Ultrafast Vector Doppler Using RF Sub-Nyquist Sampling

International audienceBackground, Motivation and ObjectiveUltrafast ultrasound has contributed to a renewed interested in vector Doppler. Sampling RF signals at a rate 4 times the carrier frequency is the standard procedure since this rule complies with the Nyquist sampling theorem, regardless of the transducer bandwidth. RF acquisition with a high-performance multi-channel system generates massive datasets,especially in 3-D ultrafast ultrasound. The objective of this in vitro and in vivo study was to demonstrate that sub-Nyquist sampling can lead to substantial lossless data reduction in vector Doppler.Statement of Contribution/MethodsVector Doppler was generated from unsteered plane waves (5-MHz linear array). Two receive sub-apertures were used, with receive angles of ±15°. A staggered dual-PRF sequence (PRF2 = ⅔ PRF1) doubled the Nyquist velocity. The effect of RF data undersampling on vector Doppler was investigated in a rotating disc and in a carotid bifurcation. The RF signalswere sampled at 20 MHz (center frequency × 4), then downsampled to simulate sub-Nyquist sampling. We used downsampling ratios up to 13; a ratio of 13 means that the RF signals were sampled at 20/13 = 1.54 MHz. The ratios were chosen so that the positive and negative frequency components did not overlap within the -10 dB bandwidth. After I/Qdemodulation and beamforming, the Doppler velocities were estimated using an auto-correlator. The Doppler-derived velocity vectors were compared with the actual vector fields. The in vitro phantom rotated at angular velocities up to 15 RPS (maximum outer speed = 1.5 m/s). In vivo vector flow images of the carotid bifurcation were produced in onevolunteer. We used a high-frame-rate duplex sequence (B-mode + color Doppler) based on plane wave imaging. A spatially-adaptive polynomial regression filter was used to remove the clutter components. Since no in vivo reference was available, the velocity vectors produced from the undersampled RF signals were compared with those obtained from thestandard Nyquist sampling.Results/DiscussionThe velocity vector errors due to sub-Nyquist sampling were marginal, which illustrates that vector Doppler can be correctly computed with a drastically reduced amount of RF samples. In our study, a 11-fold data reduction was obtained. Sub-Nyquist sampling can be a method of choice in vector Doppler to avoid information overload and reduce data transfer and storage