基于^(18)F-FDG PET/CT建立孤立性肺空洞的数学模型

Establishment of a mathematical model for solitary pulmonary cavity based on ^(18)F-FDG PET/CT

目的利用Logistic回归分析建立^(18)F-FDG PET/CT的数学模型用以诊断孤立性肺空洞的良恶性。方法分析2020年12月至2023年2月在吉林大学中日联谊医院PET/CT中心行^(18)F-FDG PET/CT检查的孤立性肺空洞病例共92例,其中男61例,女31例,对病例的一般资料及图像特征进行单因素分析及多因素二元Logistic回归分析,筛选出与孤立性肺空洞良恶性相关的独立危险因素,建立数学模型,并通过比较受试者工作曲线下面积(AUC)比较该模型、CT、PET及PET/CT对孤立性肺空洞的诊断效能。结果92例患者中良性32例(34.78%)、恶性60例(65.22%),单因素分析结果提示以下因素在良恶性组间差异有统计学意义(P<0.05):病灶SUV_(max)值、病灶SUV_(max)/肝脏SUV_(mean)、病灶SUV_(max)/脾脏SUV_(mean)、病灶最大径、空洞壁不均匀、空洞壁分叶、胸膜牵拉、空洞内壁结节及偏心性空洞。通过Logistic回归分析构建数学模型为:y=exp(x)/[1+exp(x)],x=-2.022+0.419×(病灶SUV_(max)/脾脏SUV_(mean))-(0.462×病灶最大径)+2.023×空洞壁分叶+1.341×偏心性空洞;选取y=0.51为截断值,灵敏性、特异性、准确性、阳性预测值及阴性预测值分别为91.70%、75.00%、83.35%、78.58%及90.04%;该模型的AUC为0.875[90%可信区间(CI):0.795~0.995],显著大于CT的AUC(Z=2.424,P<0.05)及病灶SUV_(max)/脾脏SUV_(mean)的AUC(Z=3.276,P<0.05),与^(18)F-FDG PET/CT的AUC比较,差异无统计学意义(Z=-0.249,P>0.05)。结论利用Logistic回归分析建立的PET/CT数学模型对孤立性肺空洞良恶性鉴别具有较好诊断效能及较高临床价值。

Objective To establish a mathematical model of ^(18)F-FDG PET/CT for the diagnosis of benign and malignant solitary pulmonary cavity by Logistic regression analysis.Methods We analyzed 92 patients with solitary pulmonary cavity who underwent ^(18)F-FDG PET/CT examination in the PET/CT Center of China-Japan Union Hospital of Jilin University from December 2020 to February 2023,including 61 males and 31 females.The general data and image features of these patients were analyzed by univariate analysis and multivariate Logistic regression analysis.The independent predictors of malignancy were estimated,then the mathematical model was established.The diagnostic capabilities of the model,CT,PET and PET/CT for solitary pulmonary cavity were compared by comparing the area under the receiver operating characteristic curve(AUC)of them.Results 34.78%of the cavities were benign,and 65.22%of the cavities were malignant in 92 patients.Univariate analysis showed that the following factors were statistically significant between the benign and malignant groups:SUV_(max),SUV_(max)/SUV_(mean) liver,SUV_(max)/SUV_(mean) spleen,endocaval tubercle and eccentric cavity.The mathematical model established by Logistic regression analysis was y=exp(x)/[1+exp(x)],x=-2.022+0.419×(SUV_(max)/SUV_(mean) spleen)-[0.462×(maximum diameter of lesion)+2.023×lobulation+1.341×eccentric cavity].When the cut-off value was set at y=0.51,the sensitivity,specificity,accuracy,positive predictive value and negative predictive value of our model were 91.70%,75.00%,83.35%,78.58%and 90.04%,respectively.The AUC of the model was 0.875[95%confidence interval(CI):0.795 to 0.955],which was greater than that of CT(Z=2.424,P<0.05)and SUV_(max)/SUV_(mean) spleen(Z=3.276,P<0.05),and had no significant difference from that of PET/CT(Z=-0.249,P>0.05).Conclusion The mathematical model of PET/CT established by Logistic regression analysis has good diagnostic performance in the diagnosis of benign and malignant solitary pulmonary cavity and high clinical value.