基于人工智能的头颈CTA在诊断颅内动脉瘤中的应用价值

Application value of AI-based head and neck CTA in diagnosis of intracranial aneurysm

目的探究基于AI的头颈CTA在诊断颅内动脉瘤中的应用价值。方法回顾性收集右江民族医学院附属医院2021年5月至2022年5月共96例疑似颅内动脉瘤患者的临床资料。分别通过AI后处理技术(AI组)与人工后处理技术(手动组)分析CTA数据,并记录两组后处理时间以及颅内动脉瘤的检出和径线测量结果。以DSA为金标准,计算两组诊断颅内动脉瘤的敏感度和准确率,并分别与DSA动脉瘤测量结果进行比对。结果96例患者中,经DSA确诊71例,共计92个动脉瘤。①后处理时间:手动组为380(342~434)s,AI组为3.05(2.79~3.22)s,差异有统计学意义(P<0.001)。②检出效能:以患者为单位,诊断颅内动脉瘤的敏感度、特异度和准确率,AI组分别为88.73%、80.65%、91.66%,手动组分别为91.55%、80.00%、96.87%;以动脉瘤个数和位置为单位,AI组误诊及漏诊个数多于手动组。③定量分析:与DSA测量数值对比,AI组在瘤颈宽度的测量数值上差异具有统计学意义(P<0.05),而手动组在瘤颈宽度的测量数值差异无统计学意义(P>0.05);手动组与AI组与DSA测量数值对比在瘤体<3 mm的测量数值上差异均具有统计学意义(P<0.05),而对于瘤体长径3~5 mm及>5 mm测量数值差异均无统计学意义(P>0.05)。结论基于AI的头颈CTA在诊断颅内动脉瘤中具有较高的价值,AI技术可作为进一步推广及应用的筛查手段。同时也存在一定的假阳性率和假阴性率,需灵活联合传统人工诊断,从而达到最佳的正确诊断效果。

Objective To explore the application value of AI-based head and neck CTA in the diagnosis of intracranial aneurysm.Methods This study retrospectively collected the clinical data of 96 patients with suspected intracranial aneurysm in The Affiliated Hospital of Youjiang Medical University for Nationalities from May 2021 to May 2022.Their CTA data were analyzed by AI post-processing(AI group)and manual post-processing(manual group),respectively,with the time of post-processing,the detection and diameter measurement of intracranial aneurysms being recorded.With DSA as the gold standard,the sensitivity and accuracy of diagnosing intracranial aneurysms in the two groups were calculated,and were compared with the DSA measurement results for aneurysms.Results Of the 96 patients,71 were diagnosed and confirmed by DSA with a total of 92 aneurysms.①Post-processing time:380(342~434)s in the manual group and 3.05(2.79~3.22)s in the AI group,and the difference was statistically significant(P<0.001).②Detection efficiency:with one patient as a statistical unit,the sensitivity,specificity and accuracy of the diagnosis of intracranial aneurysms were 88.73%,80.65%and 91.66%in the AI group and 91.55%,80.00%and 96.87%in the manual group,respectively.In terms of the number and location of aneurysms,the number of misdiagnosis and missed diagnosis in the AI group was higher than that in the manual group.③Quantitative analysis:compared with DSA measurements,the measured values of tumor neck width were significantly different in the AI group(P<0.05),while no significant difference was found in the manual group(P>0.05).Compared with DSA measurements,there was significant difference in the measured values of tumors<3 mm in the manual group and also in the AI group(P<0.05),but no significant difference was found in measured values of tumors with length 3~5 mm and>5 mm both in the manual group and the AI group(P>0.05).Conclusion AI-based head and neck CTA has a high value in the diagnosis of intracranial aneurysms,and AI technology can be used as a intracranial aneurysm-screening method for further promotion and application.At the same time,this method has a certain false positive rate and false negative rate,so it need to be flexibly combined with traditional manual diagnosis to achieve the best correct diagnosis.