人工智能辅助分析CTA影像诊断颅内动脉瘤的临床研究

目的探讨在CT血管造影(CTA)中采用传统人工、人工智能(AI)及两者联合对颅内动脉瘤的检出效能。方法纳入我院疑似颅内动脉瘤并行CTA及数字减影血管造影术(DSA)检查的患者359例,以DSA检查为金标准,分别采用传统人工、AI及两者联合分析CTA图像,计算3种方法对颅内动脉瘤诊断的灵敏度、特异度、假阳性率、假阴性率及准确度,并对传统人工、AI动脉瘤漏诊及误诊的原因进行分析。结果传统人工组、AI组、人工联合AI组对颅内动脉瘤的检出率分别为92.12%、92.12%、95.92%,组间比较差异无统计学意义(P>0.05)。人工联合AI诊断颅内动脉瘤的灵敏度、特异度及准确度分别为95.92%、92.42%及95.39%,高于单一诊断;假阳性率(7.58%)、假阴性率(4.08%)均低于单一诊断。传统人工及AI在毗邻骨质、合并动静脉畸形/烟雾病/夹闭术后复发、非责任动脉瘤方面比较差异有统计学意义(P<0.05)。结论传统人工联合AI对颅内动脉瘤诊断的灵敏度、特异度及准确度较单一方法高,其可降低颅内动脉瘤的漏诊率及误诊率。

Clinical research on diagnosis of intracranial aneurysms by CTA imaging assisted with artificial intelligence

Objective To investigate the detection efficiency of traditional manual method,artificial intelligence(AI)method and the combination of the two methods in diagnosis of intracranial aneurysms by CT angiography(CTA).Methods A total of 359 patients with suspected intracranial aneurysm who underwent CTA and digital subtraction angiography(DSA)examinations were included.With the results of DSA examination as the gold standard,the CTA images were analyzed by traditional manual method,AI method and the combination of the two methods to calculate the sensitivity,specificity,false positive rate,false negative rate and accuracy of the three methods for the diagnosis of intracranial aneurysm.The causes of missed diagnosis and misdiagnosis of traditional manual method and AI method were analyzed.ResultsThe detection rates of intracranial aneurysms in the traditional manual group,the AI group and the combined group were 92.12%,92.12%and 95.92%,respectively,and there was no statistically significant difference between groups(P>0.05).The sensitivity,specificity and accuracy of traditional manual method combined with AI in the diagnosis of intracranial aneurysms were 95.92%,92.42%,and 95.39%,respectively,which were higher than any single diagnosis method.The false positive rate(7.58%)and false negative rate(4.08%)of traditional manual method combined with AI were lower than any single diagnosis method.There were statistically significant differences between traditional manual method and AI method in adjacent bone,combined with arteriovenous malformation/moyamoya disease/recurrence occurred after clipping,and non-responsible aneurysm(P<0.05).Conclusion Traditional manual method combined with AI has higher sensitivity,specificity and accuracy in the diagnosis of intracranial aneurysm than any single diagnosis method,which can reduce the rate of missed diagnosis and misdiagnosis of intracranial aneurysm.