Suppressing Clutter Components In Ultrasound Color Flow Imaging Using Robust Matrix Completion Algorithm: Simulation And Phantom Study

Abstract

In this paper, we propose a novel technique for suppressing clutter in ultrasound Color Flow Imaging (CFI). The unexpected clutter signal originating from slowly moving tissue prevents a clear visualization of vasculature. Using eigen-based filters is a state-of-the-art technique to reject the tissue echo. However, it remains an issue to find the exact rank of clutter and blood subspaces in these algorithms. Additionally, in case of noisy data, linear eigen-based filters lead to inefficient suppression of clutter. Moreover, conventional eigen-based methods are subject to lengthy computation times. To resolve these issues, we consider the task of clutter rejection as a foreground-background separation problem where the moving blood is the foreground and the steady tissue is modeled as the background. This problem is solved by adapting the fast Robust Matrix Completion (fRMC) algorithm for suppressing clutter in ultrasound CFI. The acquired ultrasound frames are stacked into a data matrix, the rank of which is minimized using the In-face Extended Frank-Wolfe method, which extracts the sparse blood component. For this method, manual tuning to determine the rank of clutter and blood sub-spaces is not necessary. Furthermore, the algorithm substantially decreases computation time. We name the proposed technique RAPID-Robust mAtrix decomPosition for suppressIng clutter in ultrasounD. RAPID is validated against simulation and flow phantom data and the results are compared to that of the conventional Singular Value Decomposition (SVD) method. In both of the experiments, RAPID provides a better visualization of the blood vessel than SVD. Furthermore, RAPID improves the execution time by more than 12, 000%.