Prostate cancer is one of the most prevalent malignancies among men worldwide, necessitating precise diagnostic tools for effective clinical management. In this work, an application of the CNN based on the VGG16 architecture for automated prostate cancer diagnosis and Gleason grading from biopsy WSIs is described. For this study the dataset that was utilized is from the PANDA challenge which includes high-resolution WSIs representing various Gleason scores and their corresponding ISUP grades. Preprocessing involved narrowing the region of interest (ROI) using masking, followed by dividing each WSI into 16 tiles of size 224 × 224 × 3. These tiles passed to the VGG16 model which has been adapted for the given tile dimensions of input and the layers of output were modified to include softmax activation function to predict the ISUP grade 0 to 5. Adam optimizer was used in training for the 15 epochs with a learning rate of 0.001. The training dataset comprises 80% and the testing datasets comprises for 20% of the total dataset or in numerical values 8492 and 2124 WSI images respectively where 20% of the training data is kept aside for validation. On testing, the performance metrics used included accuracy, specificity, precision, recall and F1-score resulting in an overall accuracy of 89.8% across the test set. The confusion matrix demonstrated reliable classification across all ISUP grades, with particularly high recall and specificity for clinically significant cancer grades. This work affirms the possibility of employing deep learning systems into the automated diagnosis of prostate cancer and grading to increase efficiency and minification of interobserver variability in the related clinical workflow.