Abstract:Objective To investigate the value of the combination of clinical metrics,CT characteristics and radiomic features to predict the epidermal growth factor receptor (EGFR) and anaplastic lymphoma kinase (ALK) gene mutations of lung adenocarcinoma. Methods The clinical data of 143 patients with lung adenocarcinoma confirmed by operation and pathology,confirmed EGFR and ALK gene status,and performed CT within 3 months before operation were collected retrospectively between January and December 2017 in Shenzhen People′s Hospital.Single factor analysis was used to compare the correlation between EGFR and ALK gene mutations in three clinical features,16 CT features and 94 CT radiologic features.The statistically significant variables were introduced into the multivariate Logistic regression model to obtain the independent risk factors for EGFR and ALK mutation.Diagnostic efficacy was assessed by receiver operating characteristic (ROC) curve. Results Out of 143 patients with lung adenocarcinoma,63 caes(44.1%) were EGFR mutants and 80 cases(55.9%) were wild-type.There were 20 cases (14.0%) of ALK mutation and 123 cases (86.0%) of wild-type.Univariate analysis showed that smoking status (P=0.002),ground glass composition (P=0.042),peritumoral emphysema(P=0.043),mediastinal lymphadenopathy (P=0.020),and a radiomic feature named Size Zone Non Uniformity Normalized(P=0.031) were associated with EGFR mutations.Logistic regression analysis identified that smoking status (OR=2.775,P=0.014),mediastinal lymphadenopathy (OR=2.056,P=0.039) and Size Zone Non Uniformity Normalized (OR=1.906,P=0.045) were independent risk factors for EGFR gene mutations.Age (P=0.002),tumor-lung interface(P=0.002),spicule sign (P=0.021),tumor necrosis (P=0.000) were associated with ALK mutations.Age(OR=2.436,P=0.041),the boundary between tumor and lung (OR=3.367,P=0.032) and tumor necrosis (OR=7.582,P=0.002) were independent risk factors for ALK gene mutations.Areas under the ROC curve (AUC) of combination with clinical metrics,CT characteristics and radiomic features to predict EGFR were 0.694(95%CI:0.608-0.780).The AUC value of ALK gene mutation predicted by clinical and CT features was 0.763 (95% CI:0.617-0.910). Conclusions Combination of clinical metrics,CT characteristics and radiomic features can predict the EGFR and ALK gene mutation status of lung adenocarcinoma and provide decision support for clinical treatment selection.