Perinodular and intranodular radiomic features on 18F-FDG PET/CT images predict PD-L1 status in non-small cell lung cancer

Abstract Background Programmed death-ligand 1 (PD-L1) status can affect the efficacy of immunotherapy. Based on the link between PD-L1 expression and metabolism, we hypothesize that radiomics based on 18F-FDG PET/CT can provide clinical decision support for patients with non-small cell lung cancer (NSCLC). Methods In this retrospective study, 18F-FDG PET/CT images and clinical data of 394 eligible patients were divided into training (n = 315) and test sets (n = 79). Radiomic features of the CT and PET images were extracted and these were analysed using five different machine learning classifiers. Combining semantic features with the best radiomic model, a clinical-radiomic model was established to predict the PD-L1 status in patients with NSCLC. Results The radiomic model based on PET intranodular features preform the best, yielding an area under receiver operating characteristics curve (AUC) of 0.791 (95% CI: 0.788, 0.799) on the test set. By adding semantic features, the test set AUC improved to 0.795 (95% CI: 0.794, 0.803). Conclusion This study showed that 18F-FDG PET/CT-based radiomics could be used as a non-invasive biomarker to provide more accurate and personalized decision support for patients with NSCLC.