Evaluation of bubble tracking algorithms for super-resolution imaging of microvessels

Ultrasound super-resolution imaging of microvessels and quantifying flow velocities have been proposed lately by several authors [1,2,3]. The standard approach is tracking microbubbles (MB) by searching for the nearest neighbor (NN) detection in consecutive frames. In [3] a Markov chain Monte Carlo data association (MCMCDA) algorithm was implemented to handle more complex vessel morphologies and/or higher bubble concentrations. Here, we investigate the performance of the algorithms in random vessel trees with known ground truth which simulate typical measurement conditions when imaging tumors in vivo. By this, we evaluate the quality of vessel reconstruction, the accuracy of velocity estimates and the influence of microbubble concentration. We demonstrate advantages of the MCMCDA in estimation accuracy of tracks and velocities in more complex vessel morphologies as they are expected in tumor microvasculature.