摘要
目的 探讨低辐射剂量联合深度学习重建算法(DLIR)在提高肝转移图像质量和诊断能力中的应用价值。方法 前瞻性搜集因临床需要行上腹部CT增强检查的肝转移患者30例,静脉期采用标准辐射剂量扫描联合前置多模型迭代重建算法V(ASiR-V)40%算法行图像重建(对照组),加扫第二静脉期采用50%低辐射剂量扫描并联合DLIR算法三个强度水平[低(L)、中(M)、高(H)]行图像重建(实验组)。对所有图像行客观评价[图像噪声、肝转移病灶的对比噪声比(CNR)、肝脏和门静脉的信噪比(SNR)]及主观评价[总体图像质量和病灶显示能力]。采用单因素方差分析和Kruskal-Wallis H检验比较4组图像的客观和主观评价指标。结果 DLIR-L组的噪声高于ASiR-V40%组(P<0.05),病灶CNR、肝脏和门静脉SNR、总体图像质量和病灶显示能力均低于ASiR-V40%,但两组的肝脏SNR与总体图像质量差异均无统计学意义。DLIR-M组和ASiR-V40%组的噪声、病灶CNR、肝脏和门静脉SNR、总体图像质量与病灶显示能力差异均无统计学意义。与ASiR-V40%组相比,DLIR-H组的噪声减低(P<0.05),病灶CNR、肝脏和门静脉SNR和病灶显示能力相当,总体图像质量提高(P<0.05)。在DLIR 3组间,噪声随重建强度水平(DLIR-L、DLIR-M、DLIR-H)的升高而降低,总体图像质量和病灶显示能力随强度水平的提高而提高,且各组间差异有统计学意义(P<0.05)。病灶CNR、肝脏和门静脉SNR随DLIR强度的提高而降低,但仅DLIR-H与DLIR-L组差异有统计学意义(P<0.05)。结论 与ASiR-V40%相比,采用DLIR-H重建在减低50%辐射剂量条件下提高总体图像质量且保持肝转移病灶诊断能力不降低。
Objective To investigate the clinical value of low radiation dose combined with deep learning reconstruction algorithm(DLIR) in improving the image quality and diagnostic ability of liver metastasis. Methods Thirty patients who underwent abdominal enhancement CT scan were collected prospectively.In the control group, the portal venous phase images with standard radiation dose were reconstructed using ASiR-V40%(hybrid model based adaptive statistical iterative reconstruction, ASiR).In the study group, the second portal venous phase images with half radiation dose were reconstructed using DLIR at low, medium, and high strength level(DLIR L,DLIR M,and DLIR H).Quantitative parameters [image noise, contrast to noise ratio(CNR),liver and portal vein signal to noise ratio(SNR)] were measured and compared among the groups by using one way analysis of variance.Qualitative parameters(overall image quality and lesion conspicuity) were measured and compared by the Kruskal-Wallis H test. Results The image noise of DLIR L group was higher than that of ASiR-V40% group(P<0.05),and the CNR,liver SNR,portal vein SNR,overall image quality and lesion conspicuity of DLIR-L group were lower than those of ASiR-V40% group, but there was no significant difference in liver SNR and overall image quality between the two groups.There was no significant differences in image noise, CNR,liver SNR,portal vein SNR,overall image quality and lesion conspicuity between DLIR-M group and ASiR-V40% group.Compared with ASiR-V40% group, DLIR-H group showed lower image noise than ASiR-V40% group(P<0.05),but similar CNR,liver SNR,portal vein SNR and lesion conspicuity to ASiR-V40% group.DLIR-H group showed higher overall image quality than ASiR-V40% group(P<0.05).In the three DLIR groups, the image noise decreased with the increase of strength level, and the overall image quality and lesion conspicuity increased with the increase of strength level, with statistical significance among all groups(P<0.05).The CNR,liver SNR and portal vein SNR were decreased w
作者
刘娜娜
吕培杰
刘星
余娟
王会霞
高剑波
LIU Nana;LYU Peijie;LIU Xing(Department of Radiology,the First Affiliated Hospital of Zhengzhou University,Zhengzhou,Henan Province 450052,P.R.China)
出处
《临床放射学杂志》
北大核心
2022年第9期1683-1688,共6页
Journal of Clinical Radiology
基金
河南省高等学校重点科研项目(编号:22A320057)。
关键词
深度学习重建算法
迭代重建
辐射剂量
图像质量
肝转移
Deep learning image reconstruction
Hybrid model-based adaptive statistical iterative reconstruction
Radiation dose
Image quality
Liver metastasis
