| 杨帆,朱吉发,吴火林,陈媛,陈贤翔,俞林芳,黄晓娟.多项超声特征联合诊断周围神经鞘瘤的Logistic回归分析[J].中国医学影像技术,2016,32(4):605~608 |
| 多项超声特征联合诊断周围神经鞘瘤的Logistic回归分析 |
| Logistic regression analysis of ultrasonic features in diagnosis of peripheral neurilemmomas |
| 投稿时间：2015-07-28 修订日期：2015-12-11 |
| DOI：10.13929/j.1003-3289.2016.04.032 |
| 中文关键词: 神经鞘瘤 超声检查 Logistic模型 |
| 英文关键词:Neurilemmoma Ultrasonography Logistic regression model |
| 基金项目:南京军区医学科技创新经费资助项目(15MS098)、南京军区联勤第十八分部医学科技青年培育项目(18FBQN2015001). |
|
| 摘要点击次数: 2712 |
| 全文下载次数: 703 |
| 中文摘要: |
| 目的 探讨多项超声特征联合诊断周围神经鞘瘤的价值,并建立以超声特征为自变量的二分类Logistic回归模型.方法 回顾性分析浅表软组织肿物患者179例共181个病灶的超声表现,并依据术后病理分为神经鞘瘤组43例(45个病灶)和非神经鞘瘤组136例(136个病灶),两组间超声特征的比较采用χ2检验.以病理诊断为金标准,构建以超声特征为自变量的二分类Logistic回归模型,绘制ROC曲线,评价Logistic回归模型的预报能力.结果 神经鞘瘤多表现为形态规则、边界清晰、内部血流信号丰富,可见鼠尾征、靶征及血管伴行征,其显示率高于非神经鞘瘤组,差异有统计学意义(P<0.05);两组囊性变、后方回声增强差异无统计学意义(P>0.05).Logistic回归分析筛选出形态、鼠尾征、靶征及内部血流4个对诊断神经鞘瘤有统计学意义的特征变量,靶征的优势比高于其他超声特征,Logistic回归模型预报神经鞘瘤的正确率为89.5%,敏感度为73.3%,特异度为94.9%,ROC曲线下面积为0.875±0.036(P<0.000 1).结论 联合多项超声特征所构建的Logistic回归模型对诊断周围神经鞘瘤有重要价值. |
| 英文摘要: |
| Objective To explore the value of ultrasonic features in the diagnosis of peripheral neurilemmomas, and to contribute the binary Logistic regression model of ultrasonic features as independent variable. Methods Ultrasonic appearances of 179 patients with 181 superficial soft tissue masses were analyzed retrospectively. All patients were divided into neurilemmoma group (43 patients with 45 lesions) and non-neurilemmoma group (136 patients with 136 lesions). χ2 test was implemented for comparison the ultrasonic features of two groups. Pathologic diagnosis was used as golden standard, a binary Logistic regression model on the basis of ultrasonic features was obtained. ROC curve was used to assess the performance of the Logistic regression model. Results The ultrasonic appearances of neurilemmomas showed regular shapes, clear boundary, hypervascular, rat tail sign, target sign and vessels accompanying sign, which were higher than those of non-neurilemmomas, and the differences were statistically significant (all P<0.05), whereas the differences in the cystic degeneration and posterior echo-enhancement were not statistically significant (P>0.05). Four ultrasonic features were finally entering the Logistic regression model which were shape, rat tail sign, target sign and internal blood flow, and the odds ratio of target sign was higher than other features. The diagnostic accuracy, sensitivity and specificity of Logistic regression model for neurilemmomas were 89.5%, 73.3%, 94.9%, respectively, the area under the ROC curve was 0.875±0.036 (P<0.000 1). Conclusion Logistic regression model on the basis of ultrasonic features has important value in the diagnosis of peripheral neurilemmomas. |
| 查看全文 查看/发表评论 下载PDF阅读器 |
|
|
|
