Gijs A.G.M. Hendriks

Automated 3D ultrasound elastography of the breast: a phantom validation study

In breast cancer screening, the automated breast volume scanner (ABVS) was introduced as an alternative for mammography since the latter technique is less suitable for women with dense breasts. Although clinical studies show promising results, clinicians report two disadvantages: long acquisition times (>90 s) introducing breathing artefacts, and high recall rates due to detection of many small lesions of uncertain malignant potential. Technical improvements for faster image acquisition and better discrimination between benign and malignant lesions are thus required. Therefore, the aim of this study was to investigate if 3D ultrasound elastography using plane-wave imaging is feasible. Strain images of a breast elastography phantom were acquired by an ABVS-mimicking device that allowed axial and elevational movement of the attached transducer. Pre- and post-deformation volumes were acquired with different constant speeds (between 1.25 and 40.0 mm s(-1)) and by three protocols: Go-Go (pre- and post-volumes with identical start and end positions), Go-Return (similar to Go-Go with opposite scanning directions) and Control (pre- and post-volumes acquired per position, this protocol can be seen as reference). Afterwards, 2D and 3D cross-correlation and strain algorithms were applied to the acquired volumes and the results were compared. The Go-Go protocol was shown to be superior with better strain image quality (CNRe and SNRe) than Go-Return and to be similar as Control. This can be attributed to applying opposite mechanical forces to the phantom during the Go-Return protocol, leading to out-of-plane motion. This motion was partly compensated by using 3D cross-correlation. However, the quality was still inferior to Go-Go. Since these results were obtained in a phantom study with controlled deformations, the effect of possible uncontrolled in vivo tissue motion artefacts has to be addressed in future studies. In conclusion, it seems feasible to implement 3D ultrasound quasi-static elastography on an ABVS-like system and to reduce scan times within one breath-hold (~10 s) by plane-wave acquisitions.

Quasi-static elastography and ultrasound plane-wave imaging: The effect of beam-forming strategies on the accuracy of displacement estimations

Quasi-static elastography is a widely applied ultrasound method in which RF data acquired in tissue at different states of deformation are correlated to estimate displacements and strain (displacement gradient). A recent development is the introduction of ultrafast plane-wave imaging where element data are beam-formed after collection to reconstruct the image lines. Several beam-forming strategies are available: time-domain based delay-and-sum (DAS) based on time-of-flight, and frequency domain based techniques like Stolts f-k (Stolt) and Lus method (Lu) based on an exploding reflector model and ideal transmit-receive model, respectively. Plane-wave beam-forming allows arbitrary grid designs not limited by probe pitch etc. In this study, we investigated the influence of beam-forming strategies and grid designs on the performance of displacement estimation.

Estimation of the 2D motion induced by an acoustic radiation force push pulse in transverse cross-sections of vessel-mimicking phantoms using high frequency ultrasound displacement compounding

Quantification of local arterial wall elasticity may assist in differentiating lipid rich rupture prone atherosclerotic plaques from stable fibrous plaques. Because lipid cores can be present anywhere along the circumference, we focus on developing a noninvasive shear wave elastography technique for transverse carotid cross-sections. Tracking the induced wave is not trivial because the circular geometry will result in particle motion in both the axial as well as the lateral direction. This study investigates the possibility to measure the minute 2D motion field using a single high frequency ultrasound transducer by combining axial displacements estimated at multiple beam-steering angles.

Improved plane wave ultrasound image reconstruction using a deconvolution-based Fourier domain approach

Different from conventional focused ultrasound, ultrafast ultrasound imaging employs full-field transmission such as a plane wave to achieve frame rates in the order of 10 kHz. Image reconstruction of plane-wave ultrasound is more computational efficient to be processed in Fourier domain than in time domain. A widely-applied seismic wave migration technique, the Stolts f-k Fourier-domain method, was modified to fit the plane wave transmission-receiving process into the Exploding Reflector Model (ERM). In comparison with the ideal fitting in Lus f-k method, the fitting in the Stolts f-k is slightly imprecise for the higher lateral Fourier components that results in residual patterns that degrade the image quality. We propose a template that can be applied in Fourier domain directly to deconvolute the residual point spread function (residual PSF) induced by imprecise fitting.

Design of an angular weighting template for coherent plane wave compounding in Fourier domain

Coherent plane wave compound imaging has already shown to provide equal image quality as multi-focal conventional imaging at more than ten times higher frame rates. Furthermore, image reconstruction of plane wave ultrasound is more computational efficient in Fourier domain (e.g. Lus or Stolts f-k) than in spatial domain (e.g. Delay-and-Sum (DAS)). In this study we fully integrated the Stolts f-k method and coherent compounding in the Fourier domain to further increase its computational efficiency. Additionally, we introduced a weighting template for Fourier domain methods which is rotated as a function of plane wave transmission angle (angular weighting) and investigated how it affects contrast and resolution of coherently compounded images.

3-D ultrafast ultrasound strain imaging to improve breast cancer detection

The automated breast volume scanner (ABVS) is used for breast cancer detection and consists of a linear transducer translating over the breast while collecting ultrasound data to reconstruct a breast volume. Although the ABVS shows high sensitivity, specificity remains a limitation resulting in high recall-rates. To improve ABVS specificity, we verified if it is feasible to implement 3-D strain imaging to discriminate between malignant (firmly-bound and stiff) and benign lesions (loosely bound and less stiff) based on lesion-to-surrounding-tissue connectivity (shear strain) and on stiffness (axial strain). Three phantoms containing loosely- (1) and firmly-bound (2) three times stiffer lesions were scanned twice by the ABVS, pre- and post-deformation (0.5 mm). Data was collected by plane-wave imaging to reduce acquisition times below one breath-hold. Displacements were calculated by cross-correlation from which axial and octahedral-shear strains were derived. Shear strains were distributed tightly and glob...

3-D ultrasound elastography of the breast: First steps towards ABVS implementation

Mammography is the gold standard for breast cancer screening but less suitable for women with dense breasts. Therefore, the ultrasound automated breast volume scanner (ABVS) was introduced as an alternative. Although clinical studies showed high sensitivity of ABVS to detect breast cancer, specificity and long acquisition times (>1 minute) remain an issue. Since specificity can be improved using elastography, the aim of this study was to develop 3-D strain imaging for ABVS scanning while also reducing acquisition times by plane wave imaging. To mimic the ABVS, an ATL L12-5 50mm transducer was attached to a motorized XZ-translational setup and connected to a Verasonics V1 ultrasound system. The aim was to obtain 150 equidistant (0.5 mm) 2-D strain images (60×25 mm2) of a breast phantom (model 059, CIRS) containing multiple inclusions. To obtain these images, beam-formed ultrasound radiofrequency (RF) data were acquired before and after deformation of the breast, followed by coarse-to-fine 2-D cross-correlation and 1-D least-squares strain estimation. The performance of strain estimation was compared for two transmission schemes at two frame rates of 2 Hz and 16 Hz (one breath hold) by calculating elastographic contrast-to-noise ratios (CNRe) and signal-to-noise ratios (SNRe). In transmission scheme 1 pre- and post-deformation RF data were acquired per position, whereas in scheme 2 (ABVS-like acquisition) RF data were acquired for all positions per state of deformation. All methods provided similar strain images upon visual inspection for most positions, although first and last 15 strain images (corresponding to rounder parts of the breast) resulted in reduced quality (SNRe 0±6 dB; CNRe 15±0 dB) in scheme 2 compared to scheme 1 (SNRe 16±1 dB; CNRe 28±4 dB). For the other 120 images, the two methods were similar (CNRe 33±1 dB; SNRe 15±1 dB). The frame rates did not seem to affect the CNRe and SNRe in both methods. In summary, scheme 1 performed similarly to the ABVS-like scheme 2 in the central part of the breast. Thus we might conclude that implementation of strain imaging in ABVS is feasible, although the reduced quality at the borders of the breast has to be improved. Moreover, plane wave imaging will enable high frame rates and lateral compounding and allow 3-D elastography within one breath hold.