人工智能技术辅助分析头颈部CTA的应用价值研究  

Research on the Application Value of Artificial Intelligence Technology Assisted Analysis of Head and Neck CTA

作　　者：韦家旭 所世腾[1] 李传争 唐强强 郭鹏 王嵇[1] WEI Jia-xu;SUO Shi-teng;LI Chuan-zheng;TANG Qiang-qiang;GUO-peng;WANG-Ji(Department of Radiology,Renji Hospital,Shanghai Jiao Tong University School of Medicine,Shanghai 200127,China)

机构地区：[1]上海交通大学医学院附属仁济医院放射科,上海200127

出　　处：《中国CT和MRI杂志》2024年第8期159-161,共3页Chinese Journal of CT and MRI

摘　　要：目的以有创动脉造影(digital subtraction angiography,DSA)为参考标准,通过与手动后处理法比较,探讨在头颈部C T血管成像(ct angiography,CTA)后处理过程中人工智能(artificial intelligent,AI)辅助分析的应用价值。方法回顾性分析2020年4月-2022年9月在本院行头颈动脉CTA并于1月内行头颈动脉造影检查且符合纳入标准的42例患者的影像资料,男30例,女12例,平均年龄(67±10.9)岁。通过与手工法图像后处理进行比较,评价AI法后处理的工作效率、VR图像质量、标准符合率以及对颈动脉狭窄程度的诊断符合率。结果对42例患者的84条双侧颈动脉进行分析,人工智能平均完成时间为(1.23±0.17)min,比手工方法(7.95±2.42)min缩短约6.72min,平均时间增益率为84%。人工智能获得的VR图像质量与一般图像质量之比分别为88.1%(37/42)和9.5%(4/42),手动方法获得的VR图像质量与一般图像质量之比分别为19.0%(8/42)和69.0%(29/42)。两种方法获得的VR图像质量主观评分差异有统计学意义(P<0.001);以DSA为“金标准”,其中正常33根,闭塞10根,狭窄41根;在检测闭塞上手工和AI测量准确率均100%,狭窄率手工及人工智能测量结果分别为49.7%±20.9%、51.9%±26.3%,二者无显著性差异(P=0.24),组内相关系数为0.879;结论基于图像分割技术对头颈部CTA血管成像进行后处理,可以显著缩短后处理时间,提高诊断的准确性和效率。此外,VR后处理的图像质量优于手工方法,诊断血管闭塞和狭窄的准确性与手工方法相当,可以更好地辅助影像医生进行诊断。Objective Using digital subtraction angiography(DSA)as the reference standard,by comparing with manual post-processing,To explore the application value of artificial intelligent(AI)assisted analysis in the post-processing of head and neck ct angiography(CTA).Methods The imaging data of 42 patients who received head and carotid artery CTA in our hospital from April 2020 to September 2022 and underwent head and carotid artery angiography within 1 month and met the inclusion criteria were retrospectively analyzed,including 30 males and 12 females,with an average age of(67±10.9)years.By comparing with manual image post-processing,the work efficiency,VR image quality,standard coincidence rate and diagnosis coincidence rate of carotid artery stenosis degree of AI method were evaluated.Results 84 bilateral carotid arteries of 42 patients were analyzed.The average completion time of artificial intelligence was(1.23±0.17)min,which was about 6.72min shorter than that of manual method(7.95±2.42)min,and the average time gain rate was 84%.The ratio of VR image quality to general image quality obtained by artificial intelligence was 88.1%(37/42)and 9.5%(4/42),and the ratio of VR image quality to general image quality obtained by manual methods was 19.0%(8/42)and 69.0%(29/42),respectively.The subjective scores of VR image quality obtained by the two methods were statistically significant(P<0.001).Using DSA as the gold standard,33 were normal,10 were occluded and 41 were stenosis.The accuracy of manual and AI measurement was 100%,and the stenosis rate was 49.7%±20.9%and 51.9%±26.3%,respectively,with no significant difference(P=0.24),and the intra-group correlation coefficient was 0.879.Conclusion Post-processing of head and neck CTA angiography based on image segmentation technology can significantly shorten the post-processing time and improve the accuracy and efficiency of diagnosis.In addition,the image quality of VR postprocessing is superior to manual methods,and the accuracy of diagnosing blood vessel occlusion and st

关 键 词：人工智能 头颈CTA CT血管成像 图像后处理 

分 类 号：R323.1[医药卫生—人体解剖和组织胚胎学]