A Stacked-CNN deep learning strategy was introduced for recognizing prostate cancer grades using histopathology images from the Prostate Cancer Grade Assessment (PANDA) challenge. The dataset includes 11,616 training and 1593 validation images labeled using the Gleason grading scheme. Methods of preprocessing such as scaling, normalization, and data augmentation were employed to enhance model robustness. The CNN architecture consists of three convolutional layers, followed by MaxPooling layers and two fully connected dense layers. This Stacked-CNN model was trained over 30 epochs utilizing the Adam optimizer and categorical cross-entropy loss function. The performance was tracked using a Quadratic Weighted Kappa (QWK) evaluation callback, and class weights were incorporated to address class imbalance. Our model achieved a training loss of 0.2761 and a notable validation QWK score, demonstrating its ability to accurately predict prostate cancer grades. Comparisons with previous research show that our methodology outperforms existing approaches, achieving an area under the curve (AUC) of 0.90.