Effects of tissue mechanical and acoustic anisotropies on the performance of a cross-correlation-based ultrasound strain imaging method.

The anisotropic mechanical properties (mechanical anisotropy) and view-dependent ultrasonic backscattering (acoustic anisotropy) of striated muscle due to the underlying myofiber arrangement have been well documented, but whether they impact on ultrasound strain imaging (USI) techniques remains unclear. The aim of this study was therefore to investigate the performance of a cross-correlation-based two-dimensional (2D) USI method in anisotropic media under controlled quasi-static compression in silico and in vitro. First, synthetic pre- and post-deformed 2D radiofrequency images of anisotropic phantoms were simulated in two scenarios to examine the individual effect of the mechanical and acoustic anisotropies on strain estimation. In the first scenario, the phantom was defined to be transversely isotropic with the scatterer amplitudes following a zero-mean Gaussian distribution, while in the second scenario, the phantom was defined to be mechanically isotropic with Gaussian distributed scatterer amplitudes correlated along the principal directions of pre-defined fibers. These two anisotropies were then jointly incorporated into the ultrasound image simulation model with additional depth-dependent attenuation. Three imaging planes-the fiber plane with the fiber direction perpendicular to the ultrasound beam (TISperp_fb ), the fiber plane with the fiber direction parallel to the beam (TISpara ), and the transverse fiber plane (TISperp_cfb )-were studied. The absolute relative error (ARE) of the lateral strain estimates in TISperp_fb (20.99  ±  15.65%) was much higher than that in TISperp_cfb (4.14  ±  3.17%). The ARE in TISpara was unavailable owing to the large spatial extent of false peaks. The effect of tissue anisotropy on the performance of the 2D USI was further confirmed in an in vitro porcine skeletal muscle phantom. The best in-plane strain quality was again shown in TISperp_cfb (elastographic signal-to-noise ratio, or SNRe :  >25 dB), whereas the most unreliable strain estimates were found as expected in TISpara (SNRe :  <10 dB). The strain filter explained the effect of the mechanical anisotropy and required the underlying strain to be within an optimal range for estimation. Sonographic SNR (SNRs ) was found to be altered by the acoustic anisotropy and was much lower in TISpara (~10 dB) than in TISperp_fb (~50 dB) in vitro, which affected the accuracy of the strain estimation. Speckle size showed no evident impact on strain estimation but requires further examination.