灰阶超声诊断甲状腺微小乳头状癌的Logistic回归分析

Logistic regression analysis of papillary thyroid microcarcinoma with gray-scale ultrasound

目的探讨甲状腺微小乳头状癌的灰阶超声恶性征象,评价其应用价值。方法回顾性分析经过手术病理证实的378个甲状腺微小结节(其中良性137个,恶性241个)的灰阶超声表现,通过Logistc回归分析筛选微小结节的恶性征象,并建立Logistic回归模型。结果 Logistic回归分析显示最后进入回归模型的是微小结节的纵横比、内部结构、形态、边界、与甲状腺被膜关系、声晕及钙化类型。Logistic模型以预测概率P=0.50作为阈值,准确率达84.40%,ROC曲线下面积为0.91。结论以甲状腺微小结节的纵横比、内部结构、形态、边界、与甲状腺被膜关系、声晕及钙化类型7个诊断参数变量建立的Logistic回归模型有助于甲状腺微小乳头状癌的鉴别诊断。

Objective To approach malignant gray-scale sonographic features of papillary thyroid microcarcinoma（PTMC）and to assess the diagnostic value of sonographic findings of subcentimeter thyroid nodules.Methods The sonographic characteristics of 378 thyroid nodules（≤1cm）were reviewed retrospectively.ALogistic model was obtained for predicting PTMC.Results Anteroposterior and transverse diameter ratio,internal struction,shape,border,relation with thyroid capsula,acoustic halo and calcification were finally entering the Logistic model.If apply P=0.50 as cut-off,the diagnostic accuracy was 84.40%,the area under curve of ROC was 0.91.Conclusion The Logistic model which is based on anteroposterior and transverse diameter ratio,internal struction,shape,border,relation with thyroid capsula,acoustic halo and calcification can efficiently differentiate malignant subcentimeter thyroid nodules from benign ones.