The aim of this study was to assess if the histogram analysis of mono-exponential, bi-exponential and stretched exponential models of diffusion-weighted MRI (DW-MRI) parameters can differentiate two common subtypes of ovarian epithelial cancer (OEC): high-grade serous carcinomas (HGSCs) and clear cell carcinomas (CCCs).Twenty female patients with ovarian cancer (including 14 high-grade serous carcinomas and 6 clear cell carcinomas) prior to the treatment underwent pelvic MR examination. Scanning sequences included diffusion-weighted magnetic resonance imaging with 10 b values (0, 30, 50, 100, 150, 200, 400,800,1000,1500sec/mm²). Data was post-processed by monoexponential, bi-exponential IVIM and stretched exponential model for quantitation of standard ADC (ADC), slow ADC (D), fast ADC (D*) and perfusion fraction (F), distributed diffusion coefficient (DDC) and alpha（α）. Histogram analysis was performed by outlining entire-tumour regions of interest (ROIs) ^[1]. Based on an entire-tumour measurement, the following histogram parameters were derived from ADC, D, D* , F, DDC and α maps, respectively:(a)mean; (b)the 10th percentile (which indicated the point at which 10 % of the voxel values that form the histogram are found to the left); (c) the mean of the top 10 percent(Meantop). Two representative cases for histogram analysis of DW imaging measures are shown in Figure.1-2. The SPSS20.0 statistical software was used for data analysis in this study, P value less than 0.05 was considered statistically significant. The HGSCs and CCCs subtypes were compared using independent sample t test. And areas under ROC curve between two groups were assessed.For ADC ,D, DDC and α, the histogram mean, the 10th percentile and meantop were significantly lower in HGSCs than in CCCs（P＜0.05）. For F, meantop reflected statistically significant differences between HGSCs and CCCs while the mean and the 10th percentile did not show any significant difference between the two types. All of parameters of D* were of no significant different, the rest of the specific data was listed in Table.1. The meantop of DDC was the single best predictor for classification (AUC, 0.964), followed by the 10th percentile of DDC(AUC,0.952) and meantop of ADC(0.940), more details referent to Table.2 and Figure.3.ADC, D, DDC and α have all showed good diagnostic performance in terms of mean, the 10th percentile and meantop. But for F values, only meantop term has good diagnostic performance. Among all these parameters, the meantop of DDC may have better performance. The histogram parameters reflect the different histological features of two OEC subtypes in some extent ^[2]. The 10th percentile and meantop also allows low ADC components to be evaluated, which may represent the poorly differentiated parts of tumors. However, the difference of histogram D* and F between the two types did not reach statistical significance, which indicates that pseudo-perfusion may contribute little to the diffusivity for predicting the grade of OEC.The histogram analysis of mono-exponential, IVIM and stretched exponential models can help discriminate high-grade serous carcinomas from clear cell carcinomas, which may have an impact on clinical practice.