This paper introduces a hybrid framework for early and accurate classification and segmentation of prostate cancer using deep learning. The framework consists of two stages: classification and segmentation. In the classification stage, eight pre-trained convolutional neural networks (CNNs) are fine-tuned using the Aquila optimizer to classify patients with prostate cancer from normal ones. If a patient is diagnosed with prostate cancer, the segmentation stage uses the U-Net model to segment the cancerous spot from the overall image. The framework is trained on three different datasets to ensure generalization. The best reported classification accuracies are 88.91% using MobileNet for the "ISUP Grade-wise Prostate Cancer" dataset and 100% using MobileNet and ResNet152 for the "Transverse Plane Prostate Dataset" dataset. The U-Net model achieves an average segmentation accuracy of 98.46% and an AUC of 0.9778 using the "PANDA: Resized Train Data (512 × 512)" dataset. The results indicate the acceptable performance of the proposed framework. The paper also discusses related studies, methodology, experiments, and comparisons with state-of-the-art approaches.This paper introduces a hybrid framework for early and accurate classification and segmentation of prostate cancer using deep learning. The framework consists of two stages: classification and segmentation. In the classification stage, eight pre-trained convolutional neural networks (CNNs) are fine-tuned using the Aquila optimizer to classify patients with prostate cancer from normal ones. If a patient is diagnosed with prostate cancer, the segmentation stage uses the U-Net model to segment the cancerous spot from the overall image. The framework is trained on three different datasets to ensure generalization. The best reported classification accuracies are 88.91% using MobileNet for the "ISUP Grade-wise Prostate Cancer" dataset and 100% using MobileNet and ResNet152 for the "Transverse Plane Prostate Dataset" dataset. The U-Net model achieves an average segmentation accuracy of 98.46% and an AUC of 0.9778 using the "PANDA: Resized Train Data (512 × 512)" dataset. The results indicate the acceptable performance of the proposed framework. The paper also discusses related studies, methodology, experiments, and comparisons with state-of-the-art approaches.