2D Fully Resolved Strain Estimation Algorithm Evaluation on Simulations and on in-vitro Bovine Livers

摘要

Accurately estimating the strain remains fundamental in elastography since clinician's diagnosis as well as the quality of mechanical parameters reconstruction are directly related to those estimations. In this paper, we present a 2D fully-resolved iterative and adaptive strain estimation technique, appropriate to investigate media subjected to a wide range of strains. The method estimates axial strain while considering lateral motion. For each 2D RF region selected in the pre-compression image, its deformed version is searched in the post-compression image and its strain is estimated. In a first approximation, the deformed region is considered as a 2D shifted and time-scaled replica of the original one. This modeling considers no lateral scaling factor owing to the poor resolution in that direction, and thus reflects ultrasound imaging characteristics. The developed method focuses on achieving maximum strain estimation accuracy, and performs at each step, the deformation optimal track as follows: (i) First the 2D shift suffered by the considered ROI and induced by the deformation of regions surrounding it, is compensated for, by adaptively displacing the ROIs between the pre- and post-compression images. (ii) Strain parameters are then estimated as the arguments that maximize the normalized correlation coefficient between the original region and its deformed version compensated for the searched parameters. Because in elastography a small compression is applied to the tissue, resulting in expected small ranges of parameters feasible values, we used constrained optimization which increases the estimation robustness and accelerates the convergence. (iii) Finally axial strain images are directly deduced from the scaling factor fields. Simulation results from a mechanically homogeneous medium subjected to successive uniaxial loadings reveal an accurate estimation for strains up to 17%. Elastograms of in vitro bovine livers with harder lesions demonstrate the ability of our technique to investigate biological tissues.