Multi-resolution Data Processing for Accelerated and Robust Ultrasound Localization Microscopy

Abstract

Ultrasound localization microscopy (ULM) leverages widely-used contrast-enhancing microbubbles (MBs), localizing and tracking their movement in vivo, to permit the reconstruction of sub-diffraction images of tissue microvasculature and flow velocities. However, the technique is computationally expensive to implement, often requiring millions of MB localization and tracking events for microvascular reconstruction. Here we propose a multi-resolution localization and tracking algorithm for ULM reconstruction that selectively assigns microvascular MB flow to a high-fidelity, but slow, normalized cross-correlation (NCC)-based tracking algorithm and large vessel flow to faster processing solutions: either conventional colorflow (CF) Doppler, or a more rapid bipartite-graph and Kalman filter (BGKF)-based MB tracking algorithm. A low MB-count accumulation map was generated to assign sparse MB localization events to the NCC algorithm, with the remaining MB events sent either to the BGKF -based tracking algorithm or to conventional CF Doppler. In comparison to NCC alone, we were able to reduce processing time by approximately 20-40% using the proposed NCC+BGKF or NCC+CF combination algorithm on this imaging dataset.