Department of Diagnostic Radiology, Faculty of Life Sciences, Kumamoto University, 1-1-1 Honjyo, Kumamoto 860-8556, Japan.
Eur J Radiol. 2013 Feb;82(2):288-95. doi: 10.1016/j.ejrad.2012.11.003. Epub 2012 Dec 6.
To investigate the diagnostic performance of 256-slice cardiac CT for the evaluation of the in-stent lumen by using a hybrid iterative reconstruction (HIR) algorithm combined with a high-resolution kernel.
This study included 28 patients with 28 stents who underwent cardiac CT. Three different reconstruction images were obtained with: (1) a standard filtered back projection (FBP) algorithm with a standard cardiac kernel (CB), (2) an FBP algorithm with a high-resolution cardiac kernel (CD), and (3) an HIR algorithm with the CD kernel. We measured image noise and kurtosis and used receiver operating characteristics analysis to evaluate observer performance in the detection of in-stent stenosis.
Image noise with FBP plus the CD kernel (80.2 ± 15.5 HU) was significantly higher than with FBP plus the CB kernel (28.8 ± 4.6 HU) and HIR plus the CD kernel (36.1 ± 6.4 HU). There was no significant difference in the image noise between FBP plus the CB kernel and HIR plus the CD kernel. Kurtosis was significantly better with the CD- than the CB kernel. The kurtosis values obtained with the CD kernel were not significantly different between the FBP- and HIR reconstruction algorithms. The areas under the receiver operating characteristics curves with HIR plus the CD kernel were significantly higher than with FBP plus the CB- or the CD kernel. The difference between FBP plus the CB- or the CD kernel was not significant. The average sensitivity, specificity, and positive and negative predictive value for the detection of in-stent stenosis were 83.3, 50.0, 33.3, and 91.6% for FBP plus the CB kernel, 100, 29.6, 40.0, and 100% for FBP plus the CD kernel, and 100, 54.5, 40.0, and 100% for HIR plus the CD kernel.
The HIR algorithm combined with the high-resolution kernel significantly improved diagnostic performance in the detection of in-stent stenosis.
使用混合迭代重建（HIR）算法结合高分辨率内核，研究 256 层心脏 CT 评估支架内管腔的诊断性能。
本研究纳入 28 例 28 个支架的患者，进行心脏 CT 检查。分别使用以下三种不同的重建图像：（1）标准滤波反投影（FBP）算法结合标准心脏内核（CB），（2）FBP 算法结合高分辨率心脏内核（CD），（3）HIR 算法结合 CD 内核。我们测量图像噪声和峰度，并使用受试者工作特征分析评估观察者在检测支架内狭窄方面的性能。
FBP 加 CD 内核的图像噪声（80.2±15.5 HU）明显高于 FBP 加 CB 内核（28.8±4.6 HU）和 HIR 加 CD 内核（36.1±6.4 HU）。FBP 加 CB 内核和 HIR 加 CD 内核的图像噪声无显著差异。CD 内核的峰度明显更好。CD 内核的峰度值在 FBP 和 HIR 重建算法之间无显著差异。HIR 加 CD 内核的受试者工作特征曲线下面积明显高于 FBP 加 CB 或 CD 内核。FBP 加 CB 或 CD 内核之间的差异无统计学意义。检测支架内狭窄的平均灵敏度、特异性、阳性预测值和阴性预测值分别为 FBP 加 CB 内核 83.3%、50.0%、33.3%和 91.6%，FBP 加 CD 内核 100%、29.6%、40.0%和 100%，HIR 加 CD 内核 100%、54.5%、40.0%和 100%。
HIR 算法结合高分辨率内核可显著提高支架内狭窄检测的诊断性能。
