VMI+联合人工智能辅助诊断系统对颅内动脉瘤诊断价值的研究  

Research of Dual Energy CT VMI+Technology Combined with Artificial Intelligence-Assisted Diagnosis System in the Diagnosis of Intracranial Aneurysm

作　　者：韦鑫 陈维娟[1] 邓昊[1] 余菡 曹闻挺[1] 陈金华 刘杨[2] WEI Xin;CHEN Weijuan;DENG Hao;YU Han;CAO Wenting;CHEN Jinhua;LIU Yang(Department of Radiology,The Second Affiliated Hospital of Chongqing Medical University,Chongqing 400010,China;Department of Medical Imaging Technology,Chongqing Medical University,Chongqing 400010,China)

机构地区：[1]重庆医科大学附属第二医院放射科,重庆400010 [2]重庆医科大学医学影像技术系,重庆400010

出　　处：《中国医疗设备》2021年第10期100-103,107,共5页China Medical Devices

摘　　要：目的探讨双能CT虚拟单能量成像(Virtual Monoenergetic Imaging Plus,VMI+)技术联合人工智能(Artificial Intelligence,AI)在头颈CT血管成像(CT Angiography,CTA)颅内动脉瘤诊断中的应用价值。方法重建41例临床诊断疑似颅内动脉瘤的头颈CTA数据并获得1组标准线性融合(F_(0.6))图像以及9组VMI+(虚拟能级为40~120 keV,间隔为10 keV)图像。测量10组图像颈总动脉分叉处层面动脉血管的CT值和图像噪声,计算信噪比(Signal to Noise Ratio,SNR)和对比噪声比(Contrast to Noise Ratio,CNR),并比较标准线性融合(F_(0.6))和不同能级VMI+图像的差异。以2名高年资医师一致的诊断意见为参考标准,评价不同虚拟能级下AI对颅内动脉瘤的诊断符合率并筛选出最佳诊断能级。结果AI在40 keV、110 keV、120 keV组诊断符合率分别为51%、64%、60%,分别与高年资医师相比,以上差异均具有统计学差异(P<0.05)。在50~100 keV、F_(0.6)组的符合率分别为78%、73%、76%、71%、76%、64%、71%,分别与高年资医师相比,以上差异均无统计学意义(P>0.05),且50 keV时ROC曲线下面积最大,为0.796(P=0.001)。与F_(0.6)组相比,VMI+图像中40 keV组和50 keV组CT值、SNR和CNR均明显高于F_(0.6)组,其中50 keV组分别提高了92.47%、18.65%和11.87%。70~120 keV组CT值、SNR和CNR均低于F_(0.6)组,以上差异均具有统计学意义(P<0.05);60 keV组CT值显著高于F_(0.6)组,差异具有统计学意义(P<0.05),SNR和CNR与F_(0.6)组之间的差异无统计学意义(P=0.084、0.399)。结论AI辅助诊断系统对颅内动脉瘤的诊断符合率较高,虚拟单能量成像技术在50~100 keV能级范围能确保AI对动脉瘤的有效诊断,低能级重建可以显著提高CTA图像质量,最佳能级推荐为50 keV。Objective To explore the clinical value of dual-energy CT virtual monoenergetic imaging plus(VMI+)technology combined with artificial intelligence(AI)in the diagnosis of intracranial aneurysms in head and neck CT angiography(CTA).Methods A total of 41 cases of head and neck CTA data with clinical diagnosis as suspected intracranial aneurysms in our hospital were reconstructed and we got 1 set of standard linear fusion(F_(0.6))images and 9 sets of VMI+(virtual energy level of 40~120 keV,interval of 10 keV)images.We measured the CT value of arterial vessels and image noise at the bifurcation of the common carotid artery in the 10 sets of images,calculated the signal-to-noise ratio(SNR),contrast-to-noise ratio(CNR),and compared the differences of standard linear fusion(F_(0.6))and VMI+images.Based on the consistent diagnosis opinions of two senior physicians as the reference standard,the diagnostic coincidence rate of intracranial aneurysms by AI at different energy levels was evaluated and screen out the best diagnostic energy level.Results The diagnostic coincidence rates of AI in 40 keV group,110 keV group and 120 keV group were 51%,64%and 60%,respectively.Compared with senior doctors,the above differences were statistically significant(P<0.05).The coincidence rates of 50~100 keV and F_(0.6)groups were 78%,73%,76%,71%,76%,64%and 71%respectively(P>0.05).Compared with senior doctors,the above differences were not statistically significant(P>0.05),and the area under ROC curve was the largest at 50 keV(0.796,P=0.001).Compared with F_(0.6)group,the CT values,SNR and CNR of the 40 keV and 50 keV groups in the VMI+image were significantly higher than F_(0.6)group,the 50 keV group were increased by 92.47%,18.65%and 11.87%.And those of the 70~120 keV groups were lower than F_(0.6)group,and the differences were statistically significant(P<0.05).The CT value of 60 keV group was significantly higher than F_(0.6)group,the difference was statistically significant(P<0.05).But there was no statistically significant difference in

关 键 词：双能CT 虚拟单能量成像 人工智能 头颈CT血管成像 颅内动脉瘤 

分 类 号：R543.5[医药卫生—心血管疾病] R814.42[医药卫生—内科学]