Differentiation between primary CNS lymphoma and glioblastoma. Qualitative and quantitative analysis using arterial spin labeling MR imaging

To evaluate the diagnostic performance of arterial spin labelling perfusion weighted images (ASL-PWIs) to differentiate primary CNS lymphoma (PCNSL) from glioblastoma (GBM). ASL-PWIs of pathologically confirmed PCNSL (n = 21) or GBM (n = 93) were analysed. For qualitative analysis, tumours were visually scored into five categories based on ASL-CBF maps. For quantitative analysis, normalised CBF values were derived by contralateral grey matter (GM) in intra- and peritumoral areas (nCBF_intratumoral and nCBF_peritumoral, respectively). Visual scoring scales and quantitative parameters from PCNSL and GBM were compared. In addition, the area under the receiver-operating characteristic (ROC) curve was used to determine the diagnostic accuracy of ASL-PWI for differentiating PCNSL from GBM. Weighted kappa or intraclass correlation coefficients (ICCs) were used to assess reliability between two observers. In qualitative analysis, scores 5 (CBF_intratumoral>CBF_GM, 68.8% [64/93]) and 4 (CBF_intratumoral ∼ CBF_GM, 47.6% [10/21]) were the most frequently reported scores for GBM and PCNSL, respectively. In quantitative analysis, both nCBF_intratumoral and nCBF_peritumoral in PCNSL were significantly lower than those in the GBM (nCBF_intratumoral, 0.89 ± 0.59 [mean and SD] vs. 2.68 ± 1.89, p < 0.001; nCBF_peritumoral, 0.17 ± 0.08 vs. 0.45 ± 0.28, p < 0.001). nCBF_peritumoral demonstrated the best diagnostic performance (area under the ROC curve: visual scoring, 0.814; nCBF_intratumoral, 0.849; nCBF_peritumoral, 0.908; p < 0.001 for all). Interobserver agreements for visual scoring (weighted kappa = 0.869), nCBF_{intratumoralGM} (ICC = 0.958) and nCBF_{peritumoralGM} (ICC = 0.947) were all excellent. ASL-PWI performs well in differentiating PCNSL from GBM in both qualitative and quantitative analyses. (orig.)