A new class of aperture domain flow estimation algorithms

Current blood flow estimation algorithms operate on radio frequency (RF) or complex demodulated (IQ) data summed across a phased array, annular array, or single piston transducer. Typically either the phase delay or time delay between successively acquired signals is used to estimate the blood velocity. While this approach allows computational efficiency, velocity estimates are corrupted because different portions of the array lie at different angles with respect to the blood flow vector. While such effects can often be neglected for high f-number systems, they become increasingly important as aperture sizes increase. The angular dependence between the flow vector and the transducer elements broadens spectral flow estimates, increases estimator variance, and biases flow estimates. Flow estimates are further corrupted because different elements in the array receive slightly different speckle signals, as predicted by the Van Cittert-Zernike Theorem. Summation of these limited coherence signals causes a reduction in signal to noise ratio, further increasing estimator variance. This paper describes a new class of blood flow estimators which coherently process correlation or autocorrelation functions derived from data acquired by individual array elements. These aperture domain estimators can be formulated to process either IQ data or RF data. When operating on data acquired from one dimensional arrays these algorithms can yield two dimensional velocity estimates. Implementation on 1.5-D or 2-D arrays will allow estimation of the complete three dimensional blood velocity vector. Simulation results are presented which illustrate the limitations of current algorithms and potential of the narrow-band aperture domain velocity estimator (NAVE) and wideband aperture domain velocity estimator (WAVE). Simulation results are also presented which compare the bias and variance of WAVE to traditional blood flow estimators. Similarities between WAVE, NAVE, and other multidimensional velocity estimators, such as dual angle Doppler, are discussed. Tradeoffs between spatial resolution and performance are discussed. Finally, the implementation of wall filters for these algorithms is discussed