A Wavefield Correlation Approach to Improve Sound Speed Estimation in Ultrasound Autofocusing

TL;DR

Results show that using WFC decreases sound speed estimation error, and using the estimates for aberration correction improves image resolution and contrast, which have potential to improve pulse-echo imaging for challenging clinical scenarios.

Abstract

Aberration often degrades ultrasound image quality when beamforming does not account for wavefront distortions. In the past decade, local sound speed estimators have been developed for distributed aberration correction throughout a medium. Recently, iterative sound speed optimization approaches have achieved more accurate estimates than earlier approaches, but these newer methods still struggle with decreased accuracy for media with reverberation clutter and large sound speed changes. To address these challenges, we propose using a wavefield correlation (WFC) beamforming approach when performing sound speed optimization. WFC correlates simulated forward-propagated transmit wavefields and backwards-propagated receive wavefields in order to form images. This process more accurately models wave propagation in heterogeneous media and can decrease diffuse clutter due to its spatiotemporal matched filtering effect. This beamformer is implemented using auto-differentiation software to then perform gradient descent optimization, using a total-variation regularized common midpoint phase focus metric loss, on the local sound speed map used during beamforming. This approach is compared to using delay and sum (DAS) with straight-ray time delay calculations in the same sound speed optimization approach on a variety of simulated, phantom, and in vivo data with large sound speed changes and clutter. Results show that using WFC decreases sound speed estimation error, and using the estimates for aberration correction improves image resolution and contrast. These promising results have potential to improve pulse-echo imaging for challenging clinical scenarios.

Figures (6)

• Figure 1: Diagram of the sound speed optimization approach used in this work. The common midpoint phase error is used as the image focus criterion minimized in the optimization objective function. The beamforming process to obtain the focus criterion, using WFC, is tracked using auto-differentiation software, which is then used to backpropagate the loss gradient to update the local sound speed estimation. This iterative gradient-descent-based optimization process then repeats until the estimations have reasonably converged to a local minimum.
• Figure 2: Sound speed and final aberration-corrected image comparisons for abdominal wall simulations with a sound speed inclusion (a) and two sound speed layers (b), with the ground truth simulated sound speed shown on the leftmost panels. The sound speed estimates using WFC show clearer differentiation of different sound speed regions, with the abdominal wall, layers, and circular inclusion more accurately represented in the estimates compared to the DAS predictions.
• Figure 3: Sound speed estimations for an acquisition with chicken over a constant sound speed phantom. The ground truth sound speeds based on calibration measurements or phantom specifications are shown on the left. A small region is excluded about the boundary between the chicken and phantom (shown in white). The chicken region appears clearly in the WFC sound speed estimation, with much lower bias compared to the DAS estimate, and the phantom region has lower error using WFC compared to DAS.
• Figure 4: Sound speed estimation and aberration correction comparisons for a fabricated alcohol-gelatin phantom. The difference between the inclusion and background sound speeds is around 25 m/s higher in the WFC prediction compared to the DAS prediction, and the background sound speed is more uniform for WFC compared to DAS. While the point target resolution improves with aberration correction from using a local sound speed map compared to using a constant sound speed, the final image using the WFC-estimated sound speed map leads to better lateral point target resolution compared to using the DAS-estimated sound speed.
• Figure 5: Sound speed estimates and aberration corrected images for thyroid acquisitions from different subjects. Regions of interest for gCNR calculations are indicated by arrows. The contrast in hypoechoic and anechoic regions improves more greatly for the WFC model to compared to the straight-ray DAS model. Some stronger scatterers exhibit improvements to resolution, especially when compared to using a constant sound speed to beamform, such as the scatterer in inclusion 3 in (a) or the scatterers forming the boundary of the structure in the green box in (b).