Does large beam-steered angle velocity compounding using plane wave transmission improve blood vector velocity estimation in a 3D carotid artery flow field?

Quantification of complex flow patterns with variations in the blood flow direction throughout an image plane and strong fluctuations over the cardiac cycle, requires dedicated transmission schemes. Conventional Doppler-based velocity imaging techniques are limited, due to their angle dependence and limited frame rates. New methods have been suggested to estimate the full 2D velocity vector. These methods need to be combined with intelligent transmit schemes to increase the imaging frame rate. This study investigates the performance of two different techniques to estimate the 2D blood velocity vector, both using a 2D normalized cross-correlation based displacement estimation algorithm. Normal 0° velocity estimation is compared with compound vector velocity estimation, using plane wave transmissions. Combining finite element modeling with FIELD II provided ultrasound radio frequency data of a carotid artery flow field over the entire cardiac cycle. Ultrasound radiofrequency element data were simulated for a linear array transducer (f_c = 9 MHz, f_s = 36 MHz, pitch = 198 μm). Plane waves were transmitted at a PRF of 2 kHz at either 0° angle only, or at sequentially changing angles of 0°, -20° and 20°. RF data were beamformed in the direction of the steered plane waves by delay-and-sum beamforming. The 0° data were additionally beamformed at angles of -20° and +20°. The horizontal velocity component was determined in three different ways: 1) directly from the 0° acquisition (PW0), 2) by compounding the axial velocity component from the 0° data beamformed at +20° and -20° (PW0_C), or 3) by compounding the axial velocity component from the multi-angle acquisitions (PW_C). The vertical velocity component was determined directly from the 0° acquisition. The performance of the methods was compared by calculating the root me- n squared error (RMSE) between estimated and true velocity components. The results showed the ability of the PW0 and PW0_C methods to provide accurate 2D velocity estimates. Both methods perform similarly based on the calculated RMSEs. PW_C showed less accurate velocity estimates, i.e., higher RMSEs, which are probably caused by the reduced effective frame rate and presence of grating lobes. At present frame rate and transducer choice, velocity compounding does not seem to be beneficial for 2D velocity estimation in a carotid artery flow field.