Objective       To explore the reproducibility of non-Gaussian diffusion MR imaging using bi- and stretched-exponential models with 22 b-value (0~5 000 s/mm2) in evaluation of water diffusion microenvironment in the healthy brain white matter and gliomas.  Methods       Twelve healthy volunteers and 65 patients with pathologically confirmed glioma were recruited for this study. For the healthy volunteers

2 consecutive enhanced diffusion weighted imaging (eDWI) scans were performed and post-processed using bi- and stretched-exponetial models，respectively. The white matter located at the basal ganglion and semioval centrum was selected for the placement of region of interest (ROI)

with each section containing 4 different ROIs. For the glioma group

eDWI scan was done once. The representative section showing the largest solid tumor was identified and the tumor ROIs were delineated manually twice by two neuroradiologists separately. Mean values of slow- and fast-diffusion coefficients (Dslow and Dfast)

perfusion fraction (PF)

distributed diffusion coefficient (DDC) and intravoxel heterogeneity index α were calculated. The repeatability of these parameters obtained from 2 scans in healthy controls was analyzed by paired t test

coefficient of variability (CV) and Bland-Altman (B-A) plots. Analyses of intraclass correlative coefficient (ICC) and CV were used for testing the agreement of 2 scans of measurement in the same patient with glioma.  Results      For the 12 healthy volunteers

no significant differences in all diffusion parameters were shown between the 2 scans (P0.05). The CV values of all parameters were below 5%. B-A plots illustrated a majority (≥11/12) of  data points within 95% normalized difference for all parameters of eight ROIs in the 2 sections. For 65 cases of gliomas

satifactory agreements were acquired for all the parameters with the ICC value of more than 0.75 (P0.001). Dfast and PF had higher CV values within a range of 5.0%～11.5%

followed by DDC and Dslow. Their CV values were positively associated with the increasing gliomas grade. The α index had the smallest CV value of less than 2%.  Conclusion       Our study shows a good reproducibility of non-Gaussian derived diffusion parameters for the evaluation of water diffusion microenvironment in the healthy brain white matter

whereas the- index α has the best reproducibility among the non=Gaussian derived diffusion parameters in the application of gliomas.