Prostate cancer is one of the leading causes of cancer-related deaths among men worldwide, and early detection is critical for effective treatment and improved survival rates. Magnetic Resonance Imaging (MRI) is widely used for prostate cancer diagnosis due to its ability to provide detailed anatomical and functional imaging. This study presents a comparative analysis of various Convolutional Neural Networks (CNNs) for the detection and classification of prostate cancer using MRI scans. The study evaluates the performance of popular CNN architectures, including VGG, ResNet, and EfficientNet, in terms of accuracy, sensitivity, specificity, and computational efficiency. A curated dataset of annotated prostate MRI images was preprocessed and augmented to enhance model generalization. Transfer learning techniques were employed to fine-tune the networks, and hyperparameter optimization was conducted to improve model performance. Experimental results reveal significant differences in the effectiveness of CNN architectures, with EfficientNet achieving the highest accuracy and ResNet demonstrating a balance between accuracy and computational efficiency. This comparative analysis underscores the importance of selecting suitable CNN architectures for specific medical imaging tasks and provides insights into optimizing deep learning models for prostate cancer detection. The findings contribute to advancing AI-driven diagnostic tools for more accurate and efficient prostate cancer screening.