^(18)F-FDG PET/CT影像组学对肺腺癌KRAS突变的预测价值

Predictive value of ^(18)F-FDG PET/CT imaging for KRAS mutation in lung adenocarcinoma

目的 探讨18F-脱氧葡萄糖(FDG)PET/CT影像组学对肺腺癌鼠类肉瘤病毒癌基因(KRAS)突变的预测价值。方法 回顾性收集2019年1月至2022年5月收治于我院行^(18)F-FDG PET/CT检查的120例肺腺癌患者影像及病理资料,采用横断面分层法将患者分为建模组(70例)和验证组(40例),将建模组患者按照KRAS基因是否突变分为突变组(22例)和野生组(58例),比较两组患者临床资料及影像学特征;并根据临床资料和影像组学特征建立临床(clinic)模型、影像组学(rad)模型和联合(combin e)模型对肺腺癌患者KRAS基因突变进行预测。采用受试者工作特征(ROC)曲线、校准曲线和临床决策曲线对三种模型的预测价值进行评估。结果 建模组患者Clinic模型=1.168+0.976×性别+0.041×年龄+1.298×吸烟史;Rad模型=1.451+1.412×orig_shape_MAL+1.274×wave_LHH_GLSZM_GLNU+0.925×wave_HHH_GLDM_DE+0.753×wave_HLL_FO_Minimum+1.168×wave_HHH_FO_TE;Combine模型=0.943+0.305×性别+0.351×年龄+0.941×吸烟史+0.795×Rad-score。建模组和验证组RO C曲线显示三种模型均具有良好的预测效能,Delong检验发现,建模组和验证组com bine模型ROC曲线下面积均显著高于rad和clinic(均P<0.05)。校准曲线显示各模型在建模组与验证组中均具有良好拟合效果。临床决策曲线显示combine模型在建模组和验证组中较另两种模型具有较高的临床实用价值。结论 ^(18)F-FDG PET/CT影像组学和临床特征联合对肺腺癌患者KRAS突变具有良好的预测价值。

Objective To investigate the predictive value of ^(18)F-deoxyglucose(FDG) PET/CT imaging for murine sarcoma virus oncogene(KRAS) mutation in lung adenocarcinoma.Methods The imaging and pathological data of 120 patients with lung adenocarcinoma who received ^(18)F-FDG PET/CT examination in our hospital from January 2019 to May 2022 were retrospectively collected,The patients were divided into a modeling group(80 cases) and a verification group(40 cases) by cross-sectional stratification,and the patients in the modeling group were divided into a mutant group(22cases) and a wild group(58 cases) according to whether the KRAS gene was mutated.The clinical data and imaging characteristics of the two groups were compared.The clinics model,Rad model and Combine model were esta blished according to clinical data and imaging cha racteristics to predict the mutation of KRAS gene in lung adenocarcinoma patients.Receiver operating characteristic(ROC)curve,calibration curve and clinical decision curve were used to evaluate the predictive value of the three models.Resultss Clinic model of patients in modeling group=1.168+0.976×gender+0.041×age+1.298×smoking history,Rad model=1.451+1.412×orig_shape_MAL+1.274×wave_LHH_GLSZM_GLNU+0.925×wave_HHH_GLDM_DE+0.753×wave_HLL_FO_Minimum+1.168×wave_HHH_FO_TE,Combine model=0.943+0.305×gender+0.351×age+0.941×smoking history+0.795× Radscore.ROC curves of modeling group and validation group showed that the three models had good predictive efficiency.Delong test found that the area under ROC curve of the combine model in the modeling group and the verification group was significantly higher than that in the Rad and Clinical models(all P<0.05).Calibration cu rves show that all models have good fitting effect in modeling group and verification group.The clinical decision curve showed that the combine model had higher clinical practical value in the modeling group and the validation group than the other two models.Conclusion The com bination of ^(18)F-FDG PET/CT imaging and clinical features has a good predictive value for KRAS mutation in lung adenoca rcinoma patients.