Brain CT Image with Motion Artifact Augmentation Based on PGGAN and FBP for Artifact Detection

Abstract

Synthesis of computed tomographic images with motion artifact has many applications in assisted medical diagnosis such as artifact detection and removal. However, one of the challenges is on how to synthesize high-resolution images with motion artifact while artifact features and tissues are naturally presented within image. In this paper, we propose a new method to solve the problem by combing filtered back-projection (FBP) and progressive growing of generative adversarial networks (PGGAN), while FBP is for artifact generation and feature extraction and PGGAN is for feature augmentation. Finally, we superimpose artifact features onto artifact-free data, so to obtain a set of pre-demanded and diversified images with all kinds of motion artifacts. We quantitatively evaluate the synthetic images by training models with synthetic data for artifact detection. Our extensive experiments demonstrated that the performance of our proposed method is superior over the state-of-the-art methods.