1 Department of Radiology, Samsung Medical Center, Sungkyunkwan University School of Medicine, 50 Ilwon-dong, Gangnam-Gu, Seoul 135-710, Republic of Korea.
AJR Am J Roentgenol. 2014 Apr;202(4):730-7. doi: 10.2214/AJR.13.11419.
The purpose of this study was to evaluate the performance of an automated computer-assisted detection (CAD) algorithm to detect coronary artery stenosis on coronary CT angiography (CTA).
We investigated 128 consecutive patients (76 men, 52 women; mean [SD] age, 64 ± 11 years) who had acute chest pain and underwent 128-slice dual-source coronary CTA and invasive coronary angiography at an emergency department. All coronary CTA data were analyzed using customized software for the detection of coronary artery stenosis without human interaction. The diagnostic performance of a CAD algorithm for evaluation of stenosis of at least 50% of vessel diameter was compared with that of human interpretation of coronary CTA, with invasive coronary angiography as a reference standard.
Of the 128 patients, 25 patients were excluded because of failure of data processing (n = 9) or history of stent insertion or coronary artery bypass graft (n = 16). Invasive coronary angiography revealed significant stenosis in 62% (64/103) of the remaining patients. In detecting significant stenosis, the CAD algorithm yielded 100% sensitivity, 23.1% specificity, 68.1% positive predictive value (PPV), and 100% negative predictive value (NPV) in per-patient analysis. On per-vessel analysis, the CAD algorithm yielded 90.0% sensitivity, 62.4% specificity, 40.1% PPV, and 95.7% NPV. Human interpretation of coronary CTA yielded 98.4% and 96.7% sensitivities, 79.5% and 95.0% specificities, 88.7% and 84.5% PPVs, and 96.9% and 99.0% NPVs for diagnosing significant stenosis on per-patient and per-vessel analyses, respectively.
The CAD algorithm yields a high NPV in detecting stenosis of at least 50% on coronary CTA. As a second "reader," the CAD algorithm may help to exclude significant coronary stenosis in patients with acute chest pain at an emergency department.
本研究旨在评估一种自动计算机辅助检测（CAD）算法在冠状动脉 CT 血管造影（CTA）中检测冠状动脉狭窄的性能。
我们调查了 128 例连续患者（76 例男性，52 例女性；平均[标准差]年龄 64±11 岁），这些患者在急诊科因急性胸痛而行 128 层双源冠状动脉 CTA 和有创冠状动脉造影。所有冠状动脉 CTA 数据均使用定制软件进行分析，无需人工交互即可检测冠状动脉狭窄。CAD 算法评估至少 50%血管直径狭窄的诊断性能与冠状动脉 CTA 的人工解读进行比较，以有创冠状动脉造影为参考标准。
在 128 例患者中，有 25 例患者因数据处理失败（9 例）或支架置入或冠状动脉旁路移植术史（16 例）而被排除在外。在其余的 103 例患者中，有 62%（64/103）的患者有明显的狭窄。在检测显著狭窄时，CAD 算法在患者分析中产生了 100%的敏感性、23.1%的特异性、68.1%的阳性预测值（PPV）和 100%的阴性预测值（NPV）。在血管分析中，CAD 算法的敏感性为 90.0%，特异性为 62.4%，PPV 为 40.1%，NPV 为 95.7%。冠状动脉 CTA 的人工解读在患者和血管分析中分别产生了 98.4%和 96.7%的敏感性、79.5%和 95.0%的特异性、88.7%和 84.5%的 PPV 和 96.9%和 99.0%的 NPV，用于诊断显著狭窄。
CAD 算法在冠状动脉 CTA 上检测至少 50%的狭窄具有较高的 NPV。作为第二个“读者”，CAD 算法可以帮助排除急诊科急性胸痛患者的严重冠状动脉狭窄。
