PURPOSES: This study aims to investigate the diagnostic performance of combining 68 Ga-pentixafor PET with MRI to differentiate primary central nervous system lymphoma (PCNSL) from glioblastoma (GBM), particularly focusing on atypical lymphoma identification. PATIENTS AND METHODS: Seventy-one PCNSL and 53 GBM patients who underwent both 68 Ga-pentixafor PET/CT and MRI were retrospectively included. We evaluated the quantitative imaging parameters and MRI features of positive lesions, identifying atypical PCNSL by hemorrhage, necrosis, or heterogeneous enhancement. Logistic regression identified key variables, and the ROC-AUC evaluated their diagnostic value. Immunohistochemistry for CXCR4 was performed. RESULTS: PCNSLs, including 23 atypical cases, showed higher SUV max and TBR, and lower MTV, ADC min , and relative ADC min (rADC min ) than GBMs (all P 's <
0.05). The CXCR4 staining in PCNSL was also more pronounced in GBM ( P = 0.048). Multivariate logistic regression indicated that a combination of TBR, MTV, and ADC min (quantitative model 1) had a superior AUC of 0.913 in distinguishing PCNSL from GBM, outperforming single parameters (all P 's <
0.05). For differentiating atypical PCNSL from GBM, single quantitatively parameters showed moderate performance (AUC, 0.655-0.767). Further combining TBR with ADC min (quantitative model 2) significantly improve the AUC to 0.883. Multiparameter models, incorporating significant quantitative and qualitative MRI features, achieved AUCs of 0.953 (PCNSL vs GBM) and 0.902 (atypical PCNSL vs GBM), significantly outperforming single parameters (all P 's <
0.05). CONCLUSIONS: 68 Ga-pentixafor PET in combination with MRI provides valuable diagnostic information in differentiating PCNSL from GBM, especially for atypical PCNSL.