Reliable Motion Estimation in Super-Resolution US by Reducing the Interference of Microbubble Movement

In ultrasound localization microscopy, microbubbles are localized and tracked in contrast enhanced ultrasound sequences to reconstruct the microvasculature. The accurate localization of MBs is essential to achieve super-resolved images, but it is strongly disturbed by tissue motion. Therefore, the MB positions have to be corrected for this tissue motion. Usually, a foreground-background separation is applied, to localize the MBs in the foreground, and to estimate the tissue motion in the background for the following correction of the MB positions. However, we recently showed that strong interferences can be introduced into motion estimation, depending on the choice of separation algorithm. Therefore, we propose to apply an optical flow approach to the non-separated B-mode images. During nonrigid motion estimation, the regions of the MBs are excluded. This leads to incomplete vector fields whose gaps are estimated afterwards. With simulations, we show an improvement of the localization accuracy over other approaches proposed, e.g. a reduction of the displacement error of 40 μm to around 6 μm. On preclinical data, we achieved an increased correlation between the motion corrected frames. Furthermore, the reconstructed microvasculature appears clearer.