Abstract:Objective To calculate the DWI data of prostate MRI to obtain the parameters of DKI and SEM model, and conduct regression analysis with the prognostic grouping in the postoperative pathological results, so as to evaluate the correlation between the parameters and the prognostic grouping, so as to evaluate the value of the parameters of the above model in distinguishing the malignant degree of prostate cancer and predicting the pathological grade. Methods 71 cases of prostate cancer confirmed by pathology were collected retrospectively. The DWI in MRI scan before surgery was reconstructed to obtain MK and MD of DKI,DDC and α of SEM.ADC was automatically generated by the scanning system; Two doctors outlined ROI on each parameter map according to pathological tips,and made data statistics. After the reliability was tested by Cronbach's alpha, the distribution of parameters between adjacent prognostic groups was compared by Kruskal Wallis test; Wilcoxon rank test was performed on ADC, DDC and MD respectively; Spearman correlation analysis was used to evaluate the correlation between various parameters and prognosis groups; The relationship between various parameters and prognosis groups was curve fitted, and the fitting degree was evaluated by the size of R2. Results Cronbach's alpha coefficient of the score data of the two doctors was 0.910; There were differences in the distribution of ADC,α,DDC,MD and MK in each prognostic group (all P<0.05); there was significant difference in the distribution of ADC and DDC (P<0.01), and there was no significant difference in the distribution of ADC and MD (P>0.05); ADC, DDC and MD were negatively correlated with prognosis, r values were -0.601, -0.627 and -0.566 respectively, and the fitting degree of inverse model was high, R2 were 0.644, 0.749 and 0.643 respectively; MK was positively correlated with prognosis, r=0.537 and R2 of inverse model was 0.345; The correlation and curve fitting between α and prognosis were low, r=0.239. Conclusions In the prediction of prognosis grouping of prostate cancer, the single exponential model is still the most cost-effective method in practical application. DDC in the SEM has better performance than ADC, and MD in the DKI have the same performance with ADC.