On the Performance of Time Domain Displacement Estimators for Magnetomotive Ultrasound Imaging

Thomas Ersepke, Tim C. Kranemann, Georg Schmitz

IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control (Volume: 66, Issue:5, May 2019) , DOI: 10.1109/TUFFC.2019.2903885


In magnetomotive (MM) ultrasound (US) imaging, magnetic nanoparticles (NPs) are excited by an external magnetic field and the tracked motion of the surrounding tissue serves as a surrogate parameter for the NP concentration. MMUS procedures exhibit weak displacement contrasts due to small forces on the NPs. Consequently, precise US-based displacement estimation is crucial in terms of a sufficiently high contrast-to-noise ratio (CNR) in MMUS imaging. Conventional MMUS detection of the NPs is based on samplewise evaluation of the phase of the in-phase and quadrature (IQ) data, where a low signal-to-noise ratio (SNR) in the data leads to strong phase noise and, thus, to an increased variance of the displacement estimate. This paper examines the performance of two time-domain displacement estimators (DEs) for MMUS imaging, which differ from conventional MMUS techniques by incorporating data from an axial segment. The normalized cross correlation (NCC) estimator and a recursive Bayesian estimator, incorporating spatial information from neighboring segments, weighted by the local SNR, are adapted for the small MMUS displacement magnitudes. Numerical simulations of MM-induced, time-harmonic bulk and Gaussian-shaped displacement profiles show that the two time-domain estimators yield a reduced estimation error compared to the phase-shift-based estimator. Phantom experiments, using our recently proposed magnetic excitation setup, show a 1.9-fold and 3.4-fold increase of the CNR in the MMUS images for the NCC and Bayes estimator compared to the conventional method, while the amount of required data is reduced by a factor of 100.

[IEEE Link]