Intravoxel incoherent motion MRI in the brain: Impact of the fitting model on perfusion fraction and lesion differentiability

Purpose: To investigate the effect of the choice of the curve-fitting model on the perfusion fraction (fIVIM) with regard to tissue type characterization, correlation with microvascular anatomy, and dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) parameters. Several curve-fitting models coexist in intravoxel incoherent motion (IVIM) MRI to derive the (fIVIM). Materials and Methods: In all, 29 patients with brain lesions (12 gliomas, 11 meningiomas, three metastases, two gliotic scars, one multiple sclerosis) underwent IVIM-MRI (32 b-values, 0 to 2000 s/mm2) at 3T. fIVIM was determined by classic monoexponential, biexponential, and a novel nonnegative least squares (NNLS) fitting in 352 regions of interest (lesion-containing and normal-appearing tissue) and tested their correlation with DCE-MRI kinetic parameters and microvascular anatomy derived from 57 region of interest (ROI)-based biopsies and their capacities to differentiate histologically different lesions. Results: fIVIM differed significantly between all three models and all tissue types (monoexponential confidence interval in percent [CI 3.4–3.8]; biexponential [CI 11.21–12.45]; NNLS [CI 2.06–2.60]; all P < 0.001). For all models an increase in fIVIM was associated with a shift to larger vessels and higher vessel area / tissue area ratio (regression coefficient 0.07–0.52; P = 0.04–0.001). Correlation with kinetic parameters derived from DCE-MRI was usually not significant. Only biexponential fitting allowed differentiation of both gliosis from edema and high- from low-grade glioma (both P < 0.001). Conclusion: The curve-fitting model has an important impact on fIVIM and its capacity to differentiate tissues. fIVIM may possibly be used to assess microvascular anatomy and is weakly correlated with DCE-MRI kinetic parameters. Level of Evidence: 2. Technical Efficacy: Stage 1. J. Magn. Reson. Imaging 2017;46:1187–1199.