Takata Kei, Shibutani Naoki, Horita Hirokazu
Department of Radiology, Toyama University Hospital.
Nihon Hoshasen Gijutsu Gakkai Zasshi. 2022 Dec 20;78(12):1399-1405. doi: 10.6009/jjrt.2022-1297. Epub 2022 Oct 27.
The purpose of this study was to avoid the influence of background activity in the calculation of heart to contralateral ratio (H/CL) in Tc-pyrophosphate (PYP) scintigraphy for diagnosis of cardiac amyloidosis. Therefore, we investigated the utility of the H/CL calculation method using ray-summation (ray-sum) images created by multi-planar reconstruction and summing slices of the heart range from single-photon emission computed tomography (SPECT) images.
The subjects were 33 patients who underwent Tc-PYP planar and SPECT/CT at 3 hours after injection. Ray-summation axial (ray-sum axial) and ray-summation coronal (ray-sum coronal) images were created and H/CL was calculated. We compared the differences in sensitivity, specificity, accuracy, and area under the curve (AUC) between the conventional method and our method, and calculated the cutoff values.
Comparison of the conventional method and our method showed no significant difference in sensitivity and AUC, while specificity was significantly improved to 97% (p=0.003) of ray-sum axial image and 90% (p=0.01) of ray-sum coronal image, and accuracy was significantly improved to 94% (p=0.02) of ray-sum axial image.
The H/CL calculation method using ray-sum images had higher diagnostic performance than the conventional method, with optimal cutoff of ray-sum axial images 3.07 and ray-sum coronal images 2.77.
本研究的目的是在用于诊断心脏淀粉样变性的锝-焦磷酸盐(PYP)闪烁扫描中,避免背景活动对心脏与对侧比值(H/CL)计算的影响。因此,我们研究了使用多平面重建创建的射线求和(ray-sum)图像并对单光子发射计算机断层扫描(SPECT)图像中心脏范围的切片进行求和来计算H/CL的方法的实用性。
研究对象为33例在注射后3小时接受Tc-PYP平面和SPECT/CT检查的患者。创建射线求和轴向(ray-sum axial)和射线求和冠状(ray-sum coronal)图像并计算H/CL。我们比较了传统方法和我们的方法在敏感性、特异性、准确性和曲线下面积(AUC)方面的差异,并计算了临界值。
传统方法与我们的方法比较显示,敏感性和AUC无显著差异,而特异性显著提高,射线求和轴向图像的特异性提高到97%(p=0.003),射线求和冠状图像的特异性提高到90%(p=0.01),准确性也显著提高,射线求和轴向图像的准确性提高到94%(p=0.02)。
使用射线求和图像的H/CL计算方法比传统方法具有更高的诊断性能,射线求和轴向图像的最佳临界值为3.07,射线求和冠状图像的最佳临界值为2.77。