Abstract
Most modern ultrasound scanners use the so-called pulsed-wave Doppler technique to estimate the blood velocities. Among the narrowband-based methods, the autocorrelation estimator and the Fourier-based method are the most commonly used approaches. Due to the low level of the blood echo, the signal-to-noise ratio is low, and some averaging in depth is applied to improve the estimate. Further, due to velocity gradients in space and time, the spectrum may get smeared. An alternative approach is to use a pulse with multiple frequency carriers, and do some form of averaging in the frequency domain. However, the limited transducer bandwidth will limit the accuracy of the conventional Fourier-based estimator; this method is also known to have considerable variance. More importantly, both the mentioned methods suffer from the maximum axial velocity bound, vZmax = cfprf/4fc, where c is the speed of propagation. In this paper, we propose a nonlinear least squares (NLS) estimator. Typically, NLS estimators are computationally cumbersome, in general requiring the minimization of a multidimensional and often multimodal cost function. Here, by noting that the unknown velocity will result in a common known frequency distorting function, we reformulate the NLS estimator as an one-dimensional minimization problem confirmed by extensive simulations. The results show that the NLS method not only works better than both the autocorrelation estimator and Periodogram method for high velocities, it will also not suffer from the maximum velocity bound.