PET／C T 数学预测模型对孤立性肺结节诊断价值的 ROC曲线分析

摘要

目的：利用ROC曲线分析评价 PET／CT 数学预测模型对孤立性肺结节的诊断价值。方法回顾性分析2011年9月～2013年6月我院因诊断为SPN而行PET／CT检查的186例患者。以病灶良恶性结果为应变量，以患者的性别、年龄及病灶的大小、部位、CT值、边界、磨玻璃成分、分叶、血管集束、胸膜牵拉、毛刺、钙化、空泡、空洞、SUVmax作为自变量行单因素和多因素分析，建立诊断SPN良、恶性的回归数学模型。计算PET／CT数学预测模型诊断SPN的灵敏度、特异度、准确性、阳性预测值、阴性预测值，绘制相应ROC曲线并计算曲线下面积大小。结果经二元logistic回归分析建立PET／CT诊断SPN良、恶性的数学模型如下：p＝ex／（1＋ ex ），x＝‐8．111＋0．091×年龄＋1．351×分叶＋3．565×血管集束＋2．153×胸膜牵拉＋0．447× SUVmax。以数学模型对SPN良恶性进行预测，其诊断SPN的灵敏度、特异度、准确性、阳性预测值及阴性预测值分别为87．8％、81．0％、85．5％、90．0％及77．3％。 PET／CT 数学预测模型诊断SPN的ROC曲线下面积为0．927±0．019。结论以logistic回归分析构建的PET／CT数学预测模型诊断SPN的准确性较高，且不受人为因素干扰，因此其临床应用的可行性较高，是一种值得推荐的方法。%Objective To evaluate the value of PET/CT mathematical prediction model in diagnosis of solitary pulmonary nodule (SPN ) ,using receiver operating characteristic (ROC ) curves analysis .Methods From September 2011 to June 2013 ,186 patients who were confirmed with SPN and performed with PET /CT were retrospectively analysed in this stud‐y .The logistic mathematical prediction model was established by univariate analysis and multivariate analysis ,using malig‐nant or benign result as dependent variable and gender ,age ,nodule size ,density ,location ,boundary ,ground glass opaci‐ty ,lobulation ,vascular convergence ,retraction of pleural ,speculation ,calcification ,vacuole ,cavity as independent varia‐ble .The sensitivity ,specificity ,accuracy ,positive predictive value and negative predictive value of the mathematical mod‐el in diagnosis of SPN were calculated .The ROC curve was depicted and the area under the ROC curve (AUC) of the mathematical model was calculated .Results The mathematical prediction model established by logistic regression analysis was p=ex/(1+ ex ) ,x=‐8 .111+0 .091 × age+1 .351 × lobulation+3 .565 × vascular convergence+2 .153 × retraction of pleural+0 .447 × SUVmax .The sensitivity ,specificity ,accuracy ,positive predictive value and negative predictive value of the model for prediction of SPN were 87 .8% ,81 .0% ,85 .5% ,90 .0% and 77 .3% respectively .The AUC of mathemati‐cal model was 0 .927 ± 0 .019 .Conclusion The value of PET/CT mathematical prediction model in diagnosis of SPN is high and not affected by subjective factors .Therefore ,it is a recommended diagnostic method in clinical practice .