at a given resolution. The dynamic range is limited by decorrelation errors for large tissue strain values, and electronic noise for low strain values. Tradeoffs between different techniques used to enhance elastogram image quality may also be analyzed using the strain filter.     

Coded excitation system for improving the penetration of real-time phased-array imaging systems

Based on an analysis of the inherent signal-to-noise ratio (SNR) in medical ultrasound imaging, SNR improvements of 15-20 dB are theoretically possible for real-time phased-array imagers using coded excitation. A very simple coded excitation for phased arrays based on the principles of ;pseudochirp' excitation and equalization filtering is described. This system is capable of SNR improvements of about 15 dB with range sidelobe levels acceptable for many medical imaging applications. Such improvements permit increased operating frequencies, and hence enhanced spatial resolution, for real-time array imagers. Both simulations and measurements are used to demonstrate the efficacy of the method.     

An elasticity microscope. Part I: Methods

An elasticity microscope images tissue stiffness at fine resolution. Possible applications include dermatology, ophthalmology, pathology, and tissue engineering. In addition, if the resolution approaches cellular dimensions, then this system may be very useful in understanding tissue micromorphology. Elasticity images can be reconstructed from displacement and strain fields measured throughout the specimen during controlled external loading. High frequency ultrasound is used to obtain these images by tracking coherent speckle motion during deformation. In this paper, methods are presented to track speckle in two dimensions with near unity correlation coefficients using a high frequency, single element focused transducer. These techniques include improved means for speckle tracking. Procedures to control boundary conditions for consistent specimen deformation and scanning techniques required to obtain a plane-strain state in the imaging plane are also discussed. To test these methods, a 50 MHz elasticity microscope was constructed     

Phase rotation methods in filtering correlation coefficients for ultrasound speckle tracking

In speckle-tracking-based myocardial strain imaging, large interframe/volume peak-systolic strains cause peak hopping artifacts separating the highest correlation coefficient peak from the true peak. A correlation coefficient filter was previously designed to minimize peak hopping artifacts. For large strains, however, the correlation coefficient filter must follow the strain distribution to remove peak hopping effectively. This processing usually means interpolation and high computational load. To reduce the computational burden, a narrow band approximation using phase rotation is developed in this paper to facilitate correlation coefficient filtering. Correlation coefficients are first phase rotated to increase coherence, then filtered. Rotated phase angles are determined by the local strain and spatial position. This form of correlation coefficient filtering enhances true correlation coefficient peaks in large strain applications if decorrelation due to deformation does not completely destroy the coherence among neighboring correlation coefficients. The assumed strain used in the filter can also deviate from the true strain and still be effective. Further improvement in displacement estimation can be expected by combining correlation coefficient filtering with a new Viterbi-based displacement estimator.     

Analysis of correlation coefficient filtering in elasticity imaging

Correlation-based speckle tracking methods are commonly used in elasticity imaging to estimate displacements. In the presence of local strain, a larger window size results in larger displacement error. To reduce tracking error, we proposed a short correlation window followed by a correlation coefficient filter. Although simulation and experimental results demonstrated the efficacy of the method, it was not clear why correlation coefficient filtering reduces tracking error since tracking error increases if normalization before filtering is not applied. In this paper, we analyzed tracking errors by estimating phase variances of the cross-correlation function and the correlation coefficient at the true time lag based on statistical properties of these functions' real and imaginary parts. The role of normalization is clarified by identifying the effect of the cross-correlation function's amplitude fluctuation on the function's imaginary part. Furthermore, we present analytic forms for predicting axial displacement error as a function of strain, system parameters (signal-to-noise ratio, center frequency, and signal and noise bandwidths), and tracking parameters (window and filter sizes) for cases with and without normalization before filtering. Simulation results correspond to theory well for both noise-free cases and general cases with an empirical correction term included for strains up to 4%.     

A 100-MHz ultrasound imaging system for dermatologic and ophthalmologic diagnostics

A major design problem concerning high-frequency broad-band ultrasound imaging systems is caused by the strong dispersive attenuation of the tissue, which gives rise to images with inhomogeneous resolution and poor signal to noise ratio (SNR). To address the noise problem, strongly focused transducers with high energy density in a narrow focal region are utilized, which also provide more isotropic images due to improved lateral resolution. To account for the short depth of the focal area two suitable imaging conceptions are used: 1) synthetic aperture concept and 2) B/D-scan concept. To avoid the inhomogeneity of the images, different transmitter signals for each depth are applied, which are pseudoinversely prefiltered according to the transfer function of the tissue. To gain signal energy required for inverse filtering, a pulse compression technique with nonlinearly frequency modulated chirp signals is utilized. These procedures have been implemented in an ultrasound imaging system, which has been developed in the authors' laboratory for eye and skin examinations, It can be used with transducers in a frequency range from 20 to 250 MHz.     

Multiresolution imaging in elastography

The range of strains that can be imaged by any practical elastographic imaging system is inherently limited, and a performance measure is valuable to evaluate these systems from the signal and noise properties of their output images. Such a measure was previously formulated for systems employing cross-correlation based time-delay estimators through the strain filter. While the strain filter predicts the signal-to-noise ratio (SNR(e)) for each tissue strain in the elastogram and provides valuable insights into the nature of image noise, it understated the effects of image resolution (axial resolution, as determined by the cross-correlation window length) on the noise. In this work, the strain filter is modified to study the strain noise at multiple resolutions. The effects of finite window length on signal decorrelation and on the variance of the strain estimator are investigated. Long-duration windows are preferred for improved sensitivity, dynamic range, and SNR(e). However, in this limit the elastogram is degraded due to poor resolution. The results indicate that for nonzero strain, a window length exists at which the variance of strain estimator attains its minima, and consequently the elastographic sensitivity, dynamic range and SNR(e) are strongly affected by the selected window length. Simulation results corroborate the theoretical results, illustrating the presence of a window length where the strain estimation variance is minimized for a given strain value. Multiresolution elastography, where the strain estimate with the highest SNR(e) obtained by processing the pre- and post-compression waveforms at different window lengths is used to generate a composite elastogram and is proposed to improve elastograms. All the objective elastogram parameters (namely: SNR(e), dynamic range, sensitivity and the average elastographic resolution-defined as the cross-correlation window length) are improved with multiresolution elastography when compared to the traditional method of utilizing a single window length to generate the elastogram. Experimental results using a phantom with a hard inclusion illustrates the improvement in elastogram obtained using multiresolution analysis.     

Monitoring of thermal therapy based on shear modulus changes: II. Shear wave imaging of thermal lesions

The clinical applicability of high-intensity focused ultrasound (HIFU) for noninvasive therapy is currently hampered by the lack of robust and real-time monitoring of tissue damage during treatment. The goal of this study is to show that the estimation of local tissue elasticity from shear wave imaging (SWI) can lead to a precise mapping of the lesion. HIFU treatment and monitoring were respectively performed using a confocal setup consisting of a 2.5-MHz single element transducer focused at 34 mm on ex vivo samples and an 8-MHz ultrasound diagnostic probe. Ultrasound-based strain imaging was combined with shear wave imaging on the same device. The SWI sequences consisted of 2 successive shear waves induced at different lateral positions. Each wave was created with pushing beams of 100 μs at 3 depths. The shear wave propagation was acquired at 17,000 frames/s, from which the elasticity map was recovered. HIFU sonications were interleaved with fast imaging acquisitions, allowing a duty cycle of more than 90%. Thus, elasticity and strain mapping was achieved every 3 s, leading to real-time monitoring of the treatment. When thermal damage occurs, tissue stiffness was found to increase up to 4-fold and strain imaging showed strong shrinkages that blur the temperature information. We show that strain imaging elastograms are not easy to interpret for accurate lesion characterization, but SWI provides a quantitative mapping of the thermal lesion. Moreover, the concept of shear wave thermometry (SWT) developed in the companion paper allows mapping temperature with the same method. Combined SWT and shear wave imaging can map the lesion stiffening and temperature outside the lesion, which could be used to predict the eventual lesion growth by thermal dose calculation. Finally, SWI is shown to be robust to motion and reliable in vivo on sheep muscle.     

A comparison of the performance of time-delay estimators in medical ultrasound

Time-delay estimation (TDE) is a common operation in ultrasound signal processing. In applications such as blood flow estimation, elastography, phase aberration correction, and many more, the quality of final results is heavily dependent upon the performance of the time-delay estimator implemented. In the past years, several algorithms have been developed and applied in medical ultrasound, sonar, radar, and other fields. In this paper we analyze the performances of the widely used normalized and non-normalized correlations, along with normalized covariance, sum absolute differences (SAD), sum squared differences (SSD), hybrid-sign correlation, polarity-coincidence correlation, and the Meyr-Spies method. These techniques have been applied to simulated ultrasound radio frequency (RF) data under a variety of conditions. We show how parameters, which include center frequency, fractional bandwidth, kernel window size, signal decorrelation, and signal-to-noise ratio (SNR) affect the quality of the delay estimate. Simulation results also are compared with a theoretical performance limit set by the Cramer-Rao lower bound (CRLB). Results show that, for high SNR, high signal correlation, and large kernel size, all of the algorithms closely match the theoretical bound, with relative performances that vary by as much as 20%. As conditions degrade, the performances of various algorithms differ more significantly. For signals with a correlation level of 0.98, SNR of 30 dB, center frequency of 5 MHz with a fractional bandwidth of 0.5, and kernel size of 2 /spl mu/s, the standard deviation of the jitter error is on the order of few nanoseconds. Normalized correlation, normalized covariance, and SSD have an approximately equal jitter error of 2.23 ns (the value predicted by the CRLB is 2.073 ns), whereas the polarity-coincidence correlation performs less well with a jitter error of 2.74 ns.     

Complex principal components for robust motion estimation

Bias and variance errors in motion estimation result from electronic noise, decorrelation, aliasing, and inherent algorithm limitations. Unlike most error sources, decorrelation is coherent over time and has the same power spectrum as the signal. Thus, reducing decorrelation is impossible through frequency domain filtering or simple averaging and must be achieved through other methods. In this paper, we present a novel motion estimator, termed the principal component displacement estimator (PCDE), which takes advantage of the signal separation capabilities of principal component analysis (PCA) to reject decorrelation and noise. Furthermore, PCDE only requires the computation of a single principal component, enabling computational speed that is on the same order of magnitude or faster than the commonly used Loupas algorithm. Unlike prior PCA strategies, PCDE uses complex data to generate motion estimates using only a single principal component. The use of complex echo data is critical because it allows for separation of signal components based on motion, which is revealed through phase changes of the complex principal components. PCDE operates on the assumption that the signal component of interest is also the most energetic component in an ensemble of echo data. This assumption holds in most clinical ultrasound environments. However, in environments where electronic noise SNR is less than 0 dB or in blood flow data for which the wall signal dominates the signal from blood flow, the calculation of more than one PC is required to obtain the signal of interest. We simulated synthetic ultrasound data to assess the performance of PCDE over a wide range of imaging conditions and in the presence of decorrelation and additive noise. Under typical ultrasonic elasticity imaging conditions (0.98 signal correlation, 25 dB SNR, 1 sample shift), PCDE decreased estimation bias by more than 10% and standard deviation by more than 30% compared with the Loupas method and normalized cross-correlation with cosine fitting (NC CF). More modest gains were observed relative to spline-based time delay estimation (sTDE). PCDE was also tested on experimental elastography data. Compressions of approximately 1.5% were applied to a CIRS elastography phantom with embedded 10.4-mm-diameter lesions that had moduli contrasts of -9.2, -5.9, and 12.0 dB. The standard deviation of displacement estimates was reduced by at least 67% in homogeneous regions at 35 to 40 mm in depth with respect to estimates produced by Loupas, NC CF, and sTDE. Greater improvements in CNR and displacement standard deviation were observed at larger depths where speckle decorrelation and other noise sources were more significant.     

Two-dimensional temperature estimation using diagnostic ultrasound

A two-dimensional temperature estimation method was developed based on the detection of shifts in echo location of backscattered ultrasound from a region of tissue undergoing thermal therapy. The echo shifts are due to the combination of the local temperature dependence of speed of sound and thermal expansion in the heated region. A linear relationship between these shifts and the underlying tissue temperature rise is derived from first principles and experimentally validated. The echo shifts are estimated from the correlation of successive backscattered ultrasound frames, and the axial derivative of the accumulated echo shifts is shown to be proportional to the temperature rise. Sharp lateral gradients in the temperature distribution introduce ripple on the estimates of the echo shifts due to a thermo-acoustic lens effect. This ripple can be effectively reduced by filtering the echo shifts along the axial and lateral directions upon differentiation. However, this is achieved at the expense of spatial resolution. Experimental evaluation of the accuracy (0.5 degrees C) and spatial resolution (2 mm) of the algorithm in tissue mimicking phantoms was conducted using a diagnostic ultrasound imaging scanner and a therapeutic ultrasound unit. The estimated temperature maps were overlaid on the gray-scale ultrasound images to illustrate the applicability of this technique for image guidance of focused ultrasound thermal therapy.     

Coded pulse excitation for ultrasonic strain imaging

Decorrelation strain noise can be significantly reduced in low echo-signal-to-noise (eSNR) conditions using coded excitation. Large time-bandwidth-product (>30) pulses are transmitted into tissue mimicking phantoms with 2.5-mm diameter inclusions that mimic the elastic properties of breast lesions. We observed a 5-10 dB improvement in eSNR that led to a doubling of the depth of focus for strain images with no reduction of spatial resolution. In high eSNR conditions, coded excitation permits the use of higher carrier frequencies and shorter correlation windows to improve the attainable spatial resolution for strain relative to that obtained with conventional short pulses. This paper summarizes comparative studies of strain imaging in noise-limited conditions obtained by short pulses and four common aperiodic codes (chirp, Barker, suboptimal, and Golay) as a function of attenuation, eSNR and applied strain. Imaging performance is quantified using SNR for displacement (SNRd), local modulation transfer function (LMTF), and contrast-to-noise ratio for strain (CNRepsilon). We found that chirp and Golay codes are the most robust for imaging soft tissue deformation using matched filter decoding. Their superior performance is obtained by balancing the need for low-range lobes, large eSNR improvement, and short-code duration.     

Comparison of 2-D speckle tracking and tissue Doppler imaging in an isolated rabbit heart model

Ultrasound strain imaging has been proposed to quantitatively assess myocardial contractility. Cross-correlation-based 2-D speckle tracking (ST) and auto-correlation-based tissue Doppler imaging (TDI) [often called Doppler tissue imaging (DTI)] are competitive ultrasound techniques for this application. Compared with 2-D ST, TDI, as a 1-D method, is sensitive to beam angle and suffers from low strain signal-to-noise ratio because a high pulse repetition frequency is required to avoid aliasing in velocity estimation. In addition, ST and TDI are fundamentally different in the way that physical parameters such as the mechanical strain are derived, resulting in different estimation accuracy and interpretation. In this study, we directly compared the accuracy of TDI and 2-D ST estimates of instantaneous axial normal strain and accumulated axial normal strain using a simulated heart. We then used an isolated rabbit heart model of acute ischemia produced by left descending anterior artery ligation to evaluate the performance of the two methods in detecting abnormal motion. Results showed that instantaneous axial normal strains derived using TDI (0.36% error) were less accurate with larger variance than those derived from 2-D ST (0.08% error) given the same spatial resolution. In addition to poorer accuracy, accumulated axial normal strain estimates derived using TDI suffered from bias, because the accumulation method for TDI cannot trace along the actual tissue displacement path. Finally, we demonstrated the advantage 2-D ST has over TDI to reduce dependency on beam angle for lesion detection by estimating strains based on the principal stretches and their corresponding principal axes.     

Elasticity reconstruction from displacement and confidence measures of a multi-compressed ultrasound RF sequence

Ultrasound elasticity imaging shows promise as a new way for early detection of cancers by assessing the elastic characteristics of soft tissue. So far the commonly used approach involves solving the so-called inverse elasticity problem of recovering elastic parameters from displacement measurements. We propose a finite-elementbased nonlinear scheme to estimate the elasticity distribution of soft tissue from multi-compressed ultrasound radio frequency (RF) data. An experimental ultrasound workstation has been developed to acquire multi-compressed data. A composite probe was employed as the compression plate. The contact forces and torques were acquired at the same time as imaging. Axial displacements under different static loads are estimated from the RF data before and after deformation using a cross-correlation technique. The confidence of displacement estimates is employed as a weighting factor in solving the objective function describing the inverse elasticity reconstruction problem. A novel splitand- merge strategy is employed over the image sequence in which strain images are used to provide a priori knowledge of the relative stiffness distribution of the tissue to constrain the inverse problem solution. The experimental study has allowed us to investigate the performance of our approach in the controlled environment of simulated and phantom data. For a simulated single inclusion model with 5% axial displacement estimation error, the L2-error between the target and the reconstructed Young's modulus was found to be about 1%. In vivo validation of the proposed method has been carried out and some preliminary results are presented.     

Linear approach to axial resolution in elasticity imaging

Thus far axial resolution in elasticity imaging has been addressed only empirically. No clear analytical approaches have emerged because the estimator is non-linear in the data, correlation functions are nonstationary, and system responses vary spatially. This paper describes a linear systems approach based on a small-strain impulse approximation that results in the derivation of a local impulse response (LIR) and local modulation transfer function (LMTF). Closed-form solutions for strain LIR are available to provide new insights on the role of instrumentation and processing on axial strain resolution. Novel phantom measurements are generated to validate results. We found that the correlation window determines axial resolution in most practical situations, but that the the same system properties that determine B-mode resolution ultimately limit elasticity imaging.