Liver tumor segmentation is vital for early detection and treatment planning, but traditional methods lack precision and explainability. This research proposes a hybrid model combining Hybrid U-Net, SHAP, Grad-CAM, and Large Language Models (LLMs) to enhance segmentation accuracy and interpretability. The Hybrid U-Net ensures precise segmentation, while SHAP and Grad-CAM++ provide visual explanations of the model’s decisions. LLMs generate natural language insights, offering clinicians intuitive, human-readable explanations for the predictions. By emphasizing key regions of interest and delivering local explanations, the model bridges the gap between AI-driven analysis and clinical trust. Experimental results demonstrate high segmentation accuracy, computational efficiency, and robustness, making it suitable for real-world applications. This work advances interpretable AI in medical diagnostics, with potential extensions to multi-organ segmentation and real-time clinical validation.