Low-Rank Adaptive Clutter Filtering for Robust Ultrasound Vector Flow Imaging

Abstract

A low-rank based adaptive clutter filter is proposed to improve the quality of high frame rate vector flow imaging. The low rank characteristics of the tissue and sparsity of flow signals are used in the proposed model to formulate a convex optimization problem, which is a minimization of nuclear norm of the tissue matrix and $l_{1}$ norm of the flow matrix. The high frame rate plane wave based vector projectile imaging (VPI)is implemented using an FIR filter and the proposed low-rank filter for clutter removing and then compared each other. The number of columns of the matrix in the low-rank filter, i.e., up to the time length of processed flow signals in each filtering, is adaptively determined based on the mean estimated velocities in the current and previous frames. The weighting coefficient between the nuclear norm and the $l_{1}$ norm in the convex problem, is also adaptively selected based on the number of displayed vectors by the VPI in the current and previous frames. Results from a common carotid artery shows that the VPI has better performance using the proposed low-rank adaptive filter than that by the conventional FIR filter.