基于头颈部CTA的人工智能辅助诊断系统对颅内动脉瘤的应用价值探讨  

Discussion about the application value of the aided diagnosis system of artificial intelligence based on head and neck CTA for intracranial aneurysms

作　　者：胡斌 荐志洁[1] 张蕴[1] 刘军[1] Hu Bin;Jian Zhijie;Zhang Yun;Liu Jun(Department of Medical Imaging,the First Affiliated Hospital of Xi'an Jiaotong University,Xi'an 710061,Shaanxi Province,China)

机构地区：[1]西安交通大学第一附属医院医学影像科,西安710061

出　　处：《中国数字医学》2022年第1期33-37,26,共6页China Digital Medicine

摘　　要：目的:探究基于头颈部CT血管成像(CTA)的人工智能(AI)诊断系统对颅内动脉瘤的应用价值。方法:回顾性收集2020年1月至2021年3月的57例因临床疑诊动脉瘤而行头颈部CTA的患者。分别通过人工智能诊断系统与影像医师独立人工分析CTA数据,使用卡方检验评价AI与人工分析方法的诊断准确率,通过组内一致性检验(ICC)及Bland-Altman散点图比较AI和人工分析在动脉瘤形态学测量的一致性。结果:共纳入57例患者、456支血管、75个可疑病灶,AI检出阳性患者44例、阳性血管56支、动脉瘤61个,人工检出阳性患者39例、阳性血管50支、动脉瘤51个。AI基于患者、血管及病灶3个层面对动脉瘤诊断准确率分别为63.2%、93.9%、60.0%。卡方检验示AI与人工分析对于患者及病灶层面动脉瘤的检出无统计学差异(P>0.05),而对于血管段层面存在统计学差异(P<0.05);两种方法所得动脉瘤瘤体长径、短径及瘤颈ICC系数分别为0.976、0.972、0.924,95%可信区间分别为0.953~0.988、0.946~0.986、0.832~0.956,Bland-Altman散点图示大部分点集中于1.96倍标准误范围内。结论:基于头颈部CTA的AI诊断系统与人工分析对于临床颅内动脉瘤高危患者的检出能力相当,对于动脉瘤形态学评估的一致性较高,具有推广应用价值。Objective To explore the application value of the diagnostic system of artificial intelligence(AI)based on head and neck CT angiography(CTA)for intracranial aneurysms.Methods 57 patients receiving the head and neck CTA due to clinically suspected aneurysms from January 2020 to March 2021 are collected retrospectively.The CTA data is analyzed independently by the diagnostic system of artificial intelligence and imaging doctors.The accuracy rate of diagnosis of AI and artificial analysis method is evaluated by chi-square test,and the intra-group consistency test(ICC)and Bland Altman scatter diagram are used to define the consistency of AI and artificial analysis in the morphological measurement of aneurysms.Results A total of 57 patients,456 vessels and 75 suspected lesions are included.44 patients,56 vessels and 61 aneurysms are detected to be positive by AI.39 patients,50 vessels and 51 aneurysms are detected to be positive manually.Based on patients,vessels and lesions,the AI’s accuracy rate of diagnosis of aneurysms is 63.2%,93.9%and 60.0%,respectively.The chi-square test shows that there is no significant difference between AI and manual analysis in the detection of aneurysms at the patient and lesion level(P>0.05),but there is significant difference at the level of vascular segment(P<0.05).The long diameter,short diameter and ICC coefficient of the tumor neck of aneurysms obtained with the two methods are 0.976,0.972 and 0.924,respectively.The 95%confidence intervals are 0.953~0.988,0.946~0.986 and 0.832~0.956,respectively.Bland-Altman scatter diagram shows that most of the points are concentrated in the range of 1.96 times of the standard error.Conclusion AI diagnostic system and manual analysis based on head and neck CTA have equal detection capabilities for patients with a high risk of intracranial aneurysms.It has high consistency for the morphological evaluation of aneurysms,and has the value of popularization and application.

关 键 词：颅内动脉瘤 人工智能 头颈部CT血管成像 

分 类 号：R319[医药卫生—基础医学]