基于常规超声及超声造影预测浅表淋巴结恶性的危险因素及其列线图模型的应用价值  

Prediction of Risk Factors of Malignant Superficial Lymph Nodes Based on Conventional Ultrasound and Contrast-Enhanced Ultrasound and Application Value of Nomogram Prediction Model

作　　者：张剑[1] 陈卉[1] 徐斌[2] 张津楠 卢叶君[1] 卜军[1] Zhang Jian;Chen Hui;Xu Bin;Zhang Jinnan;Lu Yejun;Bo Jun(Department of Ultrasound,The Third Affiliated Hospital of Soochow University,Changzhou,Jiangsu 213000,China;Department of Tumor Biotherapy,The Third Affiliated Hospital of Soochow University,Changzhou,Jiangsu 213000,China)

机构地区：[1]苏州大学附属第三医院超声科,江苏省常州市213000 [2]苏州大学附属第三医院肿瘤生物诊疗中心,江苏省常州市213000

出　　处：《中国超声医学杂志》2023年第11期1268-1272,共5页Chinese Journal of Ultrasound in Medicine

基　　金：常州市卫生计生委青年人才科技项目(No.QN201705)。

摘　　要：目的本研究基于常规超声及超声造影(CEUS)参数构建模型预测浅表淋巴结良恶性,并采用列线图对预测模型进行应用。方法采用病例对照研究方法,收集病理证实的65例患者共96个浅表淋巴结的常规超声、CEUS特征。将96个淋巴结的超声特征通过随机抽样分为训练集及验证集,差异分析筛选出训练集中在良恶性淋巴结中有显著差异的变量,且变量的受试者工作特征(ROC)曲线下面积(AUC)需≥0.7,通过R软件建立Logistic回归模型,并依据模型绘制列线图。利用训练集和验证集数据对模型进行验证。结果训练集中淋巴结门是否正常、有无淋巴结中央和边缘血流、造影剂进入方式、峰值强度(PI)、下降斜率(DS)、时间-强度曲线的曲线下面积(AUCT)在良恶性组间的差异有统计学意义(P<0.05),边缘血流、PI、AUCT的AUC依次为0.72、0.72、0.70;验证集中边缘血流、PI、AUCT的AUC依次为0.70、0.85、0.75,将上述参数拟合Logistic回归模型,模型对训练集和验证集中淋巴结良恶性预测的灵敏度依次为95.74%和82.35%,特异度为60.00%和71.43%,准确度为83.33%和79.17%。结论边缘血流、PI、AUCT是判断浅表增大淋巴结良恶性的重要参数,以此构建的预测模型对预测浅表淋巴结恶性程度具有较好的临床应用价值。Objective A model was constructed based on conventional ultrasonography and contrast-enhanced ultrasound(CEUS)parameters to predict benign and malignant superficial lymph nodes,and the prediction model was constructed with the nomogram.Methods Conventional ultrasonography and CEUS features of 96 superficial lymph nodes in 65 pathologically proven patients were collected by case-control study.The ultrasonic features of 96 lymph nodes were randomly divided into training set and validation set,and the variables with significant differences between benign and malignant lymph nodes in the training group were screened by analysis of variance,The receiver operating characteristic curve(ROC)curves were drawn,and the area under the ROC curve of the variables should be≥0.7.The logistic regression model was established by R software,and the nomograms were drawn based on the model.The model was validated by training and validation set data.Results In the training set,there were statistically significant differences in lymphatic hilum(normal or not),presence(or not)of central and marginal blood flow in the lymph nodes,contrast filling mode,peak intensity(PI),descending slope(DS),and area under the curve of the time-intensity curve(AUCr)between benign and malignant groups(P<0.05).The AUC of marginal blood flow,PI,and AUCr were 0.72,0.72,and 0.70,respectively.In the validation set,the area under the curve(AUC)of marginal blood flow,PI,and AUCr were 0.70,0.85,and 0.75,respectively.The logistic regression model was fit to the above indicators,and the sensitivity of the model for the prediction of lymph node benignity and malignancy in the training and validation sets was 95.74%and 82.35%,the specificity was 60.00%and 71.43%,and the accuracy was 83.33%and 79.17%.Conclusions Visible marginal blood flow,PI and AUCr are important parameters for judging the benign and malignant superficial lymph nodes.The nomogram prediction model could predict the malignant potential of superficial lymph nodes,which is helpful for providing further

关 键 词：浅表淋巴结 恶性 超声造影 预测模型 列线图 

分 类 号：R445.1[医药卫生—影像医学与核医学] R733.1[医药卫生—诊断学]