基于超声乳腺影像报告和数据系统与Logistic回归分析的良恶性乳腺肿块鉴别诊断模型的构建

Construction of differential diagnosis model of benign and malignant breast masses based on ultrasonic breast image reporting and data system with Logistic regression analysis

目的基于超声乳腺影像报告和数据系统(BI-RADS)与Logistic回归分析构建良恶性乳腺肿块鉴别诊断模型,并分析其应用价值。方法选取156例和67例乳腺肿块患者,分别作为模型组和验证组。依据病理结果将156例模型组患者分为良性组(n=87)和恶性组(n=69),记录两组患者的临床特征和BI-RADS超声影像特征,乳腺恶性肿块的影响因素采用Logistic回归分析。根据影响因素构建良恶性乳腺肿块鉴别诊断模型,绘制受试者工作特征(ROC)曲线,计算曲线下面积(AUC),分析该模型的诊断效能。结果单因素分析结果显示,良性组和恶性组患者年龄、肿块最大直径、淋巴结肿大情况、BI-RADS分类、超声弹性评分、方位、形态、边缘、内部回声、后方回声、血供情况、结构扭曲情况和微钙化情况比较,差异均有统计学意义(P﹤0.01)。多因素Logistic回归分析结果显示,年龄≥40岁、肿块最大直径≥3 cm、方位不平行、形态不规则、边缘成角、边缘毛刺、内部回声不均匀和微钙化均是乳腺恶性肿块的独立危险因素(P﹤0.05)。根据上述8个影响因素构建良恶性乳腺肿块的鉴别诊断模型,Hosmer-Lemeshow拟合优度检验结果显示,χ^(2)=12.512,P=0.130;该模型诊断乳腺恶性肿块的AUC为0.896(95%CI:0.844~0.948),灵敏度为85.50%,特异度为85.10%,约登指数为0.706,表明该模型拟合优度良好,具有较好的诊断效能。将验证组67例患者代入诊断模型中,该模型诊断乳腺恶性肿块的AUC为0.986(95%CI:0.962~1.000),灵敏度为93.30%,特异度为100%,约登指数为0.933,说明该模型的诊断效能较好。结论基于超声BI-RADS与Logistic回归分析构建的良恶性乳腺肿块鉴别诊断模型具有较高的临床应用价值,可作为辅助手段对乳腺癌进行筛查。

Objective To construct a differential diagnosis model of benign and malignant breast masses based on breast image reporting and data system(BI-RADS)with Logistic regression analysis and analyze its application value.Method A total of 156 and 67 patients with breast masses were selected as model group and validation group,respectively.According to the pathological results,156 patients in model group were divided into benign group(n=87)and malignant group(n=69),the clinical features and BI-RADS ultrasound image features of the two groups were recorded,and the influencing factors of malignant breast masses were analyzed by Logistic regression.The differential diagnosis model of benign and malignant breast masses was constructed according to the influencing factors,the receiver operating characteristic(ROC)curve was plotted and the area under the curve(AUC)was calculated to analyze the diagnostic efficiency of the model.Result Univariate analysis showed that there were statistically significant differences in age,maximum mass diameter,lymph node enlargement,BI-RADS classification,ultrasonic elasticity score,orientation,morphology,edge,internal echo,posterior echo,blood supply,structural distortion and micro-calcification between benign group and malignant group(P<0.01).Multivariate Logistic regression analysis showed that age≥40 years,maximum diameter≥3 cm,nonparallel orientation,irregular shape,marginal angulation,marginal burrs,uneven internal echo and micro-calcification were independent risk factors for malignant breast masses(P<0.05).According to the above eight factors,the differential diagnosis model of benign and malignant breast masses was constructed.The Hosmer-Lemeshow goodness of fit test showed thatχ^(2)=12.512,P=0.130.The AUC of this model for the diagnosis of malignant breast tumors was 0.896(95%CI:0.844-0.948),the sensitivity was 85.50%,the specificity was 85.10%,and the Youden index was 0.706,indicating that the model had good goodness of fit and diagnostic efficacy.The 67 patients in validation group was included in the diagnostic model,the AUC of this model for the diagnosis of malignant breast masses was 0.986(95%CI:0.962-1.000),the sensitivity was 93.30%,the specificity was 100%,and the Youden index was 0.933,which showed that the diagnostic efficacy of this model was good.Conclusion The differential diagnosis model of benign and malignant breast masses based on ultrasonic BI-RADS with Logistic regression analysis has high clinical application value,and can be used as an auxiliary means for breast cancer screening.