乳腺结节样病变超声诊断的Logistic回归分析

Ultrasonic Diagnosis of Breast Nodular Lesions by Logistic Regression

目的建立乳腺结节样病变超声诊断的Logistic回归模型.方法对经手术病理证实的205个乳腺病变的二维超声、彩色多普勒超声声像特征进行回归分析,建立Logistic回归模型,用ROC曲线法评价Logistic回归模型的预报能力.结果9个超声特征进入Logistic模型初步筛选,即病灶后方回声改变、病灶活动度、病灶内血流信号、毛刺征、病灶内微小钙化、强回声晕征、包膜、腋窝淋巴节结构改变、纵横径比.经筛选后,具有显著性的病灶后方回声改变、病灶活动度、病灶内血流信号3因素再进一步Logistic回归分析,改善拟合优度. Logistic回归模型ROC曲线下面积为0.981.结论超声声像特征的Logistic 回归模型有助于乳腺良、恶性病变的鉴别诊断.

Objective To establish a Logistic regression model based on ultrasonographic characteristics and to diagnose breast nodular lesions.Methods The characteristics of gray-scale ultrasonography （ US）,color Doppler flow imaging （ CDFI） and some clinical symptoms were evaluated in 205 breast nodular lesions confirmed by surgical pathology on a retrospective study .A Logistic model for predic-ting malignancy of the breast nodular lesions on the basis of ultrasonographic characteristics and clinical symptoms were obtained .A receiver operating characteristic（ROC） curve was used to assess the performance of the Logistic model .Results Nine ultrasonographic characteristics entered the Logistic model.They were rear echo change,mass movement,color Doppler flow grade within lesion ,spicule sign,strong echo halo sign,micro-calcification,envelope,aspect ratio,and axillary lymph nodes structural change respectively .After screening,rear echo change, mass movement and color Doppler flow grade within lesion were done again to improve the goodness of fit .The area under the ROC curve was 0.981.Conclusion The Logistic regression model can help differentiate malignant breast lesion from benign one .