Reflection-mode speed-of-sound imaging using soft-prior limits

The diagnostic accuracy of classical gray-scale ultrasound (US) can be improved by complementing with new multimodal information. One promising candidate is speed-of-sound (SOS) imaging that reveals disease-related changes of tissue composition and structure. Computed ultrasound tomography in echo mode (CUTE) determines the spatial distribution of SoS based on pulse-echo signals by measuring the changing phase of an echo that is detected under a variety of transmit/receive settings. The SoS is then reconstructed via a regularized inversion of a forward model linking the SoS to the echo phase shift. In-vivo, clutter and aberration lead to phase noise that causes strong artefacts in the reconstructed SoS when using the previously proposed regularization of the spatial gradient of SoS (SG). To solve this shortcoming, we propose a soft-prior (SP) regularization that includes a statistical a priori description of the samples mean SoS variability of SoS. Both regularization approaches are compared in a phantom study mimicking the abdominal wall and liver tissue, where the SP regularization proves a much higher stability against phase noise. In an in-vivo scenario imaging a volunteer's liver, only the SP regularization leads to reproducible SoS reconstructions of the liver's SoS independent, of the scanning location.