影像组学结合临床对非小细胞肺癌基因状态的预测

Predictive Power of Radiomic Features Combined with Clinical features for Gene Mutations in Non-small Cell Lung Cancer Patients

目的探讨影像组学特征在非小细胞肺癌EGFR基因突变中的预测价值。方法回顾性分析127例病理证实为非小细胞肺癌患者的PET/CT及临床实验室资料,对CT和PET图像上的肿瘤病灶进行手动分割并提取影像组学特征,以EGFR突变状态进行分组,采用t检验、Wilcoxon rank-sum检验及卡方分析分析组间影像组学特征及临床特征的差异。采用Logistic回归分析建模,比较影像组学模型、临床模型、联合模型对基因状态的预测价值。结果 EGFR突变与年龄、大小、最大标准摄取值(SUVmax)、癌胚抗原(CEA)无显著相关(P>0.05),与吸烟史、性别、糖类抗原125(CA125)显著相关(P<0.05)。CT影像组学模型可以区分EGFR状态[曲线下面积(AUC)=0.775],结合该模型与临床特征(AUC=0.786)可以提高预测的准确性(AUC=0.867),但不具有显著性(P>0.05)。PET影像组学模型可以很好区分EGFR状态(AUC=0.819),结合临床模型(AUC=0.786),显著地提高了性能(AUC=0.927)。结论基因突变可以驱动不同的影像学表现,基于影像组学特征和临床特征的高级生物标志物可能用于预测EGFR突变状态。

Objective This study investigated the association and predictive power of18 F-FDG PET-based radiomic features for EGFR mutations in non-small cell lung cancer(NSCLC)patients.Methods The PET/CT images and clinical data of 127 patients diagnosed with non-small cell lung cancer were retrospectively analyzed.The groups were matched by EGFR mutation status and we analyzed the differences in radiomics and clinical data.Results There was no significant difference between EGFR mutation and age,size,SUVmax,and carcinoembryonic antigen(CEA)(P>0.05),but there was correlation with smoking history,gender,and carbohydrate antigen(CA125)(P<0.05).The CT radiomic model successfully discriminated EGFR status(AUC=0.775).Combining this model with a clinical model(AUC=0.786)improved prediction accuracy(AUC=0.867),but there was no significance(P>0.05).The PET radiomic model successfully discriminated EGFR status(AUC=0.819).When combined with a clinical model(AUC=0.786),it substantially improved its performance(AUC=0.927).Conclusion Gene mutations can drive different tumor imaging phenotypes,and advanced biomarkers based on radiomics features and clinical features may be used to predict EGFR mutation status.