MRI-based Radiomics As Response Predictor To Radiochemotherapy For Metastatic Cervical Lymph Node In Nasopharyngeal Carcinoma

Objective: To explore the feasibility of predicting the efficacy of chemoradiotherapy for nasopharyngeal carcinoma with cervical metastatic lymph nodes based on MRI radiomic.Materials and method: A total of 145 consecutive patients with NPC were recruited at Sichuan Cancer Hospital between October 2016 and November 2019 including 102 in primary cohort and 43 in validation cohort.Metastatic lymph nodes were diagnosed according to radiologic criteria and treatment response was evaluated according to the Response Evaluation Criteria in Solid Tumors.A total of 2704 radiomic features were extracted from contrast-enhanced T1-weighted imaging（CE-T1WI）and T2-weighted imaging（T2WI）for each patient,and were selected to construct radiomic models for model 1（CE-T1WI）,model 2（T2WI）,and model 3（combined CE-T1 WI and T2WI）,respectively.The area under curve（AUC）of receiver operating characteristic,sensitivity,specificity,accuracy,positive predictive value and negative predictive value were used to estimate the performance of these radiomic models in predicting treatment response of metastatic lymph node.The calibration curve was performed to assess the goodness-offit of radiomic models.Finally,decision curve analysis was conducted to determine the clinical usefulness of the radiomics model by quantifying the net benefits at different threshold probabilities in the validation dataset.Results: No significant difference of AUC was found among radiomic models of CET1 WI,T2WI,and combined CE-T1 WI and T2 WI in the primary and validation cohorts（all P > 0.05）.For model 3,the AUC,sensitivity,specificity and accuracy were 0.927[95%CI,0.878-0.975],0.911 [95%CI,0.804-0.970],0.826 [95%CI,0.686-0.922] and0.872 [95%CI,0.792-0.930] in primary cohort,and were 0.772 [95%CI,0.624-0.920],0.792 [95%CI,0.578-0.929],0.790 [95%CI,0.544-0.939] and 0.791 [95%CI,0.640-0.900] in validation cohort.The Hosmer-Lemeshow（H-L）tests yielded non-significant results in the both primary and validation cohorts,which suggested no departure from a perfect fit.While the probability of predicting response group ranges from 15% to 95%,using the proposed radiomics model to predict response group showed a greater advantage than either the scheme in which all patients were assumed to response group or the scheme in which all patients are Non-response group.Decision curve analysis demonstrated that the radiomics model was clinically useful.Conclusion: The MRI radiomic model can be used as a biomarker to evaluate the effect of chemoradiotherapy on cervical metastatic lymph nodes of nasopharyngeal carcinoma,and provide support for clinical decision-making of patients with nasopharyngeal carcinoma.