Comparison of apparent diffusion coefficient and intravoxel incoherent motion for differentiating among glioblastoma, metastasis, and lymphoma focusing on diffusion-related parameter

Abstract

Background and Purpose Brain tumor cellularity has been assessed by using apparent diffusion coefficient (ADC). However, the ADC value might be influenced by both perfusion and true molecular diffusion, and the perfusion effect on ADC can limit the reliability of ADC in the characterization of tumor cellularity, especially, in hypervascular brain tumors. In contrast, the IVIM technique estimates parameter values for diffusion and perfusion effects separately. The purpose of our study was to compare ADC and IVIM for differentiating among glioblastoma, metastatic tumor, and primary CNS lymphoma (PCNSL) focusing on diffusion-related parameter. Materials and Methods We retrospectively reviewed the data of 128 patients with pathologically confirmed glioblastoma (n = 55), metastasis (n = 31), and PCNSL (n = 42) prior to any treatment. Two neuroradiologists independently calculated the max imum IVIM-f (f max) and min imum IVIM-D (D min) by using 16 different b-values with a bi-exponential fitting of diffusion signal decay, min imum ADC (ADC min) by using 0 and 1000 b-values with a mono-exponential fitting and max imum normalized cerebral blood volume (nCBV max). The differences in f max, D min, nCBV max, and ADC min among the three tumor pathologies were deter min ed by one-way ANOVA with multiple comparisons. The f max and D min were correlated to the corresponding nCBV and ADC using partial correlation analysis, respectively. Results Using a mono-exponential fitting of diffusion signal decay, the mean ADC min was significantly lower in PCNSL than in glioblastoma and metastasis. However, using a bi-exponential fitting, the mean D min did not significantly differ in the three groups. The mean f max significantly increased in the glioblastomas (reader 1, 0.103; reader 2, 0.109) and the metastasis (reader 1, 0.105; reader 2, 0.107), compared to the primary CNS lymphomas (reader 1, 0.025; reader 2, 0.023) (P < .001 for each). The correlation between max and the corresponding nCBV was highest in glioblastoma group, and the correlation between D min and the corresponding ADC was highest in primary CNS lymphomas group. Conclusion Unlike ADC value derived from a mono-exponential fitting of diffusion signal, diffusionrelated parametric value derived from a bi-exponential fitting with separation of perfusion effect doesn't differ among glioblastoma, metastasis, and PCNSL.