| 基于CT表现的孤立性肺结节良恶性预测模型的研究 |
| |
| 引用本文： | 易芹芹,周宙,黄国鑫?.基于CT表现的孤立性肺结节良恶性预测模型的研究[J].CT理论与应用研究,2019,28(6):677-683. |
| |
| 作者姓名： | 易芹芹 周宙 黄国鑫? |
| |
| 作者单位： | 深圳市人民医院放射科,广东深圳 518001;深圳市人民医院放射科,广东深圳 518001;深圳市人民医院放射科,广东深圳 518001 |
| |
| 基金项目： | 深圳市科技计划项目（JCYJ20150403101028197）。 |
| |
| 摘 要： | 目的:筛选并分析影响肺结节良恶性的因素，建立预测模型、验证该模型并与梅奥模型、Brock模型对比。方法:回顾性分析2015年1月至2017年12月深圳市人民医院有病理结果的孤立性肺结节病例319例，其中229例作为建模组（A组），90例作为验证组（B组），分析A组病例性别、年龄、直径、吸烟史、毛刺、位于上叶、边界不清楚、分叶征、空泡征、血管集束征、胸膜凹陷征、含磨玻璃密度及钙化，通过单因素分析及Logistic回归分析，筛选出独立影响因子，并建立回归方程。用B组资料进行验证并将B组资料分别代入本研究模型、梅奥模型及Brock模型进行对比。结果:单因素分析示年龄、直径、毛刺、上叶、边界不清楚、分叶、空泡、血管集束征、胸膜凹陷征、是否含有磨玻璃密度在良恶性结节中的差异具有统计学意义（P<0.05），Logistic回归分析示有毛刺、有分叶、边界不清楚和含有磨玻璃密度为恶性孤立性肺结节的独立影响因素，并据此建立的回归方程ROC曲线下面积为0.894，其灵敏度为91.3％，特异度为77.3％，阳性似然比为4.02，阴性似然比为0.11，阳性预测值为80.8％，阴性预测值为89.5％；本研究模型与梅奥模型的差异有统计学意义（P=0.0049），与Brock模型差异没有统计学意义（P=0.79）。结论:有毛刺、有分叶、边界不清楚和含有磨玻璃密度为恶性孤立性肺结节的独立影响因素，据此建立的回归方程具有较高的诊断效能。本研究建立的模型诊断效能优于梅奥模型，与Brock模型诊断效能相当。
|
| 关 键 词： | 孤立性肺结节 CT logistic回归分析 |
| 收稿时间： | 2019-07-10 |
A Predicting Model to Estimate the Probability of Malignancy in Solitary Pulmonary Nodules Basing on CT Images |
| |
| Affiliation: | Department of Radiology, Shenzhen People's Hospital, Shenzhen 508001, China |
| |
| Abstract: | Objective: To establish a predicting model using multivariate logistic regression analysis for estimating the probability of malignancy in solitary pulmonary nodules, and to compare our model with Mayo model and Brock model. Methods: From January 2015 to December 2017, 319 patients with SPNs identified by histopathology in Shenzhen peoples' hospital were analyzed retrospectively. Among 319 cases, 229 patients were in modeling group (group A), and 90 patients were in validating group (group B). We analyzed gender, age, diameter, smoking history, spiculation, upper location, unclear border, lobulation, vacuole sign, vessel convergence sign, pleural indentation, ground glass opacity and calcification in patients of group A, selected independent influencing factors by univariate analysis and multivariate logistic regression analysis and established a predicting model. Our model was verified with the date of group B, and was compared with Mayo model and Brock model. Results: The age, diameter, upper location, unclear border, lobulation, vacuole sign, vessel convergence sign, pleural indentation, and ground glass opacity were shown statistically significance between malignant and benign SPNs in univariate analysis (P<0.05). The spiculation, unclear border, lobulation, and ground glass opacity were independent influencing factors in multivariate logistic regression analysis. When group B data was substituted into the established formula, the area under the ROC curve was 0.894, sensitivity was 91.3%, specificity was 77.3%, positive likely ratio was 4.02, negative likely ratio was 0.11, positive predictive value was 80.8%, negative predictive was 89.5%. The difference between our model and Mayo model was statistically significant (P=0.0049). The difference between our model and Brock model was not statistically significant (P=0.79). Conclusion: The spiculation, unclear border, lobulation, and ground glass opacity are independent influencing factors between benign and malignant solitary pulmonary nodules. This logistic regression equation has favorable effective functions for the diagnosis of SPNs. For patients in this study, our model is better than Mayo model, and is same as Brock model. |
| |
| Keywords: | |
| 本文献已被 CNKI 万方数据 等数据库收录！ |
| 点击此处可从《CT理论与应用研究》浏览原始摘要信息 |
|
点击此处可从《CT理论与应用研究》下载全文 |
|
