This study introduces a deep learning system that automatically identifies and categorizes liver tumors in CT images using convolutional neural networks (CNNs). The approach includes several preprocessing steps such as picture scaling, histogram equalization, bilateral filtering, and K-means segmentation. The CNN is used for binary classification, distinguishing between benign and malignant tumors. When a malignancy is detected, a second CNN-based system categorizes the tumors into Early, Intermediate, and Metastatic stages. This multi-step approach not only automates tumor detection but also provides a finer-grained analysis of malignant cases, offering valuable insights into the progression of liver tumors. The methodology combines advanced image processing techniques with deep learning classification, showcasing a comprehensive framework for efficient and detailed liver tumor analysis in medical imaging. This method simplifies tumor diagnosis and provides important insights into the progression of liver tumors.