Prediction model based on 18F-FDG PET/CT radiomic features and clinical factors of EGFR mutations in lung adenocarcinoma.

The aim of this study was to build a prediction model for epidermal growth factor receptor (EGFR) mutations in lung adenocarcinoma. A retrospective analysis was performed on 88 patients with lung adenocarcinoma. All patients underwent an 18F-FDG PET/CT scan and genetic testing of EGFR before the treatment. In the training set, the radiomic features and clinical factors were screened out, and model-1 based on CT radiomic features, model-2 based on PET radiomic features, model-3 based on clinical factors, and model-4 based on radiomic features combined with clinical factors were established, respectively. The performance of the prediction model was assessed by area under the receiver operating characteristic (ROC) curve (AUC). The DeLong test was used to compare the performance of the models to screen out the optimal model, and then built the nomogram of the optimal model. The effect and clinical utility of the nomogram was verified in the validation cohort. In our analysis, model-4 was superior to the other prediction models in identifying EGFR mutations. The AUC was 0.864 (95% CI: 0.777-0.950), with a sensitivity of 0.714 and a specificity of 0.784. The nomogram of model-4 was established. In the validation cohort, the concordance index (C-index) value of the calibration curve of the nomogram model was 0.778 (95%CI: 0.585-0.970), and the nomogram had a good clinical utility. We demonstrated that the model based on 18F-FDG PET/CT radiomic features combined with clinical factors could predict EGFR mutations in lung adenocarcinoma, which was expected to be an important supplement to molecular diagnosis.