【摘要】 目的基于2017版ACR TI-RADS,通过二元Logistic回归分析,预测最大径≥1 cm甲状腺结节的恶性风险。方法回顾性分析1 226例患者(1 334个结节)的超声图像。以粗针穿刺活检病理为金标准,对纳入病例依据2017版ACR TI-RADS对结节的结构、回声、形态、边缘、强回声灶等超声特征的评分进行Logistic回归分析。根据回归模型绘制ROC曲线,计算预测概率判定甲状腺结节为恶性的诊断界值。结果 Logistic回归显示回声、形态、边缘和强回声灶等四种特征的评分为判定甲状腺恶性结节的独立影响因素。ROC曲线下面积为0.744。预测概率判定甲状腺结节为恶性的诊断界值为56.4%。结论基于2017版ACR TI-RADS建立的Logistic回归模型能够很好的预测最大径≥1 cm甲状腺结节的恶性风险。更多还原

【Abstract】 Objective To predict the malignant risk of thyroid nodules with maximum diameter ≥1 cm by Logistic regression model established on the basis of the 2017 ACR TI-RADS. Methods The ultrasonographic images of 1226 patients(1334 nodules) were reviewed. Logistic regression was performed on the scores of composition, echogenicity, shape, margin and echogenic foci by using core needle biopsy pathology as reference. The ROC curve was depicted according to the Logistic regression model and the cut-off value for predictive probability to confirm malignancies was calculated. Results The scores of echo, shape, margin and echogenic foci were defined as independent risk factors for predicting malignancies by Logistic regression. Area under the ROC curve was 0.744. The cut-off value for predictive probability to confirm malignancies was 56.4%. Conclusions The Logistic regression model established on the basis of the 2017 ACR TI-RADS can be used to predict the malignant risk of thyroid nodules with maximum diameter≥1 cm.更多还原