摘要
计算机断层成像(CT)作为一种无损的成像方式,被广泛应用于临床诊断和科学研究。几何校正的精度是影响系统成像质量的关键因素,但在一些特殊成像环境下,如高分辨CT成像系统中,由于机械运动系统的加工精度难以保证,可能会在扫描过程中出现非重复性的运动误差,这些误差难以通过常规的几何校正算法准确获取,从而会在成像过程中引入条状伪影并造成图像模糊。本文针对CT系统的非重复性运动误差造成的几何伪影,以扇形束CT为实验对象,探讨基于生成对抗网络(GAN)去除图像中这类几何伪影的可行性。实验结果表明,生成对抗网络可以有效去除重建图像中的由非重复性运动误差造成的几何伪影,提高图像质量。
As a nondestructive imaging method,computed tomography(CT)is widely used in clinical diagnosis and scientific research.Geometric calibration is an important factor that affects the imaging quality of a CT system.However,in some special imaging environments such as in high-resolution CT imaging systems,irreproducible motion errors may occur during the scanning due to the difficulty in guaranteeing the machining precision of the mechanical motion system.These errors are difficult to calibrate accurately since they are irreproducible.They can introduce streak artifacts and cause image blurring.In this paper,we investigate the feasibility of removing geometric artifacts caused by irreproducible motion errors in fan-beam CT systems based on Generative Adversarial Networks(GAN).The experimental results show that GAN can effectively remove the geometric artifacts caused by irreproducible motion errors and significantly improve image quality.
作者
陈雪
周子腾
Chen Xue;Zhou Ziteng(Biological Science and Medical Engineering Department,Southeast University,Nanjing 210096,China)
出处
《信息化研究》
2022年第6期33-37,共5页
INFORMATIZATION RESEARCH
基金
国家自然科学基金项目(No.62001112)
江苏省重点研发计划项目(No.BE2021609)
关键词
扇形束CT
几何伪影
生成对抗网络
Fan-beam CT
geometric artifact
Generative Adversarial Network
作者简介
陈雪(1998-),女,硕士研究生,主要研究方向为显微CT校正等。
