Fast simulation of realistic pseudo-acoustic nonlinear radio-frequency ultrasound images

Abstract

In the medical ultrasound (US) community, only few simulators are able to fully simulate nonlinear wave propagation. In our recently developed software Creanuis [1], realistic nonlinear radio-frequency US images can be simulated. Unfortunately, the rather long computation time represents an important limitation, and is far from being comparable with the fastest simulation tools based on convolution strategies. In this work, a strategy combining Creanuis with a convolution appoach is proposed. This pseudo-acoustic nonlinear image strategy (PANIS) produces linear as well as nonlinear images. It consists to first simulate a set of punctual scatterers to locally extract a nonlinear point-spread function (PSF). Then, the 2D convolution of each of these PSF and the full medium is performed. The final PANIS image is obtained by selecting the specific part of each elementary convolved images. This final image contains the whole spectrum evolution with a depth dependent resolution and signal-to-noise ratio. The root-mean square error, between the statistics of PANIS and Creanuis images is kept under 1% and validates the model. The computation time is kept under 10 secondes for the convolution part.