Performance analysis of a regularized algorithm for elasticity imaging

We investigated mage quality features of a regularized optical flow (ROF) algorithm for strain imaging - noise, contrast and spatial resolution. ROF image features were compared to those obtained using a standard multi-resolution cross-correlation (MRCC) algorithm under equivalent conditions. Comparisons were made quantitatively using resolution phantom data and qualitatively using in vivo breast and flow phantom data. With a small applied compression, the axial resolution limit (ARL), peak strain value and standard deviation of the noise (SDN) for strain measured at the layer were respectively 1.78 lp/mm, 0.52%, 0.04% (ROF) and 1.78 lp/mm, 0.52%, 0.54% (MRCC). With a large compression, the ARL, strain and SDN were respectively 2 lp/mm, 3.15% and 0.29% (ROF) and 2 lp/mm, 3.08% and 1.05% (MRCC). These results show that the ROF algorithm provides up to an order of magnitude reduction in image noise with comparable contrast and spatial resolution. Flow phantoms with a large dynamic range for stiffness also showed that ROF effectively eliminates decorrelation noise without a significant reduction in contrast or resolution.