¹⁸F-FDG PET/CT Based Radiomics In Predicting The Prognosis For Patients With Natural Killer/T Cell Lymphoma

Objective:To develop an ¹⁸F-fluorodeoxygulucose(¹⁸F-FDG)positron emission tomography/computed tomography（PET/CT）radiomics-based model for predicting the prognosis of patients with extranodal natural killer/T cell lymphoma（ENKTL）and evaluate the predictive value of this radiomics-based model.Materials and methods:37 patients with pathologically proven extranodal NK/T cell lymphoma who underwent ¹⁸F-FDG PET/CT examination at the First Affiliated Hospital of Fujian Medical University from December 2014 to August 2020 were retrospectively included.We used LIFEx software to delineate the tumor area and extract 50 PET and 48 CT radiomic features in the regions of interest and collected the corresponding clinical data（including the age,gender,laboratory examination,etc.）.The least absolute shrinkage operator（LASSO）and 10-fold cross-validation were performed for selecting the most significant and valuable features screening and building a radiomics model.According to clinical follow-up and imaging examination,the patient’s progression free survival（PFS）and overall survival（OS）were determined.The receiver operating characteristic curve（ROC）and area under curve（AUC）were used to evaluate the predictive performance of the model.Additionally,based on whether the enrolled patients had a death outcome,these patients were further divided into two groups,and the average radiological signature score（Rad-score）of the two groups was calculated.COX proportional hazards regression analysis was performed to obtain the prognostic factors of OS and PFS.The performance of the radiomics-based modelwasevaluatedusingthe Kaplan-Meier analysis.Results:With the use of LASSO algorithm and 10-fold cross-validation,5（3 PET and 2 CT）and 7（4 PET and 3 CT）radiomic features were selected as the best subsets for predicting OS and PFS,respectively.For predicting the OS of the patients with ENKTL,the AUC value and the corresponding 95%confidence interval（confidence interval,CI）,sensitivity,specificity and accuracy were 0.783（95%CI:0.624-0.942）,0.647,0.955,and 0.744,respectively,from this ¹⁸F-FDG PET-CT based-radiomics model.In the prediction of PFS,the AUC value and the corresponding 95%CI,sensitivity,specificity,and accuracy were 0.805（95%CI:0.643-0.967）,0.765,0.909,and 0.769,respectively.We found a significant difference in Rad-score between these two groups（P=0.0042,P<0.05）.COX risk regression was performed on the Rad-scores of OS and PFS,in which OS and PFS were divided into high-risk and low-risk groups.K-M analysis showed that significant differences between these groups（P=0.000013 and P=0.0000091,both P<0.05）.Conclusion:¹⁸F-FDG PET radiomics-based model was highly useful for predicting the prognosis of patients with ENKTL.