Predictive Value Of MRI Radiomics And Relevant Clinical Variable Models For Local Recurrence Of Nasopharyngeal Carcinoma

Objective:To establish and validate a comprehensive nomogram model based on magnetic resonance imaging（Magnetic resonance imaging,MRI）omics features and clinical variables to predict the risk of local recurrence of nasopharyngeal carcinoma（Nasopharyngeal carcinoma,NPC）after treatment.Methods:The 169 patients with nasopharyngeal carcinoma who were pathologically diagnosed in our hospital from January 2014 to October 2021 were retrospectively collected,and all patients were randomly divided into training set（n=118）and validation set（n=51）according to the ratio of 7:3.Based on axial T₂WI and T₁WI enhanced MRI images,this study used 3D Slicer open source software to delineate nasopharyngeal carcinoma whole tumor lesions,and then used the Radiomics plugin in 3D Slicer software to extract radiomics features.Univariate analysis and Least Absolute Shrinkage and Selection（Least absolute shrinkage and selection operators,Lasso）analysis were used to screen radiomics features associated with local recurrence,Logistic regression methods were used to construct radiomics signatures（Rad-Score）,and univariate and multivariate regression analysis was used to select clinical variables that independently predict local recurrence in nasopharyngeal carcinoma.Finally,a visual nomogram for predicting local recurrence in nasopharyngeal carcinoma was established and validated based on Rad-Score and relevant clinical variables,and the area under the curve（AUC）and calibration curve were used to evaluate the predictive performance of the model.A separate validation set is used for validation.Results:There are 169 cases in this paper,118 cases in the training set,and 51 cases in the validation set.After Lasso analysis and Logistic regression method,a Rad-Score containing 6 radiomics features was established,which showed good predictive ability in both the training set and the validation set,AUC values were0.887（95%CI,0.825～0.950）,0.650（95%CI,0.452～0.847）.And the nomogram combining Rad-Score and clinical variables showed better predictive power than Rad-Score alone（AUC value 0.910 vs 0.887）,these results indicated that radiomic features and clinical stage were independent predictors of local recurrence of NPC.In addition,the calibration curve of the nomogram also showed a satisfactory agreement between the probability of local recurrence of NPC predicted by the nomogram and the probability of recurrence actually observed.Conclusion:The nomogram based on MRI radiomics features and clinical variables can be used as a visualization tool for predicting local recurrence in patients with NPC,which can help clinically identify individuals with high risk of local recurrence,provide evidence for early clinical intervention,and improve NPC patients overall survival of patients.