Integrating PET and CT information to improve diagnostic accuracy for lung nodules: A semiautomatic computer-aided method.
Integrating PET and CT information to improve diagnostic accuracy for lung nodules: A semiautomatic computer-aided method.
复制标题
DOI：
--
复制
发表时间：
2006-07
期刊：
影响因子：
--
通讯作者：
Y. Nie;Qiang Li;Feng Li;Y. Pu;D. Appelbaum;K. Doi
中科院分区：
文献类型：
--
作者：
Y. Nie;Qiang Li;Feng Li;Y. Pu;D. Appelbaum;K. Doi
我们的目标是开发和评估3种半自动计算机辅助诊断（CAD）方案，通过使用CT、18F-FDG PET以及CT和18F-FDG PET提取的特征来区分肺结节的良恶性。方法我们回顾性收集了92例同时接受胸部CT和全身PET/CT检查的肺结节（<3 cm）患者。经病理检查和临床随访证实，恶性结节42个，良性结节50个。CT和PET检查间隔时间均小于1个月。四个临床参数，包括患者的年龄，性别，吸烟状况，和以前的恶性肿瘤史，用于CAD方案。根据结节的大小、形状、边缘和内部结构，由2名胸部放射科医师独立主观评定16个CT特征。在PET/CT工作站上查看了四个PET特征。然后使用基于临床参数连同CT特征、PET特征以及CT和PET特征的CAD方案来区分良性结节和恶性结节。最后，从CAD方案的输出进行了评估，通过使用接收机工作特性分析。结果当我们使用临床参数和CT特征作为输入单位（CAD方案1）时，CAD方案的受试者工作特征曲线下面积（A（z）值）为0.83。当我们使用临床参数和PET特征作为输入单位（CAD方案2）时，计算机输出的A（z）值为0.91。然而，当我们使用所有数据作为输入单位（CAD方案3）时，计算机输出的A（z）值为0.95。CAD方案3的性能优于CAD方案1或2。CAD方案3和2的A（z）值之间存在统计学显著性差异（P = 0.037），CAD方案3和1的A（z）值之间存在统计学显著性差异（P = 0.015）。结论我们的基于PET和CT的CAD方案比单独基于PET和单独基于CT的CAD方案能够更好地区分肺结节的良恶性。
UNLABELLED Our objective was to develop and evaluate 3 semiautomatic computer-aided diagnostic (CAD) schemes for distinguishing between benign and malignant pulmonary nodules by use of features extracted from CT, 18F-FDG PET, and both CT and 18F-FDG PET. METHODS We retrospectively collected 92 consecutive cases of pulmonary nodules (<3 cm) in patients who underwent both thoracic CT and whole-body PET/CT. Forty-two of the nodules were malignant and 50 benign, as confirmed by pathologic examination and clinical follow-up. The interval between CT and PET was less than 1 mo. Four clinical parameters, including patient age, sex, smoking status, and history of previous malignancy, were used for the CAD schemes. Sixteen CT features based on size, shape, margin, and internal structure of nodules were independently rated subjectively by 2 chest radiologists. Four PET features were viewed on a PET/CT workstation. CAD schemes based on clinical parameters together with CT features, PET features, and both CT and PET features were then used to differentiate benign from malignant nodules. Finally, the output from the CAD schemes was evaluated by use of receiver-operating-characteristic analysis. RESULTS When we used clinical parameters and CT features as input units (CAD scheme 1), the area under the receiver-operating-characteristic curve (A(z) value) of the CAD scheme was 0.83. When we used clinical parameters and PET features as input units (CAD scheme 2), the A(z) value for the computer output was 0.91. However, when we used all data as input units (CAD scheme 3), the A(z) value for the computer output was 0.95. The performance of CAD scheme 3 was better than that of CAD scheme 1 or 2. A statistically significant difference existed between the A(z) values of CAD schemes 3 and 2 (P = 0.037) and between those of CAD schemes 3 and 1 (P = 0.015). CONCLUSION Our CAD scheme based on both PET and CT was better able to differentiate benign from malignant pulmonary nodules than were the CAD schemes based on PET alone and CT alone.