Differentiation Of Glioblastoma And Primary Central Nervous System Lymphoma Based On The Multi-parametric MR Imaging: A Radiomics Study

Objectives:The aim of this study is to evaluate the feasibility of using radiomics to differentiate the glioblastoma(GBM)and primary central nervous system lymphoma(PCNSL)based on multi-parametric MR imaging.Materials and Methods:A total of 135 patients(GBM=84,PCNSL=51)with newly diagnosed GBM and PCNSL between January 2015 and August 2021 were included retrospectively.The patients were subjected to radiomics analysis using the multi-parametric MRI(contrast-enhanced T1-weighted imaging(CE-T1WI),T2-weighted imaging(T2WI),diffusion-weighted imaging(DWI)and apparent diffusion coefficients(ADC)).The lesions were manually delineated using ITK-SNAP software,and radiomics features were performed using the 3Dslicer software.The patients were randomly split into training set(70% of all patients)and testing set(30% of all patients).Independent-sample t-test and the least absolute shrinkage and selection operator(LASSO)method were used to select the most useful features from the training set,and the ordinary single-sequence radiomics model and multi-sequence radiomics model were constructed.Subsequently,we use Bayesian Information Criterion(BIC)to select the optimal model from all combination models,and rebuilt four single-sequence radiomics models and eleven multi-sequence radiomics models.Finally,the area under the receiver operating characteristic curve(AUC)was used to verify the predictive performance of these models in the test set.Results:For LASSO combined BIC method,the predictive performance of the new model was higher than that of using LASSO alone.The CE-T1 WI model achieved the best performance among all single-sequence models with the AUC of 0.956,accuracy of 0.933,sensitivity of 1.000,and specificity of 0.808.For the multisequence radiomics models,the CE-T1 WI + ADC model achieved the best performance with the AUC of 0.967,accuracy of 0.949,sensitivity of 0.958,and specificity of 0.958.Conclusions:The radiomics base on multi-parametric MRI could well distinguish GBM and PCNSL,which assistant doctors in clinical diagnosis.Since there is no uniform model to obtained best performance for every specific data set,it is necessary to try different combination methods.