Prostate cancer (PCa) is a severe type of cancer and causes major deaths among men due to its poor diagnostic system. The images obtained from patients with carcinoma consist of complex and necessary features that cannot be extracted readily by traditional diagnostic techniques. This research employed deep learning long short-term memory $(LSTM)$ and Residual Net $(ResNet-101)$ , independent of hand-crafted features, and is fine-tuned. The results were compared with hand-crafted features such as texture, morphology, and gray level co-occurrence matrix $(GLCM)$ using non-deep learning classifiers such as support vector machine $(SVM)$ Gaussian Kernel, k-nearest neighbor-Cosine $(KNN-Cosine)$ , kernel naive Bayes, decision tree $(DT)$ and RUSBoost tree. This study reduces the features of carcinoma images, employed machine learning and deep learning approaches. For validation of training and testing data, a jack-knife ten-fold cross-validation method was used. The performance was measured using a confusion matrix such as sensitivity, specificity, positive predictive value $(PPV)$ , negative predictive value $(NPV)$ , accuracy $(AC)$ , Mathews Correlation Coefficient ( $MCC$ ), and area under the curve $(AUC)$ . The most remarkable performance was obtained using non-deep learning methods with $GLCM$ features using KNN-Cosine with sensitivity (98.00%), specificity (99.25%), PPV (98.99%), NPV (99.11%), accuracy (99.07%), and AUC (0.998). The LSTM deep learning method yields performance with sensitivity (98.33%), specificity (100%), PPV (100%), NPV (99.26%), accuracy (99.48%), MCC (0.9879) and AUC (0.9999), where using Deep learning method $ResNet-101$ , we obtained (100%) Accuracy and AUC (1) for Kernel Naive Bayes, SVM Gaussian and RUSBoost Tree. The results show that $ResNet-101$ deep learning outperformed than non-deep learning methods and $LSTM$ . Thus, the deep learning method $ResNet-101$ could be used as a better predictor for the detection of prostate cancer.

Prostate cancer is the second most commonly diagnosed malignancy in men, yet interpretation of multiparametric MRI (mpMRI) suffers from substantial inter-observer variability and depends on scarce expert radiologists. Expert readers follow a hierarchical, coarse-to-fine visual strategy: they first localize the prostate gland and then direct focused attention to tissue within it. Inspired by this biological selective-attention mechanism, we propose a segmentation-guided framework in which a modality-aware 3D Vision Transformer (ViT) first delineates the prostate gland from T2-weighted, ADC, and DWI volumes, and the resulting anatomical mask then constrains downstream cancer detection. Global self-attention models long-range anatomical context, analogous to the integration of peripheral and foveal information in human vision. Detection was evaluated through two pathways: gray-level co-occurrence matrix (GLCM) texture features with classical classifiers, and a ViT classifier. On 200 patients from the public PROSTATEx cohort, the proposed segmentation model achieved a Dice similarity coefficient of [0.905] versus [0.831] for an optimized U-Net. The segmentation-guided ViT achieved an AUC of [0.987] and an accuracy of [0.998] for clinically significant cancer detection. Ablation experiments confirmed that anatomical guidance improved both pathways. These results show that emulating expert visual search strategies yields accurate and interpretable computer-aided prostate cancer diagnosis.