Differentiable Beamforming for Distributed Attenuation Estimation and Spatial Gain Compensation (SGC)

Ultrasound imaging with automatic differentiation has emerged as a technique to map many model-based approaches for estimating physical beamforming or tissue parameters onto a differentiable computational graph. Using a tomographic model, we adapt this approach to enable highly precise and accurate distributed attenuation estimation. We present a novel sub-aperture difference method for attenuation estimation on simulated and experimental ultrasound channel data. The presented attenuation estimation model estimates attenuation with less than 44 mdB/cm/MHz on simulated inclusion data, and 119 mdB/cm/MHz on simulated uniform attenuation data, but is challenged by experimental data with low signal-to-noise ratio and the selection of an a priori estimate of the attenuation.