Predicting Progression-Free Survival Using MRI-Based Radiomics for Patients With Nonmetastatic Nasopharyngeal Carcinoma.

Objectives To explore the predictive value of MRI-based radiomics model for progression-free survival (PFS) in non-metastatic nasopharyngeal carcinoma (NPC). Methods 327 non-metastatic NPC patients (training cohort (n=230) and validation cohort (n=97)) were enrolled. The clinical and MRI data were collected. The least absolute shrinkage selection operator (LASSO) and recursive feature elimination (RFE) were used to select radiomics features. Five models (Model 1: clinical data, Model 2: overall stage, Model 3: radiomics, Model 4: radiomics + overall stage, Model 5: radiomics + overall stage +EBV DNA) were constructed. The prognostic performances of these models were evaluated by Harrell’s concordance index (C-index). The Kaplan-Meier method was applied for the survival analysis. Results The model 5 incorporating radiomics,overall stage and EBV DNA yielded the highest C-indices for predicting PFS in comparision with model 1, model 2, model 3 and model 4(training cohorts: 0.805 vs 0.766 vs 0.749 vs 0.641 vs 0.563, validation cohorts: 0.874 vs 0.839 vs 836 vs 0.689 vs 0.456). The survival curve showed that the high-risk group yielded a lower PFS than the low-risk group. Conclusions The model incorporating radiomics, overall stage and EBV DNA showed better performance for predicting PFS in non-metastatic NPC patients.