High-Quality Reconstruction of Plane-Wave Imaging Using Generative Adversarial Network

Abstract

Coherent plane wave compounding (PWC) using tens of steered plane waves (PWs) can obtain high-quality ultrasound images but reduces the gain in frame rate. Recently a new strategy using convolutional neural network (CNN) was proposed to recover high-quality images from only 3 PWs. Considering the excellent performance of generative adversarial network (GAN) in image reconstruction, we propose to use GAN to reconstruct high-quality ultrasound images from 3 PWs. Phantom and in vivo experiments are performed. The results of GAN using 3 PWs, in terms of contrast ratio and lateral resolution are competitive with those of CNN using 3 PWs and coherent compounding using 31PWs, which demonstrates the feasibility of this method.