Diffusional Kurtosis Imaging for Differentiating between High-Grade Glioma and Primary Central Nervous System Lymphoma

Background: The aim of this study was to assess the diagnostic accuracy of diffusion kurtosis magnetic resonance imaging parameters for differentiating high-grade gliomas (HGGs) from primary central nervous system lymphomas (PCNSLs). Methods: Diffusion parameters, including fractional anisotropy (FA), mean diffusivity (MD), axial diffusivity (lambda(//)), radial diffusivity (lambda(perpendicular to)); and kurtosis parameters, including mean kurtosis (MK), axial kurtosis (K-//), and radial kurtosis (K-perpendicular to), were normalized to contralateral normal-appearing white matter (NAWMc) to decrease inter-individual and inter-regional changes across the entire brain, and then compared with the solid parts of 20 HGGs and 11 PCNSLs [median 95% confidence interval (CI), P< 0.004; 0.05/14], significance level, Kolmogorov-Smirnov test, Bonferroni correction]. Results: FA, MD, lambda(//), and lambda(perpendicular to) values were higher in HGGs than in PCNSLs, but not significantly [HGGs: 0.209 (95% CI, 0.134-0.338), 1.385 (95% CI, 1.05-1.710), 1.655 (95% CI, 1.30-2.060), 1.228 (95% CI, 0.932-1.480), respectively; PCNSLs: 0.143 (95% CI, 0.110-0.317), 1.070 (95% CI, 0.842-1.470), 1.260 (95% CI, 0.960-1.930), 1.010 (95% CI, 0.782-1.240)], respectively; P50.120, 0.010, 0.004, and 0.004, respectively). However, MK and K-// were significantly higher in PCNSLs compared with HGGs [PCNSLs: 0.765 (95% CI, 0.697-0.890), 0.787 (95% CI, 0.615-1.030), respectively; HGGs: 0.531 (95% CI, 0.402-0.766), 0.532 (95% CI, 0.432-0.680], respectively; P50.001, 0.000, respectively); but not K-perpendicular to [0.774 (95% CI, 0.681-0.899) for PCNSLs; 0.554 (95% CI, 0.38920.954) for HGGs; P50.024]. Conclusion: There were significant differences in kurtosis parameters (MK and K-//) between HGGs and PCNSLs, while differences in diffusion parameters between them did not reach significance; hence, better separation was achieved with these parameters than with conventional diffusion imaging parameters.