In Taiwan, occurrence rate of prostate cancer has been going up over the past few decades. In order to help urologists to detect prostate cancer, a prostate cancer detection system in dynamic MRIs is proposed in this paper. Dynamic MRIs are commonly used for auxiliary tool in clinical study and helpful for diagnosing prostate cancer. Firstly, an ACM (Active Contour Model) is trained and used to segment the prostate. Secondly, 136 features are extracted from the dynamic MRIs after injection at different time (0, 20, 60 and 100 second respectively) and transformed them into RIC curves. Thirdly, 10 discriminative features are selected by FDR (Fisher's Discrimination Ration) and SFFS (Sequential Forward Floating Selection). Finally, the SVM classifier is adopted to classify the segmented prostate into two categories: tumor and normal. Experimental results showed that the accuracy of the proposed method is up to 94.7493%.