Permeability measurement using dynamic susceptibility contrast magnetic resonance imaging enhances differential diagnosis of primary central nervous system lymphoma from glioblastoma

Objectives: To test if adding permeability measurement to perfusion obtained from dynamic susceptibility contrast MRI (DSC-MRI) improves diagnostic performance in the differentiation of primary central nervous system lymphoma (PCNSL) from glioblastoma. Materials and methods: DSC-MRI was acquired in 145 patients with pathologically proven glioblastoma (n = 89) or PCNSL (n = 56). The permeability metrics of contrast agent extraction fraction (E_x), apparent permeability (K_a), and leakage-corrected perfusion of normalized cerebral blood volume (nCBV_res) and cerebral blood flow (nCBF_res) were derived from a tissue residue function. For comparison purposes, the leakage-corrected normalized CBV (nCBV) and relative permeability constant (K₂) were also obtained using the established Weisskoff-Boxerman leakage correction method. The area under the receiver operating characteristics curve (AUC) and cross-validation were used to compare the diagnostic performance of the single DSC-MRI parameters with the performance obtained with the addition of permeability metrics. Results: PCNSL demonstrated significantly higher permeability (E_x, p <.001) and lower perfusion (nCBV_res, nCBF_res, and nCBV, all p <.001) than glioblastoma. The combination of E_x and nCBV_res showed the highest performance (AUC, 0.96; 95% confidence interval, 0.92–0.99) for differentiating PCNSL from glioblastoma, which was a significant improvement over the single perfusion (nCBV: AUC, 0.84; nCBV_res: AUC, 0.84; nCBF_res: AUC, 0.82; all p <.001) or E_x (AUC, 0.80; p <.001) parameters. Conclusions: Analysis of the combined permeability and perfusion metrics obtained from a single DSC-MRI acquisition improves the diagnostic value for differentiating PCNSL from glioblastoma in comparison with single-parameter nCBV analysis. Key Points: • Permeability measurement can be calculated from DSC-MRI with a tissue residue function-based leakage correction. • Adding E_xto CBV aids in the differentiation of PCNSL from glioblastoma. • CBV and E_xmeasurements from DSC-MRI were highly reproducible.