Adaptively learning Speed of Sound for improving Single Plane Wave Ultrasound Imaging

Conventionally, Delay and Sum (DAS) has been used to beamform Plane Wave (PW) imaging for ultrafast Ultrasound (US) applications. Although computationally efficient, single-PW DAS suffers from low contrast and resolution. Many adaptive Deep Learning (DL) approaches have been proposed for single PW that improve the image quality by learning certain pixel or channel weights to be combined with the DAS beamformed image. In this work, we explore an overlooked performance alternative for adaptively perturbing the time delays used in DAS by learning the underlying Speed of Sound (SoS) via a Neural Network (NN). Thus, we aim to directly modify the DAS beamforming algorithm and demonstrate that the resulting method produces images of better quality than DAS and other adaptive DL approaches.