Risk Factors for Misclassification in Predicting EGFR Mutation Status Using PET/CT Imaging Features in Non-Small Cell Lung Cancer Patients

Jiali Li¹Zihang Zeng²Jie Chen¹Tianxing Fang¹Hongjun Liu¹Yong He^1*

• ¹Zhongnan Hospital of Wuhan University Department of Nuclear Medicine, Wuhan, China
• ²Zhongnan Hospital of Wuhan University Department of Radiation and Medical Oncology, Wuhan, China

Objective: This study aims to develop 10 machine learning models based on PET/CT radiomic features to predict epidermal growth factor receptor (EGFR) mutations in non-small cell lung cancer (NSCLC) patients, and identify risk factors contributing to model misclassification. Methods: This study included 277 NSCLC patients from Zhongnan Hospital, Wuhan University, who underwent pre-treatment 18F-FDG PET/CT and EGFR mutation testing. A PET-CT signature (PCS)-nomogram was developed by comparing 10 machine learning algorithms for EGFR prediction. Leave-one-out cross-validation generated model-specific EGFR mutation probabilities of individual patient, with performance disparities analyzed across clinical subgroups. Model performance was assessed via receiver operating characteristic curve, Youden's index, decision curve analysis, and DeLong's test. Results: The PCS-nomogram model constructed with partial least squares generalized linear models (plsRglm) algorithm achieved optimal performance of EGFR mutation prediction in NSCLC patients (training cohort: AUC 0.80; validation cohort: AUC 0.82). Smoking history caused statistically significant performance deterioration in 7/10 machine learning models (|ΔYouden's index| ≥0.1). PCS model demonstrated higher predictive performance in never-smokers than in smokers (AUC 0.90 vs. 0.64, P < 0.05). Conclusion: A plsRglm-based PCS-nomogram model was proposed for non-invasive prediction of EGFR mutations in NSCLC patients. Compared to smokers, radiomics-based EGFR mutation prediction demonstrated superior performance in never-smokers.