Accurate blood peak velocity estimation using spectral models and vector doppler

Ultrasound blood peak velocity estimates are routinely used for diagnostics, such as the grading of a stenosis. The peak velocity is typically assessed from the Doppler spectrum by locating the highest frequency detectable from noise. The selected frequency is then converted to velocity by the Doppler equation. This procedure contains several potential sources of error: the frequency selection is noise dependent and sensitive to the spectral broadening, which, in turn, is affected by the Doppler angle uncertainty. The result is, often, an inaccurate estimate. In this work we propose a new method that removes the aforementioned errors. The frequency is selected by exploiting a mathematical model of the Doppler spectrum that has recently been introduced. When a very large sample volume is used, which includes all the vessel section, the model is capable of predicting the exact threshold to be used without the need of broadening compensation. The angle ambiguity is solved by applying the threshold to the Doppler spectra measured from two different directions, according to the vector Doppler technique. The proposed approach has here been validated through Field II simulations, phantom experiments, and tests on volunteers by using defocused waves to insonify a large region from a linear array probe. A mean error lower than 1% and a mean coefficient of variability lower than 5% were measured in a variety of experimental conditions.