American Institute of Ultrasound in Medicine <strong>Proceedings</strong> J Ultrasound Med 32(suppl):S1–S134, 2013SCIENTIFIC SESSIONSMONDAY, APRIL 8, 2013, 11:00 AM–12:30 PMBasic Science: Tissue Characterization, Part 1Moderators: Keith Wear, PhD, Mark Holland, PhD1539806 A Feasibility Study of Ultrasound Strain Imaging for RiskAssessment of Carotid Atherosclerotic Plaques Validatedby Magnetic Resonance ImagingXiaochang Pan, 1 Shengzhen Tao, 1 Lingyun Huang, 2 ManweiHuang, 3 Xihai Zhao, 1 Le He, 1 Chun Yuan, 1,4 Jianwen Luo, 1 *Jing Bai 1 1 Biomedical Engineering, School of Medicine,Beijing, China; 2 Philips Research Asia, Shanghai, China;3Ultrasonography, Meitan General Hospital, Beijing, China;4Radiology, University of Washington, Seattle, Washington USAObjectives—Conventional B-mode ultrasound images andDoppler/color flow measurement are mostly used to evaluate the degreeof carotid atherosclerotic plaques, but they have a lack of histologic validation,while the correspondence between multicontrast magnetic resonanceimaging (MRI) sequences and histology has been constructed. Wepropose a comprehensive MRI and ultrasound feasibility study to quantitativelymeasure morphologic and mechanic properties of carotid atheroscleroticplaques and develop a risk indicator for plaques.Methods—Sequences of raw ultrasound (radiofrequency) datawere acquired from a 65-year-old male patient with carotid plaques on aPhilips iU22 ultrasound system. The interframe strain of the plaques wasestimated to indicate relative stiffness of different plaque compositions,using a coarse-to-fine 2D speckle-tracking algorithm based on cross-correlationand correlation filtering. The same patient underwent doubleblindedMRI scanning on a Philips Achieva 3T TX MR scanner using amulti–contrast imaging protocol. 3D MR images of this patient were reconstructed,and slices at the same position of the ultrasound incidenceangle were selected and registered. Carotid plaque tissue compositions onMR images were characterized according to published criteria, whileechogenicity and strain values in the ultrasound images were investigatedand compared with MRI results.Results—The plaques with intraplaque hemorrhage (IPH) or alipid-rich necrotic core (LRNC) on MR images were defined as high risk.In the ultrasound results, the calcified area of the plaque showed highechogenicity and low deformation (0.2%); IPH showed mid to highechogenicity and intermediate deformation (1.2%), and the LRNC showedthe lowest echogenicity and large deformation (2.6%). The locations ofthe calcification, IPH, and LRNC were in good agreement with findingson MR images.Conclusions—We performed a quantitative measurement ofthe morphology and mechanical properties of high-risk plaques andshowed that the combination of echogenicity and strain values obtainedfrom raw ultrasound data is feasible to quantitatively evaluate the vulnerabilityof atherosclerotic plaques.1536174 A Huber-Penalized Akaike-Regularized Broken-Stick LeastSquares Regression Algorithm for Shear Wave Velocity ReconstructionAtul Ingle, 1 * Tomy Varghese 1,2 1 Electrical and ComputerEngineering, 2 Medical Physics, University of Wisconsin, Madison,Wisconsin, USAS9Objectives—Tissue stiffness has been traditionally used as aqualitative metric for localizing cancerous tumors. The aim of this studyis to obtain quantitative stiffness estimates using electrode vibration elastography(EVE). It may be easier to distinguish stiffer tumors from healthytissue from local shear moduli than inspecting B-mode images. This workproposes a statistically robust least squares fitting technique capable ofdetecting boundaries between materials with different shear wave velocities(SWVs).Methods—In an EVE tissue-mimicking phantom, an ablationelectrode attached to an inclusion and vibrated by an actuator is used togenerate shear waves. Ultrasound displacement estimators are used totrack the shear wave pulse and record its arrival time at different locations.Assuming this plot is continuous and piece-wise linear, slopes and breakpoints are detected using a least squares fit embellished with a Huberpenalty. This penalty function switches from quadratic to linear for verylarge errors, thereby mitigating the effect of outliers. An Akaike informationcriterion is then used to trade off the error in the fit and the number ofbreak points. The SWV estimates are compared with a commercial shearwave imaging system and mechanical testing.Results—SWV estimates are within 20% of those obtainedusing the commercial scanning system. Although the stiffness estimatesare lower than those obtained from mechanical testing, the boundary delineationis quite good due to sharp transitions arising from the piece-wiselinear fits.Conclusions—The use of statistically robust noise filteringtechniques shows promise for improving results in the inverse problem ofshear stiffness reconstruction from noisy ultrasound displacement data.(Supported by National Institutes of Health grants R01CA112192-05 andR01CA112192-S103.)Table 1SWV, m/sYoung Modulus, kPaRegionMechanicalof Interest EVE SSI EVE SSI TestingInclusion 3.4 ± 1.5 2.8 ± 1.1 42.2 ± 58 24.2 ± 5.8 54.4 ± 0.1Partially ablated 2.0 ± 0.3 2.3 ± 0.8 12.1 ± 4.2 13.3 ± 3.5 21.6 ± 0.3Background 1.4 ± 0.4 1.3 ± 0.4 6.5 ± 6.1 4.8 ± 0.5 3.7 ± 0.1SSI indicates supersonic shear imaging.1540416 A Comparison of Coherence of Radiofrequency DataFrom Ablated and Unablated Liver Tissue Using MultitaperEstimationNicholas Rubert,* Tomy Varghese University of Wisconsin,Madison, Wisconsin USAObjectives—Thermally ablated liver tissue presents as a zone ofmixed echogenicity on B-mode imaging, making it difficult to delineatethe extent of treatment following thermal ablation. We demonstrate a novelcontrast mechanism for ultrasonic imaging of thermal ablations based onthe spatial distribution of acoustic scatterers. The portal triads and centralvein of the liver are arranged in repeating subunits, which are hypothesizedto correspond to quasi–periodically arranged acoustic scatterers. Estimatesof the mean scatterer spacing (MSS) of these quasi–periodic scatterershave been hypothesized to be useful indicators of pathologic changes tothe liver. Mathematically, the quasi–periodic scatterers gives rise to an ultrasoundradiofrequency (RF) signal-possessing coherence.Methods—Coherence is a frequency domain quantity computedwith tapered fast Fourier transforms of the ultrasound signal. Usingsimulation, we show that a coherence estimate using multiple orthogonaltapers outperforms coherence estimates computed with a single taper. We
American Institute of Ultrasound in Medicine <strong>Proceedings</strong> J Ultrasound Med 32(suppl):S1–S134, 2013also perform receiver operating characteristic (ROC) analysis of 38 regionsof interest (ROIs) from ex vivo bovine livers. Ultrasound imagingwas performed using a 9L4 transducer on a Siemens S2000 system priorto and following RF thermal ablation in 19 independent samples cut from4 bovine livers. Each ablation was approximately 1 to 2 cm in diameter.Results—In frequency domain Monte Carlo simulations, themultiple taper (MT) method was better able to estimate the MSS ofgamma-distributed scatterers than any single taper (ST) calculations. Inanalyzing 2 different ROI sizes ex vivo, we found that for a gate length of5 mm, we achieved an area under the ROC curve of 0.89, while at a gatelength of 7 mm, we achieved an area under the ROC curve of 0.93 usingMT coherence calculations.Conclusions—ROC analysis indicates that a tissue classifierusing coherence is able to distinguish between ablated and unablatedtissue and that an MT calculation of coherence is a better estimate than anST calculation. (Supported by National Institutes of Health grantsR01CA112192-05, R01CA112192-S103, and T32 CA09206-31).1540510 Acoustic Radiation Force Impulse Delineation of HumanCarotid Atherosclerotic Plaque Composition and StructureIn VivoTomasz Czernuszewicz, 1 Melissa Caughey, 2 Peter Ford, 3Mark Farber, 3 William Marston, 3 Raghuveer Vallabhaneni, 3Jonathon Homeister, 4 Matthew Mauro, 5 Timothy Nichols, 2,4Caterina Gallippi 1 *1 Joint Department of Biomedical Engineering,2 Medicine, 3 Surgery, 4 Pathology and Laboratory Medicine,5 Radiology, University of North Carolina, Chapel Hill,North Carolina USAObjectives—Conventional atherosclerosis imaging methodsrely on detecting luminal obstruction, which may not occur until late stagesof disease progression and does not reflect plaque rupture potential. Plaquedetection may be expedited and risk assessment improved by imagingmethods that describe the plaque composition and structure. We have previouslyso demonstrated acoustic radiation force impulse (ARFI) ultrasoundpreclinically with immunohistochemical validation. The objectiveof this work was to similarly validate ARFI in human atherosclerosis. Wehypothesize that ARFI delineates the fibrous cap, lipid/necrotic components,fibrous tissue, and calcium deposits in human atherosclerotic plaquein vivo.Methods—ARFI imaging was performed using a SiemensAcuson Antares imaging system with modifiable beam sequencing and aVF7-3 transducer in the carotid arteries of 2 patients undergoing clinicallyindicated carotid endarterectomy. Imaging was performed presurgicallyby focusing on the surgical plaque. After surgery, the extracted specimenwas sectioned according to noted arterial geometry for spatial registrationto the imaging plane. The sections were stained with hematoxylin-eosin,Verhoff van Gieson, Masson trichrome, and von Kossa. Parametric 2DARFI images of peak displacement (PD) were rendered.Results—In a fibroatheromatic plaque, 3 times higher ARFIPD was measured in the position of a soft lipid/necrotic region than in theposition of a thick fibrous cap above the region or fibrotic tissue below theregion. ARFI PD was nearly zero in positions of small (5 µm in the adjacent arterial tissue.Conclusions—These spatially matched ARFI and immunohistochemicaldata suggest that ARFI is relevant to describing the atheroscleroticplaque composition and structure in humans in vivo. The resultsalso demonstrate the feasibility of collecting data to perform a larger-scalestatistical reader study to evaluate human ARFI atherosclerosis imagingperformance using spatially matched immunohistochemistry as the validatingstandard.S101539520 Echocardiography-Based Measurements of 3-DimensionalMyocardial Fiber StructureMichelle Milne, 1 Kirk Wallace, 4 Benjamin Johnson, 1 GautamSingh, 2 Ravi Rasalingam, 3 James Miller, 1 Mark Holland 1,2 *1Physics, 2 Pediatrics, 3 Internal Medicine, Washington University,St Louis, Missouri USA; 4 GE Global Research, Niskayuna,New York USAObjectives—Previous studies from our laboratory demonstratedthat quantitative measurements of myocardial fiber structure for individualhearts can be derived from analyses of echocardiographic images.Echocardiography-based measurements of fiber structure at specific transverseplanes agreed well with those obtained using diffusion tensor magneticresonance imaging methods. The objective of this study was toextend the echocardiography-based measurements to produce 3D myocardialfiber structure images of the entire heart.Methods—A series of 2D apical echocardiographic imageswere acquired from each of 7 excised intact sheep hearts using a GE Vivid7 clinical imaging system. Myocardial fiber orientations corresponding tospecific distances from the apex of the heart were generated from analysesof radial line backscatter profiles within the ventricular walls of theheart in conjunction with a previously determined relationship betweenthe backscatter level and the angle of insonification relative to myocardialfiber orientation. The fiber orientations at each measured distancefrom the apex were assembled to produce a 3D fiber orientation image ofthe entire heart. In addition, 3D volumetric apical echocardiographic imageswere acquired from a subset of the excised hearts for comparison.Results—3D images depicting myocardial fiber structure obtainedfrom analyses of echocardiographic data appear consistent with theknown fiber structure of the heart. Images demonstrate left ventricularmid-myocardial fibers oriented within the short-axis plane and graduallybecoming more longitudinally oriented toward the epicardial and endocardialsurfaces. Data from 3D volumetric apical echocardiographic imagessuggest similar results.Conclusions—These results demonstrate that measurementsof 3D myocardial fiber structure of the entire heart can be successfullyderived from analyses of echocardiographic images. Further developmentof this method may provide a method for mapping the myocardial fiberorientation in individual patients over the heart cycle and provide a meansfor assessing potentially altered fiber structure associated with congenitaland acquired heart diseases. (Supported by National Institutes of Healthgrant R01 HL040302.)1541124 Effects of Preprocessing on Reconstructed Shear WaveSpeeds in Human Liver In VivoNed Rouze, 1 * Seung Yun Lee, 1 Michael Wang, 1Mark Palmeri, 1 Manal Abdelmalek, 2 Kathryn Nightingale 11Biomedical Engineering, Duke University, Durham, NorthCarolina USA; 2 Medicine, Division of Gastroenterology, DukeUniversity School of Medicine, Durham, North Carolina USAObjectives—Time-of-flight methods are often employed to providequantitative measurement of shear wave speed (SWS) from ultrasonicallytracked displacements following acoustic radiation forceexcitation in liver. These methods estimate overall group velocities ofpropagating shear waves and generally require filtering to reduce noiseand motion artifacts. In addition, different systems employ different beamgeometries in their push excitations, which also leads to differences in theshear wave frequency content. Differences in the shear wave frequencycontent can lead to differences in the estimated SWS due to the dispersivenature of hepatic tissue. In this study, we investigated the impact of theseeffects in a nonalcoholic steatohepatitis patient population from data obtainedin an Institutional Review Board–approved protocol.Methods—Data from >170 patients with a range of fibrosisstages were processed using multiple filtering algorithms. Three types ofmotion filters were applied, including a quadratic motion filter, a high-