Fujita Masayo, Higaki Toru, Awaya Yoshikazu, Nakanishi Toshio, Nakamura Yuko, Tatsugami Fuminari, Baba Yasutaka, Iida Makoto, Awai Kazuo
Department of Diagnostic Radiology, Hiroshima University, Kasumi 1-2-3, Minami-ku, Hiroshima, Hiroshima, 734-8551, Japan.
Department of Internal Medicine, Miyoshi Central Hospital, 531 Sakaya-cho, Miyoshi, Hiroshima, 728-0023, Japan.
Jpn J Radiol. 2017 Apr;35(4):179-189. doi: 10.1007/s11604-017-0618-y. Epub 2017 Feb 14.
To investigate the diagnostic capability of ultra-low-dose CT (ULDCT) with full iterative reconstruction (f-IR) for lung cancer screening.
All underwent ULDCT and/or low-dose CT (LD-CT) on a 320-detector scanner. ULDCT images were reconstructed with f-IR. We qualitatively and quantitatively studied 95 nodules in 69 subjects. Two radiologists classified the nodules on ULDCT images as solid-, part-solid-, and pure ground-glass (PGG) and recorded their mean size. Their findings were compared with the reference standard. The observer performance study included 7 other radiologists and 35 subjects with- and 15 without nodules. The results were analyzed by AFROC analysis.
In the qualitative study, the kappa values between observers 1 and 2, respectively, and the reference standard were 0.70 and 0.83; the intra-class correlation coefficients for the nodule diameter between the reference standard and their measurements were 0.84 and 0.90. The 95% confidence interval (CI) for the area under the curve (AUC) difference for nodule detection on LDCT and ULDCT was -0.03 to 0.07. The 95% CI crossed the 0 difference in the AUC but not the pre-defined non-inferiority margin of -0.08.
The diagnostic ability of ULDCT using f-IR is comparable to LDCT.
探讨采用全迭代重建（f-IR）的超低剂量CT（ULDCT）用于肺癌筛查的诊断能力。
所有受试者均在320排探测器扫描仪上接受ULDCT和/或低剂量CT（LD-CT）检查。ULDCT图像采用f-IR重建。我们对69名受试者中的95个结节进行了定性和定量研究。两名放射科医生将ULDCT图像上的结节分类为实性、部分实性和纯磨玻璃（PGG）结节，并记录其平均大小。将他们的结果与参考标准进行比较。观察者性能研究纳入了另外7名放射科医生以及35名有结节和15名无结节的受试者。结果通过AFROC分析进行分析。
在定性研究中，观察者1和观察者2分别与参考标准之间的kappa值为0.70和0.83；参考标准与其测量的结节直径的组内相关系数分别为0.84和0.90。LDCT和ULDCT上结节检测的曲线下面积（AUC）差异的95%置信区间（CI）为-0.03至0.07。95%CI跨越了AUC的0差异，但未跨越预先定义的非劣效性界值-0.08。
采用f-IR的ULDCT的诊断能力与LDCT相当。
