Prediction of EGFR mutation status and its subtypes in non-small cell lung cancer based on 18 F-FDG PET/CT radiological features.

Abstract

Purpose: Prediction of epidermal growth factor receptor (EGFR) mutation status and subtypes in patients with non-small cell lung cancer (NSCLC) based on 18 F-fluorodeoxyglucose ( 18 F-FDG) PET/computed tomography (CT) radiomics features.

Patients and methods: Retrospective analysis of 201 NSCLC patients with 18 F-FDG PET/CT and EGFR genetic testing was carried out. Radiomics features and clinical factors were used to construct a combined model for identifying EGFR mutation status. Mutation/wild-type models were trained in a training cohort ( n  = 129) and validated in an internal validation cohort ( n  = 41) vs an external validation cohort ( n  = 50). A second model predicting the 19/21 mutation locus was also built and evaluated in a subset of EGFR mutations (training cohort, n  = 55; validation cohort, n  = 14). The predictive performance and net clinical benefit of the models were assessed by analysis of the area under curve (AUC) of the subjects, nomogram, calibration curve and decision curve.

Results: The AUC of the combined model distinguishing EGFR mutation status was 0.864 in the training cohort and 0.806 and 0.791 in the internal vs external test sets respectively, and the AUC of the 19/21 mutation site model was 0.971 and 0.867 in the training cohort and internal validation cohort respectively. The calibration curves of the individual models showed better model predictions (Brier score <0.25). Decision curve analysis showed that the models had clinical application.

Conclusion: The combined model based on 18 F-FDG PET/CT radiomics features combined and clinical features can predict EGFR mutation status and subtypes in NSCLC patients, and guiding targeted therapy, and facilitate precision medicine development.