Artificial intelligence (AI) plays a transformative role in the medical field, offering advanced tools for early diagnosis and accurate disease classification. In particular, deep learning has proven highly effective for automated feature extraction from medical images, capturing intricate patterns such as edges, textures, and shapes crucial for disease identification. Liver tumors, caused by abnormal cell growth, severely impact vital liver functions including metabolism and detoxification, making them a global health concern due to high morbidity and mortality rates. Early and accurate detection of liver tumors is essential for timely treatment and improved patient outcomes. In this study, we propose a hybrid deep learning and machine learning pipeline that integrates Convolutional Neural Networks (CNN) with XGBoost for liver tumor classification. The CNN serves as a robust feature extractor, while XGBoost leverages the extracted high-dimensional features for accurate classification. Our model achieves impressive performance with 97.71% accuracy, 96.93% precision, 96.87% recall, and 96.89% F1-score. ROC-AUC analysis confirms strong discriminative capability, highlighting the model’s suitability for real-world medical diagnostic applications.