超声多因素Logistic回归分析鉴别甲状腺结节的良恶性

Multivariate Logistic regression analysis of ultrasound features in differential diagnosis benign and malignant thyroid nodule

目的探讨Logistic回归分析在超声多因素鉴别甲状腺良恶性结节中的价值。方法选取经超声引导下穿刺活检或手术后病理证实的甲状腺恶性结节40例(47个结节)、良性结节68例(83个结节),分析结节的形态、边界、边缘、内部回声、血流分布以及超声造影增强模式、剪切波速度值等超声特征并对其进行单因素分析,将有统计学意义指标作多因素Logistic回归分析。结果单因素分析显示低回声、边界模糊、不规则边缘、垂直生长、微钙化、内部血流、不均性增强及SWV鉴别甲状腺良恶性结节差异有统计学意义(P均<0.001)。多因素分析显示低回声、微钙化、高SWV值及不均匀增强模式与甲状腺癌相关(P均<0.05)。多因素分析诊断甲状腺癌的准确率、敏感度、特异度、阳性预测值和阴性预测值分别为93.08%(121/130)、91.49%(43/47)、93.97%(78/83)、89.58%(43/48)和95.12%(78/82)。结论多因素分析可有效提高甲状腺癌诊断的准确率。低回声、微钙化、不均性增强模式以及高SWV值可作为综合判断甲状腺结节良恶性的可靠指标。

Objective To investigate the significance of Logistic regression analysis in differential diagnosis of benign and malignant thyroid nodules by ultrasound multiple factors. Methods Sixty-eight patients with 83 thyroid benign nodule and 40 patients with 47 thyroid malignant nodule confirmed by pathology were enrolled. Ultrasound features of morphology, boundary, margin, internal echogenicity, blood distribution, enhanced pattern and shear wave velocity （SWV） were ana- lyzed by univariate analysis, and the multivariate Logistic regression analysis was performed on statistically significant difference indexes. Results On univariate analysis, hypoechoic, blurred boundary, spiculated/microlobulate margin, mi- crocalcificaition, vertical growth, internal blood, heterogeneous enhanced pattern and high SWV had statistically significant difference between benign and malignant thyroid nodules （all P〈0. 001）. On multivariate analysis, hypoechoic, microcal- cificaition, high SWV and heterogeneous enhanced pattern correlated with thyroid malignant nodule （all P〈0. 05）. Based on multiple factors analysis, the accuracy, sensitivity, specificity, positive predictive value and negative predictive value of ultrasound in diagnosis of thyroid carcinoma were 93.08% （121/130）, 91.49% （43/47）, 93.97% （78/83）, 89, 58% （43/ 48） and 95.12%（78/82）, respectively. Conclusion Multiple factors analysis can efficiently improve the accuracy of thy- roid malignant nodule from benign nodule. Hpoechoic, microcalcificaition, high SWV and heterogeneous enhanced pattern would be the reliable indexes for differential diagnosis of thyroid carcinoma.