An Impoved U-GAT-IT model for Enhancing the Realism of Simulated Ultrasound Images

Abstract

Ultrasound (US) simulation has been widely used in clinical training and multi-modal image registration. However, there is a gap between the realism of simulated US images and real US images. In this paper, we propose an improved U-GAT-IT model by adding the spatial and channel “Squeeze and Excitation” (scSE) module to improve the feature extraction ability of generator. The model is used to enhance realism in simulated US images from computed tomography (CT) images. By comparing the quantitative indicators such as Peak signal-to-noise Ratio (PSNR), Structural Similarity (SSIM) and Fréchet Inception Distance (FID) with the previous models, the experimental results show that the proposed model is conducive to enhancing realism of simulated US images, and the simulation effect of fine structures is more realistic.