Distributed aberration correction via differentiable beamforming

Distributed aberration correction has long been a goal in adaptive imaging. While early models of distributed aberration and simulations of wave propagation could adequately capture the distortion imposed by a heterogeneous medium, the ability to correct for distributed aberration in ultrasound imaging remained challenging. In recent years, advances in sound speed estimation techniques coupled with synthetic aperture focusing have allowed for the development of a wide variety of distributed aberration correction techniques that incorporate wave propagation models. In this talk, we describe distributed aberration correction using the differentiable beamforming model. This model iteratively beamforms signals from a synthetic aperture sequence using an a priori sound speed map, calculates a phase error loss from signals sharing a common midpoint, and backpropagates the loss gradient to update the sound speed map used for calculating the beamforming time delays. Iterations are continued until the phase error converges. Beamforming time delays depend on the wave propagation model used, which includes straight ray, bent ray path, or numerical implementations of the wave equation. We show examples of distributed aberration correction with the differentiable beamformer in both heterogeneous tissue-mimicking phantoms and in vivo tissue, demonstrating correction of both local and global image distortion.