摘要
目的应用二分类Logistic回归模型分析乳腺肿块良恶性的超声鉴别诊断。方法选择经手术病理证实的151个乳腺病灶的二维灰阶超声、二维彩色多普勒超声、三维灰阶超声、三维彩色多普勒超声、超声弹性成像的各诊断指标进行多因素回归分析，建立Logistic模型。用ROC曲线法评价Logistic模型的预报能力。结果经前进法逐步回归的多变量二分类Logistic回归分析，筛选引入方程的超声检查指标包括：弹性成像评分、形态、内部回声、阻力指数、后方回声和汇聚征。Logistic模型的预报正确率为97．35％，ROC曲线下面积为0.996。结论二分类Logistic回归多元分析模型能很好地描述和分析良恶性乳腺肿块的超声鉴别诊断的过程，并能筛选出有意义的鉴别诊断指标。
Objective To evaluate the application of the binary Logistic regression model to analyze uhrasonographic indexes of the solid breast tumors. Methods The indexes of two dimensional gray scale uhrasonography,two dimensional color Doppler flow imaging, three dimensional gray scale ultrasonography, three dimensional color Doppler flow imaging and ultrasonic elastography were evaluated in 151 breast lesions confirmed by surgical pathology. A Logistic regression model for predicting breast malignancy on the basis of uhrasonographic indexes was obtained. A receiver operating characteristic（ROC） curve was used to assess the performance of the Logistic regression model. Results Six ultrasonic indexes were finally entering the Logistic regression model. They were elasticity score, shape, internal echo, RI, enhancement of posterior acoustic alteration and the converging pattern in the coronal plane. The area under the ROC curve was 0. 996. The percentage correct of prediction was 97.35 %. Conclusions The multivariate analysis model of binary Logistic regression can describe and analyze the process of differential diagnosis of malignant and benign solid breast tumors by ultrasonography and can select out the valuable indexes of differential diagnosis.
出处
《中华超声影像学杂志》
CSCD
2008年第7期601-603,共3页
Chinese Journal of Ultrasonography
