CT Radiomic Features Predicting Epidermal Growth Factor Receptor Mutation In Peripheral Lung Adenocarcinoma

Objective:To predict epidermal growth factor receptor(EGFR)mutation status in patients with lung adenocarcinoma using quantitative radiomic biomarkers and representative clinical variables.Method:A retrospective analysis of 104 patients of pathologically proved as lung adenocarcinoma,who did lung surgery and detectived EGFR mutation were enrolled from 2016 to 2018 in the First Hospital of Jilin University.Using a radiomic method,radiomic features that reflect the heterogeneity and phenotype of tumors were extracted from patients' pre-therapy computed tomography(CT)scans.Afterwards,the least absolute shrinkage and selection operator(LASSO)method was applied to select the features that were most distinguishable.Two logistic regression models were established to predict the mutation status of EGFR by using the radiomic biomarkers and the combination of radiomic biomarkers and clinical characteristics.The prediction efficacy of the two models was evaluated with the 5-fold cross-validation,ROC analysis and AUC to obtain the optimal model for predicting the EGFR mutation.Results:1025 3D radiomic features were extracted and reduced to 13 features as the most important predictors to build the radiomics signatures.The radiomic biomarkers combined with clinical features model was developed with radiomics signatures,sex,smoking,vascular_infiltration and pathological_type,whose AUC for the training cohort was 0.88±0.03(mean±standard deviation),the AUC for the verification cohort was 0.88±0.12 and the AUC for the test cohort was 0.913,which were superior to prediction model that used radiomic variables alone,whose training cohort was 0.92±0.01,the AUC for the verification cohort was0.84±0.04 and the AUC for the test cohort was 0.837.Conclusion:The radiomic biomarkers combined with clinical features model consisting of radiomics signatures,sex,smoking,vascular_infiltration and pathological_type,presenting with radiomics nomogram,could effectively predict the EGFR mutation status in the peripheral lung adenocarcinoma patients,which had better predictive performance than the simple radiomics features model.