深度学习重建算法结合低管电压技术提高常规腹部增强CT图像小血管显示水平的价值  

Value of the deep learning image reconstruction algorithm combined with low-kV technique in improving small vessel display level at routine abdominal contrast enhanced CT imaging

作　　者：陈楚韩 綦维维[1] 刘晓怡[1] 陈雷[1] 程瑾[1] 洪楠[1] Chen Chuhan;Qi Weiwei;Liu Xiaoyi;Chen Lei;Cheng Jin;Hong Nan(Department of Radiology,Peking University People′s Hospital,Beijing 100044,China)

机构地区：[1]北京大学人民医院放射科,北京100044

出　　处：《中华放射学杂志》2022年第11期1168-1174,共7页Chinese Journal of Radiology

基　　金：国家自然科学基金(81901819);北京市自然科学基金(7202217);北京大学医学部大学生创新实验项目。

摘　　要：目的评价深度学习重建(DLIR)算法结合低管电压技术相对于自适应统计迭代重建(ASiR-V)算法对腹部增强CT动脉晚期图像中微小动脉的显示情况。方法前瞻性收集2021年12月至2022年1月就诊于北京大学人民医院需排查腹部疾病并接受腹盆腔增强CT扫描的患者。根据CT扫描不同管电压以简单随机法将患者分为80 kV低管电压(LV)组和120 kV高管电压(HV)组。根据不同重建算法, 每组再进一步分为DLIR-H(D)和ASiR-V 50%(A)2个亚组。对患者进行CT增强扫描时均采用自动管电流调节技术, 噪声指数统一设置为9。对重建层厚为0.625 mm的动脉晚期图像进行主观评价和客观指标评价, 并记录动脉晚期扫描的辐射剂量。结果共纳入168例患者, 其中男76例, 女92例, 年龄18~85(53±15)岁, 体质指数(24±3)kg/m^(2);LV组91例、HV组77例。LV组主动脉和肝总动脉CT值明显高于HV组(t=-14.20, P<0.001;t=-0.95, P<0.001)。管电压相同时, D亚组动脉晚期图像噪声明显低于A亚组, 肝脏、主动脉、肝总动脉信噪比(SNR)和对比噪声比(CNR)均明显高于A亚组(P均<0.001)。LV-D亚组中主动脉和肝总动脉SNR和CNR明显优于LV-A、HV-D、HV-A亚组(P均<0.001)。腹部血管显示的主观评价方面, LV-D亚组肝总动脉、肠系膜下动脉、左结肠动脉升支边缘的空间分辨力以及左结肠动脉升支对比度均显著优于LV-A、HV-D、HV-A亚组(P<0.05), 脾区边缘动脉显示率(54.9%, 50/91)显著优于HV-D(24.7%, 19/77)和HV-A亚组(32.5%, 25/77)(校正后P<0.05)。LV组与HV组间有效辐射剂量分别为(4.91±1.97)、(5.43±1.78)mSv, 差异无统计学意义(P>0.05)。结论低管电压技术结合DLIR算法的腹部增强CT扫描能够有效提升肠系膜下动脉发出的左结肠动脉升支及脾区边缘动脉的显示水平, 为类似微小血管的评估带来更多的可能性。Objective To evaluate the presentation of small arteries in abdominal contrast-enhanced CT late-arterial images using the deep learning image reconstruction(DLIR)combined with low tube voltage(kV)technique relative to the adaptive statistical iterative reconstruction V(ASiR-V)algorithm.Methods Patients who were admitted to Peking University People′s Hospital from December 2021 to January 2022 and needed to be screened for abdominal diseases and receive abdominal and pelvic contrast-enhanced CT scan were prospectively collected.The patients were divided into low-voltage(LV)with 80 kV and high-voltage(HV)with 120 kV groups.According to two different reconstruction algorithms,each group was further divided into DLIR-H(D)subgroup and ASiR-V 50%(A)subgroup.The automatic tube current adjustment technique was used for CT enhanced scanning of patients,and the noise index value was uniformly set to 9.Subjective and objective evaluations were performed on the late-arterial images with a constructed slice thickness of 0.625 mm,and the radiation doses were recorded.Results A total of 168 patients were included,including 76 males and 92 females,aged 18-85(53±15)years old,body mass index(24±3)kg/m2;91 patients in the LV group and 77 in the HV group.The CT values of the aorta and common hepatic artery in the LV group were significantly higher than those in the HV group(t=-14.20,P<0.001;t=-0.95,P<0.001).When the tube voltage was the same,the late-arterial image noise in subgroup D was significantly lower than that in subgroup A,and the signal to noise ratio(SNR)and contrast to noise ratio(CNR)of the liver,aorta and common hepatic artery were significantly higher than those in subgroup A(all P<0.001).The SNR and CNR of the aorta and common hepatic artery in the LV-D subgroup were significantly better than those in the LV-A,HV-D,and HV-A subgroups(all P<0.001).In the subjective evaluation of abdominal vascular display,the special resolution of the common hepatic artery,inferior mesenteric artery and the edge of the ascending b

关 键 词：体层摄影术 X线计算机 深度学习 图像重建算法 微小血管 

分 类 号：R816.5[医药卫生—放射医学]