Early and accurate cancer diagnosis remains one of the major challenges in modern healthcare, where diagnostic errors, limited interpretability of deep learning models, and the lack of integrated clinical reasoning can adversely affect patient outcomes. Although Convolutional Neural Networks (CNNs) have achieved remarkable performance in medical image analysis, they operate as black-box models and cannot effectively utilize clinical text or structured medical knowledge. This paper proposes a novel Self-Adaptive Hybrid CNN-LLM Framework that integrates visual, textual, and symbolic reasoning for cancer detection and clinical decision support. The proposed framework combines three complementary components: a deep CNN backbone for extracting spatial features from MRI, CT, and X-ray images; a biomedical Large Language Model (LLM) for interpreting clinical reports and generating diagnostic reasoning; and a Medical Knowledge Graph (MKG) based on the Unified Medical Language System (UMLS) for semantic validation of predictions. An attention-based transformer fusion module integrates information from the three modalities into a unified representation, while a Grad-CAM visualization module and an LLM-based explanation engine provide interpretable Explainable Artificial Intelligence (XAI) reports for clinicians. In addition, a self-adaptive reinforcement learning mechanism continuously refines the framework using clinician feedback without requiring complete model retraining. Experimental evaluation on three publicly available cancer datasets demonstrates that the proposed framework achieves 97.3% accuracy, 96.8% precision, 96.4% recall, and an AUC of 0.983, outperforming CNN-only and CNN+LLM baseline models by 4.2–7.9% across key performance metrics while providing substantially improved interpretability and clinical transparency. These results demonstrate that the proposed framework represents a reliable and trustworthy clinical decision support system that effectively bridges the gap between data-driven artificial intelligence and clinically meaningful medical reasoning.