摘要
目的:探讨人工智能(AI)在冠状动脉CT血管成像(CCTA)中诊断冠状动脉狭窄的准确性及应用价值。方法:收集2019年4月至10月110例同时行CCTA及有创冠状动脉造影(ICA)病人的影像资料,110例共1484段血管纳入评价范围。狭窄程度分为无狭窄、轻度狭容(<50%)、中度狭(50%~70%)重度狭窄(>70%).AI软件自动对CCTA图像进行重建及计算分析。以ICA结果为金标准,计算AI在CCTA中诊断冠状动脉狭容的敏感度、特异度、阳性预测值及阴性预测值。对AI与ICA结果进行Kappa值一致性检验。结果:①AI检出冠状动脉狭窄的敏感度、特异度、阳性预测值及阴性预测值分别为92.97%.97.91%,88.53%,96.36%,准确性为93.60%,AI与ICA检出冠状动脉狭窄一致性好(Kappa值0.86).②AI诊断冠状动脉狭窄程度准确性为66.13%,与ICA一致性一般(Kappa值0.58)。诊断轻度狭窄准确性较高,诊断中重度狭窄特异度较高。结论:AI在CCTA中对冠状动脉狭窄节段的检出及诊断轻度狭窄具有较高准确性,诊断中重度狭窄特异度较高,可作为医师辅助诊断手段.
Purpose:To evaluate the accuracy and application value of artificial intelligence in the diagnosis of coronary artery stenosis in CCTA.Methods:Imaging data of 110 patients undergoing CCTA and invasive coronary angiography from April 2019 to October 2019 were collected.A total of 1484 segments were included in 110 patients.The degree of stenosis was divided into no stenosis,mild stenosis(<50%),moderate stenosis(50%〜70%)and severe stenosis(>70%).CCTA images were automatic reconstructed,calculated and analyze of by artificial intelligence software automatically.Based on the golden standard of ICA,the sensitivity,specificity,positive and negative predictive value of AI in the diagnosis of coronary artery stenosis in CCTA were calculated.The agreement between AI and ICA results was analyzed.Results:①The sensitivity,specificity,positive and negative predictive value,accuracy of AI in detecting coronary artery stenosis were 92.97%,97.91%,88.53%,96.36%and 93.60%,respectively.The consistency between AI and ICA in detecting coronary artery stenosis was well(Kappa value 0.86).②The accuracy of AI in the diagnosis of coronary artery stenosis was 66.13%.The consistency between AI and ICA in diagnosing coronary artery stenosis was good(Kappa value 0.58).Conclusion:AI has high accuracy in the detection and diagnosis of mild coronary artery stenosis in CCTA,and high specificity in the diagnosis of moderate and severe stenosis.It can be used as an auxiliary diagnostic method for physicians.
作者
李浚利
韩丹
段慧
黄益龙
闵蕊
蔡雅倩
张正华
LI Jun-li;HAN Dan;DUAN Hui;HUANG Yi-long;MIN Rui;CAI Ya-qian;ZHANG Zheng-hua(Department of Medical Imaging,First Affiliated Hospital of Kunming Medical University.)
出处
《中国医学计算机成像杂志》
CSCD
北大核心
2020年第2期120-124,共5页
Chinese Computed Medical Imaging
基金
云南基础研究计划昆医联合专项基金项目(2018FE001-208)。
关键词
人工智能
冠状动脉CT血管成像
狭窄
诊断
Artificial intelligence
Coronary computed tomography angiography
Stenosis
Diagnosis