Purpose:The purpose of the study was to develop and evaluate a method aiming at reducing the estimation error of the intravoxel incoherent motion (IVIM) model parameters reduced-error IVIM (reIVIM) and to compare the estimated parameters to those measured using the conventional "segmented" IVIM (seIVIM) algorithm.Materials and Methods: The prospective clinical component of this study was approved by the Institutional Review Board; all patients signed informed consent.10 patients who underwent pre-treatment magnetic resonance (MR) imaging between July 2014 and May 2015 at 3-Tesla as part of a prospective study of the effect of radiation treatment on IVIM parameters were included.Diffusion-weighted MR images at 17 b-values (0, 10, 40, 70, 90, 100, 110, 120, 170, 210, 240, 270, 390, 530, 620, 750, and 1000 s/mm 2 ) were acquired at 4-8 consecutive time points.Using pre-treatment data, IVIM parameters (pseudo-diffusion D*, true diffusion D, and perfusionfraction f) were estimated from tumor regions of interest using the seIVIM and reIVIM algorithms.The repeatability of IVIM parameters for each algorithm was evaluated through assessment of the coefficient of variation (CV).For simulated data, precision and accuracy were evaluated as a function of noise.Results: Differences in CV for D and f between seIVIM and reIVIM (mean ± standard deviation %: 5.95 ± 2.63 and 5.94 ± 2.78 for D and 13.26 ± 4.94 and 10.54 ± 3.83 for f, respectively) were not significant (P = 0.99 and P = 0.18 for D and f, respectively).Differences in CV for D* and f × D* between seIVIM and reIVIM (40.88 ± 14.80% and 20.51 ± 7.71% for D* and 45.10 ± 13.40% and 22.50 ± 8.48% for f × D*, respectively) were significant (P = 0.0003 and P = 0.0003 for D* and f × D*, respectively).Conclusion: The proposed approach to the voxel-wise analysis of IVIM data results in the improved estimation of the pseudo-diffusion parameter by reducing the variability of the measurement.