摘要
目的:对比三低方案结合深度学习图像重建(DLIR)算法肝门静脉血管成像(SCTP)与常规协议结合自适应统计迭代重建Veo(ASiR-V)算法SCTP图像质量。方法:将210例上腹部SCTP检查患者分为4组:标准方案A组(n=90)、低体脂标准方案A1组(n=30)、三低方案B组(n=90)、低体脂超三低方案C组(n=30);其中A组、A1组采用60%权重ASiR-V(A-AV60,A1-AV60),80%权重ASiR-V(A-AV80,A1-AV80)重建;B组、C组:采用60%权重ASiR-V(B-AV60,C-AV60),80%权重ASiR-V(B-AV80,C-AV80),DLIR算法中等级别(B-DM,C-DM),DLIR算法高等级别(B-DH,C-DH)重建。结果:与A组相比,B组的有效剂量、对比剂剂量和注射速度分别减少41.3%(P<0.001)、32.6%(P<0.001)和30.0%(P<0.001),B-DH的标准差(SD)最低,信噪比(SNR)及对比噪声比(CNR)最高,与A-AV80比较,差异无统计学意义。B-DH综合主观评分最高医生1为4.30(4.20,4.40)分,医生2为4.20(4.15,4.40)分(P<0.001),且两位医生的主观评分表现出较强的一致性(Kappa=0.806,P<0.001)。与A1组相比,C组的有效剂量、对比剂剂量和注射速度分别减少64.8%(P<0.001)、36.3%(P<0.001)和30%(P<0.001),A1-AV80的SD值最低(P<0.001),与C-DH比较,差异无统计学意义;A1-AV80的SNR,CNR最高(P<0.001),与C-DH比较无统计学差异;与A1组相比,C-DH综合主观评分最高,医生1为3.80(3.60,4.00)分,医生2为3.80(3.60,4.00)分,且两名医生对总体主观评分具有较强的一致性(Kappa=0.777,P<0.001)。结论:三低方案结合DLIR算法可以降低辐射剂量41.3%、对比剂剂量32.6%、注射速度33.0%并提高图像质量;在低体脂样本中,超三低方案结合DLIR算法可以降低辐射剂量64.8%、对比剂剂量36.3%、注射速度33%并提供不弱于标准协议的图像质量。
Objective:To compare the image quality of hepatic portal vein angiography(SCTP)combined with deep learning image reconstruction(DLIR)algorithm and conventional protocol combined with adaptive statistical iterative reconstruction Veo(ASiR-V)algorithm.Methods:A total of 210 patients with upper abdominal SCTP were divided into 4 groups:standard regimen A(n=90),low body fat standard regimen A1(n=30),triple low regimen B(n=90),and low body fat over triple low regimen C(n=30).Group A and group A1 were reconstructed with 60%and 80%weighting ASiR-V(A-AV60,A1-AV60,A-AV80,A1-AV80).Group B and group C:B-AV60,C-AV60,B-AV80,CAV8O,medium-level DLIR algorithm(B-DM,C-DM),high-level DLIR algorithm(B-DH,C-DH)were used for reconstruction.Results:Compared with group A,the effective dose,contrast dose and injection velocity of group B were decreased by 41.3%(P<0.001),32.6%(P<0.001)and 30%(P<0.001),respectively,and B-DH had the lowest Standard deviation(SD).Signal-to-noise ratio(SNR)and Contrast noise ratio(CNR)were the highest,and the difference was not statistically significant when compared with A-AV80.The comprehensive subjective score of B-DH was the highest 4.30(4.20,4.40)for doctor 1 and 4.20(4.15,4.40)for doctor 2(P<0.001),and the subjective score of the two doctors showed a strong agreement(Kappa=0.806,P<0.001).Compared with group A1,the effective dose,contrast dose and injection speed of group C were reduced by 64.8%(P<0.001),36.3%(P<0.001)and 30%(P<0.001),respectively.The SD value of A1-AV80 was the lowest(P<0.001),and there was no statistical significance when compared with C-DH.The SNR and CNR of A1-AV80 were the highest(P<0.001),and there was no statistical difference between A1-AV80 and C-DH.Compared with A1 group,the overall subjective score of C-DH was the highest 3.80(3.60,4.00)for Doctor 1 and 3.80(3.60,4.00)for Doctor 2,and there was strong agreement between the two doctors on the overall subjective score(Kappa=0.777,P<0.001).Conclusion:The combination of three-low protocol and DLIR algorithm can reduce radiation dose by 41.3%,contrast dose by 32.6%,injection speed by 33%,and improve image quality.In low body fat samples,the combination of super three-low scheme and DLIR algorithm can reduce radiation dose by 64.8%,contrast agent dose by 36.3%and injection speed by 33%,while providing image quality no worse than standard protocol.
作者
张可
张悦
邓锶锶
郭月飞
谢斯栋
孟占鳌
ZHANG Ke;ZHANG Yue;DENG Sisi;GUO Yuefei;XIE Sidong;MENG Zhanao(Department of Radiology,the Third Affiliated Hospital of Sun Yat-sen University,Guangzhou 501630,Guangdong,China)
出处
《暨南大学学报(自然科学与医学版)》
CAS
北大核心
2023年第3期305-315,共11页
Journal of Jinan University(Natural Science & Medicine Edition)
基金
广州市科技计划项目(202007030007)
中山大学附属第三医院国家自然科学基金培育专项项目(2021GZRPYM506)
中山大学附属第三医院“五个五”工程项目(2023WW605)。
关键词
深度学习
图像重建
自适应迭代统计重建-V
三低方案
肝门静脉血管成像
deep learning
image reconstruction
adaptive iterative statistical reconstruction-V
three-low protocol
hepatic portal venography(SCTP)