Noninvasive prediction of classic EGFR-activating mutations in non-small cell lung cancer using a combined spectral computed tomography model: an auxiliary tool for preoperative stratification of candidates for EGFR-TKI therapy

Background EGFR genotyping is essential for EGFR-TKI therapy in non-small cell lung cancer (NSCLC), but invasive biopsy has limitations. Noninvasive preoperative stratification tools are urgently needed. Methods We retrospectively enrolled 129 NSCLC patients to develop and internally validate a combined model integrating clinical features, conventional CT findings, and dual-layer spectral CT parameters for predicting classic EGFR-activating mutations. Performance was evaluated by ROC, DeLong test, DCA, and 1000-iteration bootstrap validation. Results The combined model achieved an AUC of 0.805, significantly outperforming the clinical-conventional CT model (AUC = 0.642, P < 0.001). Venous-phase normalized iodine concentration (NIC) was the only independent predictor (OR = 2.578, P = 0.018). The model showed robust performance in the lung adenocarcinoma subgroup (AUC = 0.810) with negligible overfitting. Conclusions This noninvasive spectral CT model provides a reliable, easy-to-implement auxiliary tool for preoperative EGFR mutation stratification to optimize EGFR-TKI therapy decisions.