Rapid measurement of intravoxel incoherent motion (IVIM) derived perfusion fraction for clinical magnetic resonance imaging

Emma M. Meeus, Jan Novak, Hamid Dehghani, Andrew C. Peet

This study aimed to investigate the reliability of intravoxel incoherent motion (IVIM) model derived parameters D and f and their dependence on b value distributions with a rapid three b value acquisition protocol. Diffusion models for brain, kidney, and liver were assessed for bias, error, and reproducibility for the estimated IVIM parameters using b values 0 and 1000, and a b value between 200 and 900, at signal-to-noise ratios (SNR) 40, 55, and 80. Relative errors were used to estimate optimal b value distributions for each tissue scenario. Sixteen volunteers underwent brain DW-MRI, for which bias and coefficient of variation were determined in the grey matter. Bias had a large influence in the estimation of D and f for the low-perfused brain model, particularly at lower b values, with the same trends being confirmed by in vivo imaging. Significant differences were demonstrated in vivo for estimation of D (P = 0.029) and f (P < 0.001) with [300,1000] and [500,1000] distributions. The effect of bias was considerably lower for the high-perfused models. The optimal b value distributions were estimated to be brain500,1000, kidney300,1000, and liver200,1000. IVIM parameters can be estimated using a rapid DW-MRI protocol, where the optimal b value distribution depends on tissue characteristics and compromise between bias and variability.