摘要
目的采用Logistic回归模型探讨剪切波弹性成像(SWE)评估颈部淋巴结良恶性的价值和筛选SWE的定量参数。方法对59例患者共95个疑似恶性颈部淋巴结在颈部淋巴结清扫前进行常规超声检查及剪切波弹性成像检查,分别比较颈部良性与恶性淋巴结的弹性值比值(E-ratio)、病灶的平均弹性值(E-mean)、病灶的最大弹性值(E-max)和标准差(SD);以病理结果为金标准,建立二分类逻辑回归模型,绘制模型预测概率的ROC曲线并计算曲线下面积,确定诊断界值,计算敏感度、特异度、准确度、阳性预测值、阴性预测值和约登指数。结果Logistic回归模型为logitic(p)=-3.653+1.760 X_1+0.235 X_2-0.207 X_3+0.168 X_4,X_1为E-ratio,X_2为E-mean,X_3为E-max,X_4为SD。模型预测概率ROC曲线的曲线下面积为0.865,以55.66%为模型预测概率的诊断界值时,准确率最高为84.21%,对应的敏感度为80.00%,特异度88.89%,阳性预测值为88.89%,阴性预测值为80.00%,约登指数为68.89%。结论运用SWE的4个定量参数建立的逻辑回归模型对颈部良恶性淋巴结的鉴别具有中等诊断价值,4个定量参数均可为良恶性淋巴结的鉴别诊断提供依据,而诊断价值最高的定量参数是E-ratio。
Objective To explore the value of shear wave elastography（SWE）in the assessment of cervical lymph nodes,and to screen the quantitative parameters of SWE using model of Logisticregression.Methods Totally 59 patients with 95 suspected malignant cervical lymph nodes underwent conventional ultrasound and SWE before cervical lymphadenectomy.Quantitative parameters（E-ratio,E-mean,E-max,SD）between benign and malignant lymph nodes were compared.According to the pathologic findings,the model of binary Logisticregression was established.ROC curves of model predictive probability were drawn and the area of under curve（AUC）was calculated.The cut-off value was determined,and the sensitivity,specificity,accuracy,positive predictive value,negative predictive value and Youdenindex were calculated.Results The Logistic regression model was logitic（p）=-3.653＋1.760 X_1＋0.235 X_2-0.207 X_3＋0.168 X_4,X_1,X_2,X_3,X_4 were E-ratio,E-mean,E-max,SD,respectively.The AUC of model predictive probability was 0.865.When cut-off value was 55.66% of model predictive probability,the highest accuracy was 84.21%,sensitivity was 80.00%,specificity was88.89%,positive predictive value was 88.89%,negative predictive value was 80.00%,and Youdenindex was 68.89%.Conclusion Using four quantitative parameters of SWE,the Logisticregression model has moderate value in the differentiation of benign and malignant cervical lymph nodes,four quantitative parameters can provide diagnostic evidences,and the most valuable one is E-ratio.
出处
《中国医学影像技术》
CSCD
北大核心
2016年第4期500-503,共4页
Chinese Journal of Medical Imaging Technology
基金
国家自然科学基金面上项目(30872996)
