Using a hybrid PET/MR system that enables precise image co-registration of 18FDG-PET and DWI, we evaluate the utility of voxel-based analysis of SUVs and ADCs to differentiate CNS lymphomas from GBs. Fifteen GBs and five CNS lymphomas were enrolled. DWI and 18FDG-PET were acquired. Volume, SUVmax, ADCmin, correlation coefficient, elliptical 95% area of bivariate normal distribution of SUVs and ADCs and elliptical 95% area/volume were evaluated. Elliptical 95% area/volume in CNS lymphomas was higher than that of GBs and showed the best diagnostic performance. Voxel-based analysis of 18FDG-PET/MR may be useful to differentiate CNS lymphomas from GBs.