Search

Many tissues, such as muscle, kidney, and breast, are mechanically anisotropic. Appropriately exploited, mechanical anisotropy can be a clinically relevant target for noninvasive imaging. An imaging method's potential for interrogating mechanical anisotropy can be experimentally evaluated using tissue-mimicking materials, also known as phantoms. The objective of this work is to demonstrate the feasibility of constructing mechanically anisotropic phantoms using 3D-printed polylactic acid (PLA) fibers embedded in gelatin hydrogel or polyvinyl alcohol (PVA) cryogel. Four identical fiber sets were printed; two were embedded in gelatin and two in PVA. Acoustic Radiation Force Impulse (ARFI) imaging was performed on the constructed phantoms, with data acquisitions at 0°, 30°, 60°, and 90° concentric orientations, where 0° and 90° corresponded to the long-axis of the spatially asymmetric ARF excitation being aligned across and along the fibers, respectively. Degree of anisotropy (DoA) was calculated as the ratio of peak displacements achieved at 90° versus 0° orientations. While both gelatin and PVA embedded fibers demonstrated elastic anisotropy, DoA values were 32% higher in gelatin. These pilot experimental results demonstrate that phantoms constructed of 3D-printed PLA fibers embedded in gelatin or PVA exhibit mechanical anisotropy as assessed by ARFI ultrasound.

Viscoelastic Response (VisR) ultrasound is a noninvasive method for interrogating the viscoelastic properties of tissue by observing, in the region of excitation, tissue displacement in response to successive acoustic radiation force impulses. Recent VisR technology advancements now enable estimation of mechanical anisotropy and texture as well as, via machine learning approaches, quantitative evaluation of elastic and viscous moduli (as in Quantitative VisR (QVisR)). The diagnostic relevance of VisR outcome measures have been demonstrated for monitoring renal transplant status, differentiating malignant from benign breast masses, and delineating dystrophic muscle degeneration. For the latter two applications, method and pilot clinical results are reviewed herein. VisR-derived relative elasticity (RE) and relative viscosity (RV) mechanical anisotropy measures differentiated malignant from benign breast masses in women with sensitivities of 0.94 and specificities of 1.00, with accuracies of 0.96. In dystrophic muscle, QVisR-derived longitudinal shear elastic modulus in rectus femoris, vastus lateralus, sartorius, and gastrocnemius muscles statistically different between dystrophic and control for ages less than six years (Wilcoxon, p

While SWEI methods require shear wave propagation to be sampled across a measurement kernel, the recently-introduced double-profile intersection (DoPIo) ultrasound can estimate shear moduli locally and on-axis by exploiting ARF-induced displacement underestimation caused by scatterer shearing under a tracking PSF. We herein perform a systematic comparison of DoPIo to MTL/STL-SWEI to identify the benefits and limitations of using multiple tracking F-numbers. We further compare DoPIo performance achieved using conventional normalized cross-correlation (NCC) to that using a blind source separation (BSS) estimator by assessing its impact of mechanical feature resolution and measurements via automatic segmentation. DoPIo shear modulus estimates in a simulated elastic phantom using BSS resulted in a contrast transfer efficiency comparable to STL-SWEI (0.81 for a 12.0 kPa feature within an 8.0 kPa background using BSS-DoPIo, compared to 0.77 for NCC, 0.76 for MTL-SWEI, and 0.81 for STL-SWEI). We also demonstrate DoPIo measurements in a calibrated phantomm using a clinical ultrasound scanner system. In doing so, we present DoPIo as a method to ultrasonically measure the shear modulus of tissue that may be performed on-axis independently of ARF amplitudes.

Double-profile intersection (DoPIo) ultrasound is a new acoustic radiation force (ARF) imaging technique that exploits scatterer shearing under two different tracking point spread functions to estimate shear elastic moduli. Previous versions of DoPIo used an empirically derived model to relate the time where two displacement profiles, each measuring displacement in identical positions using different aperture sizes. We herein propose an alternative, machine learning-based approach that obviates the need for an empirically-derived model by replacing it with a support vector machine that uses paired displacement profiles in their entirety. In this work, we describe the implementation of such a model, validate its ability in shear modulus for simulated materials, and deploy it to image a calibrated phantom.

We present a quantitative extension of Viscoelastic Response (VisR) ultrasound that estimates shear elastic and viscous moduli from on-axis VisR displacement profiles in silico. Isotropic, homogeneous, linearly viscoelastic materials ranging from 1.57-33.33 kPa shear elasticity and 0.0033-2.34 Pa.s shear viscosity subject to a VisR beamsequence at 26 focal depths were simulated. Multi-target regression machine learning models were used to estimate shear elasticity and shear viscosity given the displacement profile, focal depth, and axial depth. The best performing models achieve a shear elasticity RMSE of 0.29 kPa and a shear viscosity RMSE of 0.13 Pa.s when predictions were made on the test set. These results suggest that machine learning methods can be used to quantitatively estimate viscoelasticity from VisR displacement profiles.

Vascular hemodynamics and plaque composition are better indicators of stroke risk than degrees of stenosis alone. In silico work previously indicated that the acoustic radiation force impulse (ARFI) derived parameter, variance of acceleration (VoA), may be implemented with plane wave (PW) acquisitions to offer a high frame rate method of differentiating plaque components. Vector flow imaging methods such as multi-angle plane wave vector Doppler further allow 2D blood flow tracking and assessments of wall shear stress (WSS) to better predict rupture potential. We hypothesize ARFI VoA and plane wave vector Doppler may be efficiently combined to interrogate plaque structure and composition and estimate WSS in human carotid arteries. While focused ARFI - VoA best predicts feature size compared to PW acquisitions, PW ARFI-VoA offers comparable resolution, but contrast of lower SNR features is reduced. The PW ARFI-VoA sequence may be supplemented with a steered PW sequence to additionally visualize blood flow and estimate wall shear stress at potential real time frame rates. This work demonstrates the feasibility of the technique to an ex-vivo left carotid from a recently deceased 63-year-old female.

Delineating carotid plaque components that confer rupture risk is vital to stoke prevention. We have previously shown that combining temporal profiles of Acoustic Radiation Force Impulse (ARFI)-induced displacement, cross-correlation coefficient, and signal-to-noise ratio (SNR) as the input feature set to a machine learning classifier enabled the classifier to differentiate intraplaque hemorrhage (IPH), lipid-rich necrotic core (LRNC), collagen (COL), and calcium (CAL) plaque components, with accurate fibrous cap thickness measurement. We hypothesize that machine learning-based classification of carotid plaque structure and composition will be improved by including more information in the input feature set. This study analyzed six carotid plaques imaged in vivo in patients undergoing carotid endarterectomy CEA). Data were acquired with ECG gating to diastole and to systole, with and without ARF excitation. From these data, temporal profiles of displacement, crosscorrelation coefficient (CC), and signal-to-noise ratio (SNR) were generated and used as inputs to a support vector machine classifier. The classifier was trained and tested using spatially matched histology. Over all examined plaques, combining acquisitions at systole and diastole, with and without ARFI push, achieved CNRs that were statistically higher than any one or combination of two inputs. These results suggest that the performance of a machine learning classifier for automatic plaque feature delineation can be improved by optimizing the feature set, with clinical relevance to in vivo human carotid plaque feature delineation herein demonstrated.

Differentiating malignant from benign breast lesions remains a challenge that may be met by a new imaging biomarker: lesion-to-background mechanical anisotropy assessed by VisR ultrasound. VisR is an ultrasound imaging method that uses 2 consecutive acoustic radiation force excitations, separated in time and delivered to the same region of excitation, to directionally interrogate peak displacement (PD), relative elasticity (RE), and relative viscosity (RV). From these directional measurements, mechanical degree of anisotropy is measured. In a pilot clinical study involving 29 women with BIRADS-4 or −5 breast masses, VisR PD, RE, and RV were measured in lesions and in surrounding tissue at 0°, 30°, 60°, and 90° concentric orientations. From these measurements, lesion DoA (LDoA), surrounding tissue DoA (SDoA), and log(LDoA/SDoA) were assessed per parameter and compared between biopsy-confirmed malignant versus benign masses. Across all patients, PD-, RE-, and RV-based LDoA or SDoA alone did not achieve statistical difference between malignant and benign masses (Wilcoxon, $\mathrm{p} > 0.05$ ). However, PD-, RE-, and RV-based log(LDoA/SDoA) statistically differentiated malignant from benign masses (Wilcoxon, $\mathrm{p} ), with AUC values of 0.96 for PD-based, and 0.94 for RE- and RV-based outcomes. These results demonstrate the relevance of lesion-to-background mechanical anisotropy assessed by VisR ultrasound for differentiating malignant from benign breast masses in women, in vivo.

Stroke is often caused by carotid plaque rupture, with rupture potential indicated by plaque structure and composition. Delineating carotid plaque components that confer rupture risk is vital to stoke prevention. We have previously shown that ARFI Log Variance of Acceleration, log(VoA), discriminates intraplaque hemorrhage (IPH) from lipid-rich necrotic core (LRNC) and collagen (COL) from calcium (CAL) in vivo in humans, while ARFI peak displacement does not. We hypothesize that log(VoA) performance will be further improved by harmonic tracking of ARF-induced displacement. In a pilot clinical study of six carotid plaques imaged in vivo in patients undergoing carotid endarterectomy (CEA), plaque component contrast-to-noise ratio (CNR) achieved by harmonic log(VoA) was compared to that achieved by tracking at fundamental low (4.0 MHz, FL) and high (8.89 MHz, FH) frequencies with validation by spatially-matched histology derived from the extracted plaque specimens. Over all examined plaques, harmonic and both fundamental low and high log(VoA) distributions statistically differentiated plaque components (Wilcoxon, $\mathrm{p} , panel iv). Harmonic log(VoA) achieved the highest CNR between COL and LRNC, which is relevant to fibrous cap delineation, but FH tracking yielded the highest CNR between LRNC and IPH and between COL and CAL. These results suggest that combined harmonic and FH tracking could improve ARFI log(VoA) characterization of carotid plaque components that are correlated to rupture.

ARFI Variance of Acceleration (VoA) has been previously used in clinical applications to delineate atherosclerotic plaque morphology and detect subcutaneous hemorrhage. In silico models also demonstrated that behavior of the parameter is influenced by error in the displacement estimates, which are accentuated by decreasing SNR or correlation. This study experimentally demonstrates that in non-elastic materials, decreasing correlation, induced by a greater magnitude ARFI push, increases VoA when the material SNR is above 40–45 dB. This signal correlation dependent change is realizable with focused and PW tracking, potentially expanding the clinical utility of high framerate VoA analysis.