Deep neural networks for non-linear model-based ultrasound reconstruction

TL;DR

This paper introduces a deep learning-based method for ultrasound image reconstruction that effectively handles the non-linearity of the forward model, significantly improving image quality over traditional linear and delay-and-sum methods.

Contribution

It presents a non-iterative, neural network-based reconstruction technique that refines linear estimates to account for non-linear ultrasound forward models.

Findings

Dramatic image quality improvements over traditional methods

Effective handling of non-linear ultrasound models

Validated on simulated ultrasound data

Abstract

Ultrasound reflection tomography is widely used to image large complex specimens that are only accessible from a single side, such as well systems and nuclear power plant containment walls. Typical methods for inverting the measurement rely on delay-and-sum algorithms that rapidly produce reconstructions but with significant artifacts. Recently, model-based reconstruction approaches using a linear forward model have been shown to significantly improve image quality compared to the conventional approach. However, even these techniques result in artifacts for complex objects because of the inherent non-linearity of the ultrasound forward model. In this paper, we propose a non-iterative model-based reconstruction method for inverting measurements that are based on non-linear forward models for ultrasound imaging. Our approach involves obtaining an approximate estimate of the reconstruction using a simple linear back-projection and training a deep neural network to refine this to the actual reconstruction. We apply our method to simulated ultrasound data and demonstrate dramatic improvements in image quality compared to the delay-and-sum approach and the linear model-based reconstruction approach.