Back to Basics in Ultrasound Velocimetry: Tracking Speckles by Using a Standard Piv Algorithm

In this IUS proceeding, we describe a classical block-matching approach that we used during the 2018 SA-VFI (Synthetic Aperture 2-D Vector Flow imaging) challenge. To estimate frame-to-frame displacements, we used blockwise FFT-based ensemble cross-correlations. Subpixel displacements were obtained by parabolic peak fitting. We opted for a coarse-to-fine multiscale scheme to increase the resolution and precision. Robustness and accuracy were improved by including a robust unsupervised smoother in the estimation process. 2-D velocity vector fields were computed in several flow phantoms (from both simulations and experiments) provided by the organizers of the SAV-FI challenge. Our results showed that the standard block-matching approach provided reliable 2-D velocity vector fields, in terms of magnitude and angle. Complex flow patterns, like those occurring in the carotid bifurcation, were also estimated accurately. In summary, the long-standing block-matching technique by normalized cross-correlation is effective for flow estimation by ultrasound imaging. The final estimates returned by the method will be uploaded on the SA-VFI platform, and the results will be compared with those obtained by all the competitors during the challenge session at IEEE IUS 2018 in Kobe (Japan).