A Risk Assessment Model For Recurrence Or Metastasis Of Nasopharyngeal Carcinoma Was Established Based On MRI Radiomics And Clinical Features

Purpose:A combined risk assessment model for recurrence or metastasis of nasopharyngeal cancer（NPC）was constructed based on the clinical factors and imaging features of the patients initially diagnosed with NPC,and the efficacy of the model was evaluated.Methods:A total of 140 patients with NPC who met the inclusion and exclusion criteria and were initially diagnosed from February 2013 to December 2017 in the Affiliated Hospital of Zunyi Medical University were enrolled and followed up for more than 3 years.The primary endpoint was the recurrence or metastasis free survival（RMFS）,and the secondary endpoint was the cancer specific survival（CSS）.Clinical and pre-treatment imaging information（T stage,N stage,EBV-DNA,T₁WI sequence,T₂WI sequence,and enhanced sequence,etc.）was collected.All patients were randomly assigned to the training cohort and the validation cohort（98 and 42 patients）in a 7:3 ratio.The region of interest（ROI）of tumors was delineated by 3D-Slicer software,and the image features of tumors were extracted by Py Radiomics package.Pearson correlation coefficient（PCC）was used to screen out two features with good correlation,and random forest（RF）was used to determine the most correlated features with recurrence or metastasis.Cox multivariate regression analysis was used to construct clinical model（CM）,Radiomics model（RM）and Clinical-Radiomics model（CRM）in the training cohort.Area under curve（AUC）were used to verify the sensitivity and specificity of the models.All patients were divided into the high and low recurrence or metastasis risk group based on the median risk value in the CRM model.Kaplan-Meier curve was used to compare the 3-year recurrence or metastasis free rate（RMFR）and 3-year CSS rate between the two groups.We constructed the nomograms,calibration curves,and consistency index（C-index）of CRM to evaluate the predictive power of the model.Results:1.The median follow-up time of 140 patients was 61.36（56.89-65.85）months.Among them,41 patients had recurrence or metastasis,and 99 patients did not.2.A total of 960 imaging features were extracted from enhanced MRI.The 46 imaging features associated with recurrence or metastasis were obtained by PCC and RF.3.In the training cohort,9 factors（7 imaging features and 2 clinical factors）were screened out by Cox multivariate regression analysis.Based on the above independent factors,a CRM for evaluating the risk of recurrence or metastasis of nasopharyngeal carcinoma was successfully constructed.The model predicted the AUC value of 3-year RMFS was 0.863,（p<0.001,95%CI:0.715-0.971）,and the sensitivity and specificity were:0.80 and 0.78,respectively;the AUC value of the validation cohort was 0.888 for 3-year RMFS（p<0.001,95%CI:0.833-0.943）,sensitivity and specificity were:0.86 and 0.81,respectively.4.Kaplan-Meier curve showed that the 3-year RMFR and 3-year CSS rate in the CRM with high risk of recurrence or metastasis were significantly lower than those in the low-risk group（p<0.001）.Our Nomogram constructed using the CRM visually assess the risk of recurrence or metastasis,and the C-index of the training cohort was 0.776.The calibration curve showed that the recurrence or metastasis predicted by the model was in good consistence with the actual recurrence or metastasis.5.The CRM was analyzed for the predictive performance of different treatment schemes,and the results showed that the model had good performance in predicting the risk of recurrence or metastasis regardless of the treatment scheme adopted by the subsequently enrolled patients.6.In addition,the CRM showed that in the high-risk group,the 3-year RMFR of patients who received adjuvant chemotherapy was as high as 70.0%,which was significantly higher than that of patients who did not receive adjuvant chemotherapy（46.8%）,p=0.034.It was suggested that the patients who receive adjuvant chemotherapy may improve the 3-year RMFR in the high-risk group;for low-risk patients evaluated by this model,subsequent treatment can be considered without adjuvant chemotherapy.Conclusion:In this study,a combined model for predicting recurrence or metastasis in patients with nasopharyngeal carcinoma was constructed based on imaging features and clinical factors,which could better distinguish the patients with high risk of recurrence or metastasis from patients with low recurrence or metastasis...