| 陈豫,魏艳磊,王泽尉,等.基于人工智能的冠状动脉计算机断层扫描血管造影在诊断冠状动脉狭窄及提高放射科工作效率中的应用价值[J].安徽医药,2025,29(11):2259-2265. |
| 基于人工智能的冠状动脉计算机断层扫描血管造影在诊断冠状动脉狭窄及提高放射科工作效率中的应用价值 |
| Application value of artificial intelligence-based coronary computed tomography angiography in diagnosing coronary artery stenosis and improving radiology department efficiency |
| |
| DOI：10.3969/j.issn.1009-6469.2025.11.030 |
| 中文关键词: 冠状动脉狭窄 人工智能 冠状动脉 CT血管成像 冠状血管造影术 工作效率 |
| 英文关键词: Coronary stenosis Artificial intelligence Coronary artery computed tomography angiography Coronary angiogra-phy Work efficiency |
| 基金项目: |
|
| 摘要点击次数: 510 |
| 全文下载次数: 171 |
| 中文摘要: |
| 目的评估基于人工智能( AI)辅助的冠状动脉计算机断层扫描血管造影( CCTA)对冠状动脉狭窄程度的诊断价值，并探讨其对放射科工作效率的影响。方法回顾性分析 2019年 1月至 2023年 1月于北京市大兴区中西医结合医院接受 CCTA及侵入性冠状动脉造影( ICA)检查的 76例病人。以 ICA结果作为参考，评估了基于 AI的 CCTA在病人及血管层面诊断 ≥50%狭窄和 ≥70%狭窄的灵敏度、特异度、准确度和受试者操作特征曲线下面积( AUC)。同时，分析了 AI对狭窄病变血管分支的定位能力以及与 ICA检测狭窄程度的相关性和一致性。此外，对比了 AI与不同资历的放射科医生在 CCTA后处理时间和诊断效能上的差异。结果在 76例病人中，共有 52例病人的 83支血管狭窄程度 ≥50%，19例病人的 36支血管狭窄程度 ≥70%。与 ICA相比， AI软件对 ≥50%狭窄的病人显示出 92.3%的灵敏度、 75.0%的特异度，阳性预测值为 88.9%，阴性预测值为 81.8%；对于≥70%狭窄的病人，灵敏度为 89.5%，特异度为 77.2%，阳性预测值为 56.7%，阴性预测值为 95.7%。AI在血管及主要冠脉分支层面上对 ≥50%和≥70%狭窄的诊断效能良好，具备较高的灵敏度、特异度和阴性预测值。 Pearson相关性分析及 Bland-Altman图显示 AI与 ICA评估病人和血管狭窄程度呈显著正相关，两种方法测量一致性良好。 AI软件辅助图像阅读及后处理时间［( 326.82±30.29)s］明显短于中级年资［( 912.06±51.35)s］、高级年资［( 591.73±80.26)s］放射科医师( P<0.05)但其在冠状动脉狭窄诊断方面的能力与经验丰富的放射科医生相当。结论基于 AI的 CCTA对冠状动脉中重度狭窄的诊断性，能良好，在缩短图像处理时间方面具有显著优势，有望作为放射科辅助诊断工具以提高工作效率。 |
| 英文摘要: |
| Objective To evaluate the diagnostic value of artificial intelligence (AI)-assisted coronary computed tomography angiogra-phy (CCTA) for assessing the degree of coronary artery stenosis and to explore its impact on radiology department work efficiency.Meth. ods A retrospective analysis was conducted on 76 patients who underwent both CCTA and invasive coronary angiography (ICA) atBeijing Daxing District Hospital of Integrated Traditional Chinese and Western Medicine from January 2019 to January 2023. UsingICA results as the reference standard, we evaluated the sensitivity, specificity, accuracy, and area under the receiver operating charac-teristic curve (AUC) of AI-assisted CCTA for diagnosing ≥50% and ≥70% stenosis at both patient and vessel level. AI's capability to lo-calize stenotic lesions in vascular branches and its correlation and agreement with ICA for assessing stenosis severity were also ana-lyzed. Additionally, differences in CCTA post-processing time and diagnostic performance between AI and radiologists of different expe-rience levels were compared.Results Among the 76 patients, 52 patients had 83 vessels with ≥50% stenosis, and 19 patients had 36vessels with ≥70% stenosis. Compared with ICA, the AI software demonstrated for ≥50% stenosis at patient level a sensitivity of 92.3%,specificity of 75.0%, positive predictive value of 88.9%, and negative predictive value of 81.8%; for ≥70% stenosis, it showed a sensi-tivity of 89.5%, specificity of 77.2%, positive predictive value of 56.7%, and negative predictive value of 95.7%. AI demonstrated gooddiagnostic performance for both ≥50% and ≥70% stenosis at vessel and major coronary branch levels, with high sensitivity, specificity,and negative predictive values. Pearson correlation analysis and Bland-Altman plots indicated significant positive correlation and goodagreement between AI and ICA in assessing stenosis degree at both patient and vessel levels. AI-assisted image reading and post-pro-cessing time [(326.82±30.29) s] was significantly shorter than that of mid-career [(912.06±51.35) s] and senior [(591.73±80.26) s] radi-ologists (P < 0.05), while its diagnostic capability for coronary artery stenosis was comparable to that of experienced radiologists.Con. clusion AI-assisted CCTA demonstrates good diagnostic performance for moderate to severe coronary artery stenosis and offers signifi-cant advantages in reducing image processing time, showing potential as an auxiliary diagnostic tool in radiology departments to im-prove work efficiency. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
|
|
|
