乳腺囊实性病变良恶性超声预测模型的建立

目的 分析乳腺囊实性病变中良恶性肿块的超声声像图特点,建立预测模型。方法 回顾性分析71例经手术治疗或穿刺诊断的乳腺囊实性病变患者的超声声像图,与术后病理结果对照分析,采用Logistic回归分析筛选影响因素并建立预测模型。结果 病灶形态(OR=26.109;P=0.027)、血流(OR=23.900;P=0.021)、钙化(OR=11.642;P=0.030)、年龄(OR=1.127;P=0.002)是乳腺囊实性病变恶性的独立预测因素。通过Logistic回归分析建立预测模型Logit(P)=-12.218+3.262×形态+3.173×血流+2.455×钙化+0.12×年龄。该模型预测乳腺囊实性病变恶性的AUC值为0.945(95%CI:0.895~0.995),诊断阈值为-1.4020,约登指数、敏感度、特异度、准确率、阳性预测值、阴性预测值分别为0.804、94.4%、86.0%、88.0%、68.0%、98.0%。结论 以病灶形态、血流、钙化和患者年龄建立的预测模型可以较准确预测乳腺囊实性病变的良恶性,值得在超声诊断中应用。

Objective To analyze and summarize the ultrasound image characteristics of benign and malignant masses in cystic and solid lesions of breast,build predictive models.Methods A retrospective analysis was performed for the ultrasound images of 71 patients with cystic and solid breast lesions which diagnosed by surgery or puncture.The influencing factors were analyzed by Logistic regression analysis and a prognostic model was created.Results Morphology(OR=26.109,P=0.027),blood flow(OR=23.900,P=0.021),calcification(OR=11.642,P=0.030),age(OR=1.127,P=0.002)were independent predictor of malignancy of breast cystic lesions.The regression model was Logit(P)=-12.218+3.262×morphology+3.173×blood flow+2.455×calcification+0.12×age.The areas under the ROC curve(AUC)of the predict malignancy was 0.945(95%CI:0.895~0.995).The Youden index,sensitiv ity,specificity,accuracy,positive predictive value(PPV)and negative predictive value(NPV)with the cutoff value of-1.4020 were 0.804、94.4%、86.0%、88.0%、68.0%and 98.0%respectively.Conclusion The model based on the morphology、blood flow、calcification and age can effectively predict the benign and malignant of breast cystic and solid lesions.It is worth applying in ultrasound diagnostics.