In this study, the performance of deep learning methodologies in classification, particularly convolutional neural networks (CNN) and their various forms, is evaluated to aid in the identification of liver cancer using the different medical images taken. The dataset processed in this study consists of 2300 CT and MRI scans for identifying the liver cancer stages. The data are preprocessed, and after feature extraction, thorough model training is done. The VGG 19 model gives the best classification accuracy score at 96.75%, followed by the VGG 16 model at 93.4%, CNN at 91.2%, and ResNet 50 at 88.98%. This itself proves that deep learning models can learn and classify abnormal pathological symptoms and manifestations with respect to liver cancer. These models offer valuable tools for early detection, classification, and monitoring. The inference is that with deep learning architectures, the classifications are very accurate and highly relevant for the recognition of liver cancer and its pathological representations. Consequently, it can be inferred from this study that VGG 19 is the most accomplished model in classification of liver cancer with the highest rate of accuracy. This study brings out lots of significance for computer vision applications as well as for machine learning, and subsequent progressions in these areas will bring forth new models and methods. Systems built on these novel architectures will significantly enhance classification performance by innovative models and optimized methods with data augmentation techniques that improve efficiencies in a broad arena of disciplines.