Prostate Cancer (PCa) is considered as one of the widely cancer diseases affect men around the globe. Research found about 16.67% of men affects by PCa in their life specially after 40 ages. Studies reported that one of every six men are suffering from Prostate Cancer approximately. The traditional methods of PCa diagnosis are considered a tedious process and subjected to human errors. Therefore, research tried to develop an efficient diagnostic technique for early detection of PCa which can make a great clinical treatment impact. This paper presents an intelligent system for automated the process of detecting prostate cancer by using MRI images with convolution neural network approaches. Five pre-trained transfer-learning models are used here including: Inception-v3, Inception-v4, Inception-Resent-v2, Xception, and PolyNet. The dataset consisted of 1524 prostate MRIs, which were divided into three parts: 1067 MRI for training, 304 MRI for validation, and 153 MRI for test purposes. Initially, the dimension of MRI images was resized to reduce the complexity and computation processing. Then, the training phase conducted by forwarding MRI images into transfer learning models individually for feature extractions and PCa classification. The transfer learning models were used to classify MRI prostate images into two sets: positive (significant) and negative (nonsignificant) results. Finally, experiments were conducted to evaluate the pre-trained models using both validation and testing datasets. The experimental results showed a robust and high accuracy recognition rate of detecting PCa for each model where Inception-v3 = 98.69, Inception-v4= 96.73%, Inception-Resent-v2= 96.73%, Xception=95.42%, and PolyNet=99.34%. We found here the ability of proposed transfer learning models to detect prostate cancer using MRI images successfully with a high accuracy recognition rate.