Fast ultrasonic imaging using end-to-end deep learning

TL;DR

This paper introduces an end-to-end deep learning architecture that unifies data pre-processing, wave physics-based image formation, and post-processing in ultrasonic imaging, improving efficiency and accuracy over traditional sequential methods.

Contribution

The novel contribution is the integration of all three ultrasonic imaging steps into a single trainable deep neural network, including a differentiable DAS layer.

Findings

End-to-end model outperforms sequential approaches.

Effective on simulated data and shows potential on real data.

Demonstrates feasibility of deep learning in ultrasonic image formation.

Abstract

Ultrasonic imaging algorithms used in many clinical and industrial applications consist of three steps: A data pre-processing, an image formation and an image post-processing step. For efficiency, image formation often relies on an approximation of the underlying wave physics. A prominent example is the Delay-And-Sum (DAS) algorithm used in reflectivity-based ultrasonic imaging. Recently, deep neural networks (DNNs) are being used for the data pre-processing and the image post-processing steps separately. In this work, we propose a novel deep learning architecture that integrates all three steps to enable end-to-end training. We examine turning the DAS image formation method into a network layer that connects data pre-processing layers with image post-processing layers that perform segmentation. We demonstrate that this integrated approach clearly outperforms sequential approaches that are trained separately. While network training and evaluation is performed only on simulated data, we also showcase the potential of our approach on real data from a non-destructive testing scenario.