Logistic回归及ROC曲线综合评价超声造影对乳腺癌的诊断价值  

Comprehensive evaluation of the diagnostic value of contrast-enhanced sonography for breast cancer by ROC curve and logistic regression

作　　者：沈嫱[1] 宋光辉[1] 张建兴[1] 林勇[1] 党欢[1] 

机构地区：[1]广州中医药大学附属第二医院超声科,510120

出　　处：《实用医学杂志》2009年第7期1058-1060,共3页The Journal of Practical Medicine

基　　金：广东省医学科研基金资助项目(编号:A2006246)

摘　　要：目的:应用Logistic回归模型及受试者工作特征曲线(ROC曲线)探讨超声造影综合指标在乳腺癌诊断中的应用价值。方法:92例乳腺肿块患者术前行超声造影检查,通过Logistic逐步回归分析,对超声造影诊断乳腺癌的多项指标进行筛选,建立乳腺癌诊断的统计模型即回归方程,并应用ROC曲线评价回归方程的诊断效能和最佳诊断分界值。结果:超声造影时间-强度曲线上升支斜率和乳腺肿块内血流形态联合诊断乳腺癌的ROC曲线下面积(AUC)高于上升支斜率及血流形态单一指标的AUC(P<0.05);联合上升支斜率、血流形态的回归方程P=1/[1+e-(-3.637+0.856X+3.153A1+3.572A2)]对乳腺癌的临床诊断界值为0.659,其相应的灵敏度、特异性、正确性分别为95.8%、84.2%、91.3%。结论:联合上升支斜率、血流形态的回归方程有助于提高超声造影对乳腺癌的诊断效能,优于单一指标的检测。Objective To explore the diagnostic value of contrast-enhanced sonography for breast cancer by receiver operating characteristic curve （ROC curve） and a model of logistic regression. Methods Contrast-enhanced sonography was performed preoperatively in 92 women with breast mass. After analyzing the sonographic findings with logistic stepwise regression, we screened multiple diagnostic parameters for breast cancer and established a mathematical model for diagnosis. Then we assessed the diagnostic efficacy of the model and calculated the diagnostic cut-off points for breast cancer using the ROC curve. Results The area under ROC curve （AUC） of the rising slope combined with the morphological characteristics of blood flow was greater than that of either parameter alone （P 〈 0.05）. According to the regression equation P =1/[1＋e^-（-3.637＋0.856X＋3.153A/＋3.572A2） , the cut-off point, sensitivity, specificity, and accuracy of the combined parameters for diagnosing breast cancer were 0.659, 95.8%, 84.2%, and 91.3%, respectively. Conclusion The model of logistic regression is helpful to improve the diagnostic efficacy of contrast-enhanced sonography for breast cancer.

关 键 词：乳腺肿瘤 LOGISTIC模型 ROC曲线 超声检查 造影剂 

分 类 号：R737.9[医药卫生—肿瘤]