Prostate cancer (PCa) is the most common malignancy among males worldwide, and is the second leading cause of cancer death among men in United States[1]. DWI plays an important role in multiparametric MRI of prostate. The last decade has witnessed rapid developments in DWI using multiple b values, such as DWI based on IVIM (intravoxel incoherent motion) or kurtosis model. But stretched-exponential model (SEM) DWI for prostate is very few. Liu, X, et al [2] used SEM to compare tumor foci with normal peripheral zone and normal central gland tissues, concluding that both DDC (distributed diffusion coefficient) and α parameters from SEM-DWI differ from cancer to normal prostate tissues. As we know, PCa is the most common malignant tumor and benign prostate hyperplasia（BPH）is the most common benign disease among medium-elderly men. Previous researches usually take normal prostate of relatively young men as control group to compare with PCa. Our study firstly assesses SEM-DWI performance on differentiation between PCa and BPH, and investigates two new parameters (DDC and α), comparing to monoexponential DWI.This retrospective study was approved by the local ethics committee, and written informed consent was obtained from each participant. Subjects: 30 cases of PCa (Mean age 65.1 yrs.; age range 45-86 yrs.；TPSA range 3.0-409.6 ng/mL) and 30 cases of BPH (Mean age 63.0 yrs.; age range 52-84 yrs.；TPSA range 2.7-36.85 ng/mL) verified by TRUS-guided biopsy or histopathology following radical prostatectomy were enrolled. In PCa group, 8, 16, and 6 patients had Gleason score of 3＋3, 3＋4, >3＋4, respectively. The MR experiment was conducted on a Philips 3.0T clinical scanner (Achieva TX, Best, Netherlands). A 16-channel SENSE Torso XL coil was used for signal reception. DWI was performed using single-shot echo-planar imaging, scanning parameters were as follows: TR/TE= 2000ms/67ms, FOV= 240mm ×240 mm; flip angle= 90°, matrix= 120×160; slice thickness= 3.5 mm; NSA=4; number of slices=20. Diffusion in 3 directions was measured by using b values of 0, 500, 1000 and1500 s/mm2. Data were postprocessed by monoexponential and SEM model for quantitation of ADC (apparent diffusion coefficient), DDC and α values. Regions of interest (ROIs) were drawn on tumor foci, central gland (CG) or peripheral zone (PZ) in prostates of BPHs, using T2WI, DWI and histopathology result as references. Independent samples t-test was used to compare ADC, DDC and α value between PCa and BPH group. The correlation between ADC and DDC was assessed using Pearson’s correlation coefficient. ROIs positions of tumor, ADC, DDC and α maps are shown in Figure1. ADC, DDC, α values for different prostate tissue and independent samples t-test results are presented in Figure 2. The ADC and DDC values of PCa were both significantly lower than those of BPH, and significantly lower in CG-BPH than in PZ-BPH. But there was no significantly statistical difference (p>0.05) in parameter α among three tissues. ADC and DDC values had good correlations of each other in different tissue (0.908 for PCa, 0.820 for PZ-BPH and 0.939 for CG-BPH, p<0.0001, respectively). Our study demonstrated that α value failed to distinguish PCa from BPH, which is consistent with previous research by Toivonen, J., et al [3] but contrary to the research by Liu, X, et al[2]. In Liu’research, α values of PCa were significantly lower than those of PZ and CG in normal prostate. Firstly, in Liu’s study, ROIs in control group were drawn on the normal-looking and pathologically confirmed normal PZ and CG tissue, which can be chosen from lateral tissue or other prostatic puncture and biopsy points without cancer cells from one sample. But in our study we only chose one ROI in one sample either in PCa or BPH group, which may narrow interior-group difference and decrease the possibility of mixture of malignant and benign tissue in a same ROI. But the best solution is surely to avoid match bias between MR images and histopathology by using whole-mount prostatectomy sections. Secondly, parameter α describes the deviation of water diffusion from a single exponential decay, and associated with histological heterogeneity. However the structure of prostate cancer is more homogeneous than other tumor types like glioma, because it seldom has gross hemorrhage, necrosis or calcification. Moreover the highest b value in our study was only 1500 s/mm2 due to limitations in the SNR and scan time.Our study shows that stretched-exponential model of DWI is feasible for prostate MRI examination, and DDC can be a new parameter to distinguish cancer from BPH, while there is no significant difference in parameter α between benign and malignant tissues of prostate.