Cancer remains one of the major global health challenges, with a significant impact on morbidity and mortality worldwide, making early detection crucial for improving survival rates through more timely interventions and treatments. However, the early diagnosis of liver cancer presents complexities and often relies on technologies and methods with inherent limitations. This study aims to develop an early detection model for liver cancer from abdominal Computed Tomography scan images, applying an experimental methodology based on the phases of the CRISP-DM process. Analysis of results demonstrates that the model achieved 95% accuracy in validation tests. In a real-world evaluation, it assigned a cancer probability ≥ 80% to 77.7% of cases, reflecting high confidence in detecting high-risk lesions and suggesting that this approach could accelerate oncological diagnosis, benefit more patients, and optimize clinical resources. Subsequent phases could involve expanding the dataset, exploring additional artificial intelligence algorithms, and implementing advanced segmentation techniques using masks and multi-label classification to optimize the precise delineation of tumor areas.