深度学习重建联合Smart去金属伪影算法在口腔金属植入物患者头颈CT血管成像中的应用  

Application of deep learning reconstruction combined with Smart metal artifact removal algorithm in head and neck CT angiography of patients with oral metal implants

作　　者：唐丽[1] 刘星 吕培杰[1] 高剑波[1] TANG Li;LIU Xing;LYU Peijie;GAO Jianbo(Department of Radiology,the First Affiliated Hospital,Zhengzhou University,Zhengzhou 450052)

机构地区：[1]郑州大学第一附属医院放射科,郑州450052

出　　处：《郑州大学学报（医学版）》2024年第4期484-487,共4页Journal of Zhengzhou University(Medical Sciences)

基　　金：河南省医学科技攻关计划联合共建项目(LHGJ20210327)。

摘　　要：目的:探讨深度学习重建(DLR)联合Smart去金属伪影(MAR)算法在口腔金属植入物患者头颈CT血管成像(CTA)中的应用价值。方法:选择郑州大学第一附属医院2023年2月至6月口腔有不可拆卸金属植入物行头颈CTA的患者70例,采用以下3种方法重建图像:基于混合模型的自适应迭代重建(ASIR-V)50%算法(IR),ASIR-V50%联合Smart MAR算法(IR-S),高水平DLR联合Smart MAR算法(DLR-S)。测量不受伪影影响的颈内动脉C1段和头夹肌感兴趣区CT值的标准差(SD)2和SD4,作为图像噪声指标;计算颈内动脉C1段和舌部的金属伪影指数(AI)1和AI2;对颈内动脉C1段和口腔整体图像质量进行主观评分。结果:与IR组和IR-S组比较,DLR-S组SD2和SD4降低(P<0.05)。与IR组比较,IR-S组和DLR-S组AI1、AI2降低;与IR-S组比较,DLR-S组AI1、AT2降低(P<0.05)。与IR组比较,IR-S组和DLR-S组口腔整体和颈内动脉C1段图像质量主观评分均增高;与IR-S组比较,DLR-S组图像质量主观评分增高(P<0.05),9例患者舌部可见新的伪影。结论:Smart MAR联合DLR可减少口腔植入物造成的金属伪影,提高头颈CTA图像质量。但Smart MAR可能引入新的伪影,需联合未加入Smart MAR的图像进行分析。Aim:To investigate the value of deep learning reconstruction(DLR)combined with Smart metal artifact removal(Smart MAR)algorithm in improving neck image quality of head and neck CT angiography(CTA)of patients with oral metal implants.Methods:A total of 70 patients with non-removable oral metal implants undergoing head and neck CTA at the First Affiliated Hospital of Zhengzhou University from February to June 2023 were selected.Three reconstruction algorithms were used:adaptive statistical iterative reconstruction(ASIR-V)50%algorithm(IR group),ASIR-V50%combined with Smart MAR algorithm(IR-S group),and high-level deep learning reconstruction(DLR)combined with Smart MAR algorithm(DLR-S group).The standard deviation(SD)of the CT values in the C1 segment of the internal carotid artery unaffected by artifacts(ROI2)and the head clasp muscle(ROI4)were measured as SD2 and SD4,respectively.Metal artifact index(AI)1 and AI2 of the C1 segment of the internal carotid artery and tongue were calculated.Image quality scores for the C1 segment of the internal carotid artery and the overall oral cavity were assessed.Results:Compared with the IR and IR-S groups,the DLR-S group showed lower SD2 and SD4(P<0.05);compared with the IR group,the IR-S and DLR-S groups exhibited reduced AI1 and AI2(P<0.05);compared with the IR-S group,the DLR-S group showed further reduction in AI1 and AI2(P<0.05).Overall image quality and C1 segment image quality scores were higher in the IR-S and DLR-S groups compared with the IR group,and higher in the DLR-S group compared with the IR-S group(P<0.05).New artifacts were observed on the tongue in 9 patients.Conclusion:The combination of Smart MAR and DLR can reduce oral metal artifacts caused by oral metal implants and improve the image quality of the C1 segment of the internal carotid artery and the oral cavity in patients undergoing head and neck CTA.However,Smart MAR algorithm may introduce new artifacts,necessitating analyzing images not incorporating Smart MAR.

关 键 词：深度学习重建 口腔金属植入物 金属伪影 CT血管成像 Smart去金属伪影算法 

分 类 号：R445.3[医药卫生—影像医学与核医学] R814.42[医药卫生—诊断学]