P3D-4 3D Breast Elastography with a Combined Ultrasound/Tomography System

Abstract

We have produced high quality strain images in a breast phantom and in 7 human subjects with 3D ultrasound (US) breast elastography using a combined US/tomography system. All radiofrequency (RF) images in this study were acquired using a GE Logiq 9 scanner and a linear 1D array operating at 7.5 MHz on a stand-alone, mammography-mimicking unit. To determine 3D elastography efficacy, a breast phantom (ATS BB-1) was imaged using static compression at 0.5% axial steps up to 2.5% strain over 7-17 elevational steps with slice thicknesses 15-60% of the elevational beamwidth. RF images were correlated using 3D, phase-sensitive speckle tracking algorithms and accumulated, estimated displacements were converted to strain images. Image quality was assessed via correlation coefficient (R) and strain contrast-to-noise ratio (CNR). Results indicated that R remained high and nearly constant (0.96-0.98) for a 0.5% strain step under all conditions. Elevational slice thicknesses of les 30% of the elevational beamwidth sizes produced the highest CNR because thicker slices did not sufficiently meet Nyquist requirements. At slice thicknesses of 35% elevational beamwidth, at least 7 elevational slices were required to meet 3D speckle tracking algorithm spatial requirements in the elevational direction (filter ges kernel = 1 elevational speckle spot). Moving beyond these minimum requirements produced the greatest improvement in CNR with 3D tracking: acquiring elevational planes over 3 speckle spots produced a 90% CNR improvement over 2D analogous. Acquiring up to 4.5 speckle spots (17 elevational planes) increased the CNR by a total of 130%. Additionally, elevational slices off the center axis confirmed on-axis results. Human subject motion was addressed before applying these results in vivo. Volume data acquisition must occur within a patient breath hold (les 10 sec). Thus, all 7 human subjects (1 cancer, 6 fibroadenomas) were imaged using quasistatic elastography as they held their breath. Five axial compression steps were acquired at 0.3-0.7% strain for 7-11 elevational planes with slice thicknesses 30% of the elevational beamwidth (spatially equivalent to 2.0-4.6 speckle spots). When minimal out-of-plane motion was present, there was no significant difference in correlation coefficient values created from 3D and fast (<0.5 sec) 2D acquisition. Thus, potential motion artifacts introduced by 3D data acquisition have been minimized. Of the 7 lesions imaged, 3 were visible on both 3D and 2D and elastograms, with the 3D elastograms depicting CNRepsiv of 7-11% better than 2D. This suggests that 3D elastography on the combined system holds great potential for improving an already clinically valuable imaging technique