Abstract: Objective To assess the value of diffusion tensor imaging (DTI) combined with T2-weighted imaging (T2WI) in detecting the extra-capsular extension (ECE) of the peripheral zone in localized prostate cancer. Methods From December 2018 to August 2019,17 patients with biopsy-proven prostate cancer underwent preoperative multi-parametric (T2WI+DWI+DTI) 3 tesla MRI examination before radical prostatectomy (RP) were included in our study.All specimens were processed by whole-mount sections at apex,mid-part and base of prostate.Before analysis,slides were all divided into the right part and the left part and we got 102 objects in all.The pathology reports showed that 16 objects (15.8%) consisted of prostate peripheral zone ECE.Utilizing T2WI and PI-RADS v.2 score,we calculated sensitivity,specificity,PPV,and NPV of each parameter to diagnose prostate peripheral zone ECE.Then,ROC analysis was used to calculate the cut-off values of 4 DTI tractography parameters,including fractional anisotropy (FA) values,mean diffusivity (MD),fiber number and density.The potential values were then used in accompany with PI-RADS score to reevaluate patients and assess its sensitivity and specificity for peripheral zone ECE diagnosis. Results PI-RADS score based on T2WI demonstrated a specificity of 86%,a sensitivity of 69%,a PPV of 48%,and a NPV of 94% in detection of peripheral zone ECE.In the quantitative analysis of DTI parameters,the area under ROC curve of FA was 0.805,with a higher sensitivity of 94%,a specificity of 74%,and the optimal cut-off value was 0.301.FA demonstrated greater diagnostic performance when comparing to MD,fiber number,and fiber density.The combination of PI-RADS score and FA demonstrated a high sensitivity (100%) but a low specificity (60%).The logistic regression model that employed PI-RADS score and FA value achieved a high sensitivity (88%) and a high specificity (91%). Conclusion The combination of morphologic component of T2WI and functional aspect of DTI improves sensitivity but reduce specificity for the diagnosis of peripheral zone ECE.Logistic regression modeling provides a favorable solution with higher sensitivity and specificity.