From the Carl E. Ravin Advanced Imaging Laboratories, Department of Radiology, Duke University Medical Center, 2424 Erwin Rd, Suite 302, Durham, NC 27705.
Radiology. 2016 Apr;279(1):185-94. doi: 10.1148/radiol.2015150892. Epub 2015 Dec 1.
To determine if radiation dose and reconstruction algorithm affect the computer-based extraction and analysis of quantitative imaging features in lung nodules, liver lesions, and renal stones at multi-detector row computed tomography (CT).
Retrospective analysis of data from a prospective, multicenter, HIPAA-compliant, institutional review board-approved clinical trial was performed by extracting 23 quantitative imaging features (size, shape, attenuation, edge sharpness, pixel value distribution, and texture) of lesions on multi-detector row CT images of 20 adult patients (14 men, six women; mean age, 63 years; range, 38-72 years) referred for known or suspected focal liver lesions, lung nodules, or kidney stones. Data were acquired between September 2011 and April 2012. All multi-detector row CT scans were performed at two different radiation dose levels; images were reconstructed with filtered back projection, adaptive statistical iterative reconstruction, and model-based iterative reconstruction (MBIR) algorithms. A linear mixed-effects model was used to assess the effect of radiation dose and reconstruction algorithm on extracted features.
Among the 23 imaging features assessed, radiation dose had a significant effect on five, three, and four of the features for liver lesions, lung nodules, and renal stones, respectively (P < .002 for all comparisons). Adaptive statistical iterative reconstruction had a significant effect on three, one, and one of the features for liver lesions, lung nodules, and renal stones, respectively (P < .002 for all comparisons). MBIR reconstruction had a significant effect on nine, 11, and 15 of the features for liver lesions, lung nodules, and renal stones, respectively (P < .002 for all comparisons). Of note, the measured size of lung nodules and renal stones with MBIR was significantly different than those for the other two algorithms (P < .002 for all comparisons). Although lesion texture was significantly affected by the reconstruction algorithm used (average of 3.33 features affected by MBIR throughout lesion types; P < .002, for all comparisons), no significant effect of the radiation dose setting was observed for all but one of the texture features (P = .002-.998).
Radiation dose settings and reconstruction algorithms affect the extraction and analysis of quantitative imaging features in lesions at multi-detector row CT.
在多排 CT 中，确定辐射剂量和重建算法是否会影响肺部结节、肝脏病变和肾结石的基于计算机的定量成像特征的提取和分析。
回顾性分析了一项前瞻性、多中心、符合 HIPAA 规定、机构审查委员会批准的临床试验的数据，对 20 名成人患者（14 名男性，6 名女性；平均年龄 63 岁；范围 38-72 岁）的多排 CT 图像上的 23 个定量成像特征（大小、形状、衰减、边缘锐度、像素值分布和纹理）进行提取，这些患者因已知或疑似局灶性肝脏病变、肺部结节或肾结石而接受检查。数据采集时间为 2011 年 9 月至 2012 年 4 月。所有多排 CT 扫描均在两个不同的辐射剂量水平进行；使用滤波反投影、自适应统计迭代重建和基于模型的迭代重建（MBIR）算法进行图像重建。采用线性混合效应模型评估辐射剂量和重建算法对提取特征的影响。
在所评估的 23 个成像特征中，辐射剂量对肝脏病变、肺部结节和肾结石的 5、3 和 4 个特征有显著影响（所有比较的 P <.002）。自适应统计迭代重建对肝脏病变、肺部结节和肾结石的 3、1 和 1 个特征有显著影响（所有比较的 P <.002）。MBIR 重建对肝脏病变、肺部结节和肾结石的 9、11 和 15 个特征有显著影响（所有比较的 P <.002）。值得注意的是，MBIR 测量的肺部结节和肾结石的大小与其他两种算法显著不同（所有比较的 P <.002）。尽管重建算法显著影响病变纹理（MBIR 影响所有病变类型的 3.33 个纹理特征；所有比较的 P <.002），但除了一个纹理特征外（P =.002-.998），所有辐射剂量设置均无显著影响。
辐射剂量设置和重建算法会影响多排 CT 中病变的定量成像特征的提取和分析。
