Blind source separation-based velocity estimation (BSSVE): simulation and clinical results

We present a novel, two-dimensional tissue motion estimation method using blind source separation (BSS), which accurately measures both axial and lateral blood velocity components in the common carotid artery. Ensembles (packet length=20) of raw RF data were decomposed via BSS into orthogonal basis functions generally spanning vessel wall, blood, and noise signal subspaces separately. The depth and time projections corresponding to each basis function were computed. Employing knowledge of the imaging system´s axial and lateral spatial frequencies, two-dimensional blood velocity is calculated from the basis function time projections. The BSS technique is verified using field II pulsatile flow simulations at 45°, 60°, 70°, 80°, and 90° flow angles. The method, blind source separation-based velocity estimation (BSSVE), is demonstrated for clinical application and compared to speckle tracking methods using raw RF data collected from the carotid artery of a volunteer. BSSVE velocity and angle measurements in the simulated data are consistently below 10% error. The BSSVE clinical velocity and angle measurements are within 10% of the values computed with speckle tracking. Simulation results indicate that the method successfully measures axial and lateral flow components. Clinical results affirm the efficacy of BSS for two-dimensional in vivo tissue velocity estimation.