Applying in vitro elasticity imaging results to optimize in vivo breast lesion characterization using a combined 3D USs/digital X-ray system

Ultrasound-based reconstructive elasticity imaging has great potential for diagnosis and characterization of breast lesions. Applying external strain with a mammographic paddle as part of a combined 3D US/Digital X-ray system provides more uniform deformation and breast stability, offering opportunities to improve image fidelity. In this study, we examined phantom and in vivo strain image quality with three GE transducers (M12L, 10L, 7L) each operating at several frequencies between 5- 10 MHz and 3 TPX paddle thicknesses to predict optimal in vivo results with the combined system. Out-of-plane motion was measured by translating an ultrasonic transducer across a breast phantom (ATS, model BB-1) in 50um steps over 400um. Each image was correlated to the first in the sequence to determine rate of elevational decorrelation. Next, in-plane, strain-limited decorrelation was evaluated by correlating images at 0.25-1.0% steps up to 5% strain using two-pass 2D speckle tracking algorithms and accumulation. Adaptive strain estimation (ASE) was applied to maximize CNR throughout the final strain image. Overall the 10L transducer caused the least decorrelation due to out-of-plane motion (R = 0.97 at 7.5MHz and 400um elevational translation). In-plane decorrelation was also primarily strain- limited with the 10L transducer at 7.5MHz, with R = 0.9 for a 1.6% strain step. Accumulated strain images after ASE demonstrated a CNR = 13.6 with the 10L transducer at 7.5 MHz. Of the 3 paddles (0.25, 1.0, 2.5 mm) used in the phantom study, the 2.5mm paddle created strain images with less artifacts than the thinner paddles by providing more uniform deformation during compression. Next, we evaluated sources of in vivo chest wall motion in 7 subjects to minimize patient motion during the scan. Based on these experiments, patients were asked to breathe shallowly during exams as it caused less decorrelation due to chest wall motion (Ravg = 0.96 over 91 frames). In vivo results were acquired with the 10L at 7.5 MHz using continuous compression over 2.1 seconds. Strain images clearly distinguished between tissue types when accumulated up to 5% strain. In vivo results are limited by out-of-plane motion but are expected to improve with 3D tracking. These early successes indicate that using elasticity imaging with the combined system can potentially characterize breast lesions and monitor therapies.