A Radiomics Nomogram For The Prediction Of The EGFR Mutation Status In Peripheral Lung Adenocarcinoma

Objective: To develop and validate a radiomics nomogram for the prediction of epidermal growth factor receptor(EGFR)mutation status in peripheral lung adenocarcinomas.Materials and Methods: Three hundred and forty cases of peripheral lung adenocarcinoma confirmed by pathology with complete clinical,imaging and genetic data.Based on leave-one-out method,they were divided into a training set(n =226)and a validation set(n=114).Three hundred and ninety-six features were extracted from arterial-phase computed tomography(CT)images of each patient.Least absolute shrinkage and selection operator(LASSO)regression was used to select features and generate a radiomics signature in the training set.A radiomics nomogram was built with multivariate logistic regression by integrating the radiomics signature and clinical variables,its calibration and discrimination were evaluated.Finally,decision curve analysis was performed with the combined training and validation sets to assess the clinical usefulness of the nomogram.The descriptions of clinical features include Gender,age,smoking status,histologic subtype,TNM stage(early or Advanced)and CT signs.The objectives of CT signs include: maximum diameter on axial images(Dmax),ground-glass opacity(GGO),lobulation sign,burr sign,pleural indentation sign,vascular cluster sign,air bronchus sign and vacuole sign of tumor.The clinical risk factors of EGFR mutation were analyzed by Fisher's test or Wilcox-on rank sum test,univariate regression analysis and multi-logistic analysis,and the clinical-pathological model was established.There was a statistical significance when P was less than 0.05.Results: In this study,EGFR mutations in 220 cases(64.7%),EGFR wild-type mutations in 120 cases(35.3%).EGFR mutation rates were significantly higher in women(138 of182 [75.8%])than in men(82 of 158 [51.9%],P<0.05)and in nonsmokers(183 of 260[70.4%])than in smokers(37 of 80 [46.3%],P<0.05).EGFR mutations were also significantly more frequent in patients with early stage(203 of 289 [70.2%])than in advanced stage(17 of 51 [33.3%],P<0.05).When tumors were divided into lepidic predominant adenocarcinomas(adenocarcinoma in situ,minimally invasive adenocarcinoma,and lepidic predominant invasive adenocarcinoma)and other subtypes of dominant histologic findings(acinar,papillary,micropapillary,and solid predominant adenocarcinoma,as well as invasive mucinous adenocarcinoma,EGFR mutations were more likely to be lepidic predominant adenocarcinomas(59 of 72 [81.9%])than other subtype(161 of 268 [60.1%],P<0.05).In regard to CT signs,EGFR mutations were significantly more frequent with GGO(123 of 153 [80.4%])than without GGO(97 of 187 [51.9%],P<0.05).Except for GGO,there was no significant difference in lobulation sign,burr sign,pleural indentation sign,vascular cluster sign,air bronchus sign and vacuole sign.The results of the multiple logistic regression analysis revealed that cancer staging and tumor with GGO were independent predictors for EGFR mutation status in peripheral lung adenocarcinoma.Consisting of 13 radiomics features,the radiomics signature has favorable prediction efficacy in training set(AUC:0.739;95% CI,0.686-0.792)and the validating set(AUC:0.730;95% CI,0.624-0.835).The radiomics nomogram has a favorable discriminatory ability for the prediction of EGFR in the training set(AUC,0.80;95%CI,0.75-0.85)than with the clinical model(AUC,0.72;95%CI,0.67-0.78);in the validation set,the radiomics nomogram(AUC,0.84;95%CI,0.78-0.91)also performed better than the clinical model(AUC,0.74;95%CI,0.65-0.82).Decision-curve analysis confirmed the clinical utility of the radiomics nomogram.Conclusion: The presented radiomics nomogram,a non-invasive prediction tool that incorporates the radiomics signature and clinical variables,shows favorable predictive efficiency for predicting EGFR mutation in patients with peripheral lung adenocarcinoma.