On the use of denoing algorithms for ultrasound beamforming

Abstract

Beamforming in receive, whose objective is to estimate an image from raw RF data acquired by the probe piezoelectric elements, plays a crucial role in ultrasound imaging. The standard method, called delay and sum (DAS) and implemented in most of the commercial scanners, consists in coherently summing the RF signals, providing the backprojection solution of the inverse problem of beamforming. Despite real-time properties, DAS results into images with limited spatial resolution and contrast. The literature of ultrasound beamforming is rich and mainly consists in alternatives to DAS based on non-adaptive or adaptive (e.g., minimum variance, coherence factor) methods or image reconstruction algorithms in the Fourier domain. Furthermore, inverse problem formulations have been shown to be well-adapted to ultrasound beamforming. They consist in minimizing a cost function formed by two terms: a data fidelity term modelling the acquisition setup, and a regularization term. The choice of the latter is not straightforward in ultrasound imaging, mainly because of the need to conserve statistical properties of the speckle. In this paper, denoising algorithms are shown to be good regularizers for ultrasound beamforming, providing a good performance in spatial resolution and contrast gain, without deteriorating the quality of the speckle texture.