A Competitive Swarm Optimized SVD-based Clutter Filter

Ultrafast ultrasound imaging has promoted long-term developments and wide applications in the field of blood flow imaging. The singular value decomposition (SVD)-based clutter filter for ultrafast ultrasound highly increases the sensitivity of resolving smaller blood vessels than the conventional clutter filters, thus enabling ultrasound microvessel imaging. Currently, most of the SVD clutter filters use two cutoffs and hard thresholding in the singular values to separate the subspaces of tissue, blood flow, and noise. However, there is no strict theoretical basis to prove that SVD can ideally separate the tissue, blood flow, and noise into three independent subspaces. In this paper, we propose to estimate the contributions of blood flow in each singular value through global optimization. Specifically, we design a new SVD clutter filter by using the competitive swarm optimization (CSO) to search for the counterparts of blood flow signals in each singular value, i.e., a CSO-SVD clutter filter is proposed. We validate the feasibility of our idea and the effectiveness of the CSO-SVD clutter filter on public in-vivo cerebral datasets acquired from rat brains. The experimental results demonstrate that our filter significantly improves the contrast-to-noise ratio (CNR) of ultrafast power Doppler imaging (uPDI) when compared with the state-of-the-art SVD-based clutter filters.