基于多尺度生成对抗网络的锥束CT图像耦合伪影校正方法  

Coupled artifacts removal in cone-beam computed tomography images based on multi-scale generative adversarial network

作　　者：柴世杰[1,2,3] 黄魁东 杨富强[1,3] 赵举龙 Chai Shijie;Huang Kuidong;Yang Fuqiang;Zhao Julong(School of Mechanical Engineering,Northwestern Polytechnical University,Xi′an 710072,China;Ningbo Institute of Northwestern Polytechnical University,Ningbo 315000,China;Key Laboratory of High Performance Manufacturing of Aero Engines,Ministry of Industry and Information Technology,Northwestern Polytechnical University,Xi'an 710072,China)

机构地区：[1]西北工业大学航空发动机高性能制造工业和信息化部重点实验室,西安710072 [2]西北工业大学宁波研究院,宁波315000 [3]西北工业大学航空发动机先进制造技术教育部工程研究中心,西安710072

出　　处：《仪器仪表学报》2024年第9期44-54,共11页Chinese Journal of Scientific Instrument

基　　金：浙江省“尖兵领雁+X”研发攻关计划项目(2024C01249(SD2));中国航空发动机集团产学研合作项目(HFZL2022CXY024);航空发动机及燃气轮机基础科学中心项目(P2022-B-Ⅳ-013-001)资助。

摘　　要：针对锥束CT(CBCT)图像存在的耦合伪影难以完全校正问题,提出1种基于多尺度生成对抗网络的锥束CT图像耦合伪影校正方法。首先,根据CT图像的伪影特征构建了一套包括仿真图像和实际图像的CBCT耦合伪影数据集,用于提高模型泛化能力。再将特征金字塔结构(FPN)和基于卷积块的注意机制(CBAM)融入网络的生成器结构中,帮助网络捕获更全面特征信息,并配合多尺度判别器(MSD)搭建生成对抗网络框架,使得生成的去伪影图像更加清晰和真实。实验分析显示,经本文方法校正图像的PSNR和SSIM在仿真数据集中提高了21.595 dB、0.541,在实际数据集中提高了14.072 dB、0.274。实验结果表明,本文方法可有效校正耦合伪影。To address the issue of incomplete correction of coupled artifacts in cone-beam computed tomography(CBCT)images,a coupled artifact correction method for CBCT images based on a multi-scale generative adversarial network(GAN)is proposed.Firstly,a CBCT coupling artifact dataset comprising both simulated and real images was constructed based on the artifact characteristics of CT images to enhance the model′s generalization capability.Additionally,the generator structure of the network was improved by integrating the feature pyramid network(FPN)and convolutional block attention module(CBAM)to capture more comprehensive feature information.We also employed a multi-scale discriminator(MSD)alongside these components to from a generative adversarial network framework,producing clearer and more realistic artifact-free images.Experimental analysis showed that the PSNR and SSIM of the corrected images increased by this method increased by 21.595 dB and 0.541 in the simulated dataset,and by 14.072 dB and 0.274 in the real dataset.The experimental results indicate that the proposed method can effectively correct coupled artifacts.

关 键 词：CBCT 耦合伪影去除 多尺度 FPN 生成对抗网络 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术] TH74[自动化与计算机技术—计算机科学与技术]