Characterization of a multiscale variational optical flow method for elastography

Abstract

Part of the experiments carried out in our research have been developed using the time-domain cross-correlation estimator (TDE) to obtain the displacement field. When trying to use this method with real phantoms, decorrelation effects, mainly when dealing with pronounced tissue strains, compromises the accuracy of these methods. With this motivation we examined different ways to obtain the displacement field taking advantage of the properties of the B-mode images and the quantity and quality of different algorithms. A Multiscale Variational Optical Flow (MVOF) algorithm adapted to elastography gave the best results. Once the displacement is known, a least- squares strain estimator (LSE) is applied to estimate the tissue strain. To fully characterize the proposed algorithm, the elastographic signal-to-noise ratio (SNRe) and the contrast- to-noise ratio (CNRe) figures of merit are both obtained. It is important to remark that the input data for the algorithm comes from the ultrasound scanner standardized video output, therefore, its clinical implementation is immediate.