应用颅内动脉瘤辅助评估软件评价未破裂动脉瘤稳定性的多中心前瞻性研究  

A multicenter prospective study using intracranial aneurysm auxiliary assessment software to evaluate the stability of unruptured aneurysms

作　　者：张福峻晖 黄骥良 米热扎提·吐尔洪 刘健[1,2] 李天晓[3] 段传志[4] 李聪慧[5] 于建军[6] 吴鑫[7] 杨新健[1,2] 张鸿祺[8] Zhang Fujunhui;Huang Jiliang;Mirzat Turhon;Liu Jian;Li Tianxiao;Duan Chuanzhi;Li Conghui;Yu Jianjun;Wu Xin;Yang Xinjian;Zhang Hongqi(Beijing Neurosurgical Institute,Capital Medical University,Beijing 100070,China;Neurosurgery Center,Beijing Tiantan Hospital,Capital Medical University,Beijing 100070,China;Department of Neurosurgery,Henan Provincial People′s Hospital,Zhengzhou 450003,China;Department of Neurosurgery,Zhujiang Hospital,Southern Medical University,Guangzhou 510280,China;Department of Neurosurgery,the First Affiliated Hospital of Hebei Medical University,Shijiazhuang 050030,China;Department of Neurosurgery,Linyi People′s Hospital,Linyi 276002,China;Department of Neurosurgery,Yuhuangding Hospital,Yantai 264009,China;Department of Neurosurgery,Xuanwu Hospital,Capital Medical University,Beijing 100053,China)

机构地区：[1]首都医科大学,北京市神经外科研究所,北京100070 [2]首都医科大学附属北京天坛医院神经外科学中心,北京100070 [3]河南省人民医院神经外科,郑州450003 [4]南方医科大学珠江医院神经外科,广州510280 [5]河北医科大学第一附属医院神经外科,石家庄050030 [6]临沂市人民医院神经外科,临沂276002 [7]烟台毓璜顶医院神经外科,烟台264009 [8]首都医科大学宣武医院神经外科,北京100053

出　　处：《中华神经外科杂志》2024年第7期652-658,共7页Chinese Journal of Neurosurgery

基　　金：国家自然科学基金(82072036,82372058)。

摘　　要：目的探讨应用颅内动脉瘤辅助评估软件评价未破裂动脉瘤(UIA)稳定性的效果。方法前瞻性纳入2022年5—9月于首都医科大学附属北京天坛医院等7家中心神经外科就诊的UIA患者。分别采用软件辅助医生评估、医生独立评估、专家组评估、PHASES评分和ELAPSS评分标准评价所有患者的UIA是否稳定。使用分层随机抽样法将患者随机分为A、B两组,A组先由软件辅助医生评估,在4周洗脱期后再由医生独立评估;B组评估顺序相反。将专家组评估结果作为判断UIA稳定性的"金标准",比较动脉瘤辅助评估软件与其他评估方式的准确率、灵敏度、特异度及其与"金标准"的一致性,并根据动脉瘤最大径进行亚组分析。结果共纳入192例患者(227个动脉瘤),其中男62例,女130例;年龄[M(Q1,Q3)]为58.0(50.5,65.0)岁。专家组评估结果显示,227个动脉瘤中,稳定动脉瘤98个,不稳定动脉瘤129个。软件辅助医生评估的准确率和灵敏度(92.5%和93.8%)均高于医生独立评估(76.7%和85.3%)、PHASES评分(62.6%和39.5%)和ELAPSS评分(79.3%和76.7%)评估,差异均具有统计学意义(均P<0.05)。软件辅助医生评估的特异度(90.8%)与PHASES评分(92.9%)和ELAPSS评分(82.7%)评估的差异均无统计学意义(均P>0.05),但高于医生独立评估(65.3%,P<0.05)。一致性方面,分别与"金标准"比较,软件辅助医生评估(Kappa=0.94)、医生独立评估(Kappa=0.81)和ELAPSS评分评估(Kappa=0.83)结果的差异均无统计学意义(均P>0.05),其中软件辅助医生评估与"金标准"的一致性最好;而PHASES评分与"金标准"评估结果的差异具有统计学意义(配对χ^(2)=45.84,P<0.001)。亚组分析结果显示,在动脉瘤最大径<3 mm组、3~5 mm组和>5~7 mm组中,软件辅助医生评估与"金标准"的一致性均较好(Kappa分别为0.94、0.79和0.85),且均高于其他评估方法。软件辅助医生评估和医生独立评估A组与B组的准确率的差异均无统计学意义(均P>Objective To explore the effect of using intracranial aneurysm auxiliary assessment software to evaluate the stability of unruptured aneurysms(UIA).Methods UIA patients who visited the Neurosurgery Department of 7 centers including Beijing Tiantan Hospital,Capital Medical University from May to September 2022 were prospectively included.Software-assisted physician assessment,physician independent assessment,the expert group assessment,PHASES score,and ELAPSS score were used to evaluate whether the aneurysms of all patients were stable.Stratified random sampling was used to randomly divide patients into two groups,A and B.Group A was first evaluated by software-assisted doctors,and then independently evaluated by doctors after a 4-week washout period;group B was evaluated in the opposite order.The evaluation results of the expert group were used as the"gold standard"to judge the stability of UIA,and the accuracy,sensitivity and specificity of the aneurysm auxiliary evaluation software and other evaluation methods were compared.The consistency of these methods with"gold standard"were also compared.Subgroup analysis was performed according to the maximum aneurysm diameter.Results A total of 192 patients(227 aneurysms)were included,including 62 males and 130 females;age[M(Q1,Q3)]was 58.0(50.5,65.0)years old.Among the 227 aneurysms,98 were stable and 129 were unstable.The accuracy and sensitivity of software-assisted physician assessment(92.5%and 93.8%)were higher than those of physician independent assessment(76.7%and 85.3%),PHASES score(62.6%and 39.5%)and ELAPSS score(79.3%and 76.7%),the differences were statistically significant(all P<0.05).The difference between the specificity of software-assisted physician assessment(90.8%)and PHASES score(92.9%)and ELAPSS score(82.7%)was not statistically significant(both P>0.05),but was higher than that of physician independent assessment(65.3%,P<0.05).In terms of consistency,compared with the"gold standard",there were no statistically significant differences in the results of

关 键 词：颅内动脉瘤 人工智能 稳定性评估 诊断 计算机辅助 

分 类 号：R743[医药卫生—神经病学与精神病学]