A predictive nomogram for mutation status in lung adenocarcinoma manifesting as ground-glass nodules.

Background: The mutation status of epidermal growth factor receptor () in lung adenocarcinoma is significantly associated with postoperative progression-free survival. Computed tomography (CT)-based radiomics analysis may have potential value in predicting mutation status. This study aims to explore the predictive capacity of radiomics analysis for mutation status in lung adenocarcinomas presenting as ground-glass nodules (GGNs).

Methods: We included 199 GGNs confirmed by histopathology from 2016 to 2020. The clinical factors and radiographic characteristics were counted and evaluated. All GGNs were manually delineated and the radiomics features were extracted, using the least absolute shrinkage and selection operator for feature selection. Then the radiographic, radiomics, and combined nomogram model were constructed respectively, and compared with each other. Decision curve analysis (DCA) was used to assess the clinical usefulness of the models, while receiver operating characteristic curves and calibration curves were used to evaluate their predictive performance.

Results: Univariate analysis revealed five variables that were significantly different between the mutant and wild-type groups. Fifteen radiomics features were significantly associated with mutations. Among the three models, both the radiomics [area under the curve (AUC) =0.818] and the nomogram (AUC =0.820) had good discriminatory ability in predicting mutation status and performed consistently in the validation cohort (AUC =0.805, and 0.833, respectively), with higher predictive performance than the radiographic model. The DCA showed that when it comes to mutation status prediction, the nomogram and the radiomics model showed better overall net benefit than the radiographic model.

Conclusions: For preoperatively predicting the status of mutation in lung adenocarcinomas manifesting as GGNs, the CT-based radiomics analysis will be valuable.