Dynamic contrast-enhanced magnetic resonance imaging-based radiomics for predicting programmed death ligand 1 expression in nasopharyngeal carcinoma.

BACKGROUND: Programmed death-ligand 1 (PD-L1) is a key biomarker for nasopharyngeal carcinoma (NPC) immunotherapy. However, given that it requires an invasive approach based on immunohistochemical testing, novel non-invasive PD-L1 assessment is needed to guide the application of immunotherapy. PURPOSE: To investigate the potential of radiomics to predict the expression of PD-L1 in NPC based on dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI). MATERIALS AND METHODS: We enrolled 108 NPC patients who underwent DCE-MRI and PD-L1 immunohistochemistry. Radiomics features were extracted from DCE-MRI, efflux rate constant (Ktrans), and extracellular extravascular volume (Ve) images. Optimal radiomics features were selected using the least absolute shrinkage and selection operator. Logistic regression was utilized to develop a radiomics model, and its predictive performance was evaluated using receiver operating characteristic (ROC) curves and the area under the ROC curve (AUC). Calibration curves and the Hosmer-Lemeshow test were employed to assess the goodness-of-fit. RESULTS: Patients were randomly assigned into training cohort (n = 75) and validation cohort (n = 33). No significant differences in clinical factors were observed between PD-L1-positive and PD-L1-negative groups. Four features were finally selected to construct the radiomics model. The AUC of the radiomics model were 0.751 and 0.718 for the training and validation cohorts, respectively. The calibration curves for the radiomics model demonstrated excellent agreement between the predictions and observations in both the training (P = 0.447) and validation (P = 0.861) cohorts. CONCLUSION: The DCE-MRI-based radiomics model can predict the expression of PD-L1 in NPC, providing a non-invasive tool to select positive PD-L1 patients.