“Explainable Bone Tumor Diagnosis Using Deep CNNs and Language Models”

Abstract: Bone cancer is a critical medical condition that requires early and accurate diagnosis to improve patient outcomes. Traditional diagnostic approaches based on manual interpretation of X-ray images are often limited by inter- observer variability and the scarcity of expert radiologists, particularly in resource-constrained settings. To address these challenges, this paper proposes an explainable deep learning framework for automated bone tumor classification using X-ray images.

The proposed system leverages a pre-trained Convolutional Neural Network (CNN) to accurately detect and classify bone tumors into multiple categories. To enhance model interpretability, Gradient-weighted Class Activation Mapping (Grad-CAM) is employed to generate visual explanations by highlighting the most relevant regions in the input images that contribute to the model’s predictions. Furthermore, an integrated Large Language Model (LLM) module generates human-readable diagnostic explanations, summarizing tumor characteristics, predicted severity, and potential clinical insights.

The framework is evaluated on benchmark medical imaging datasets, demonstrating superior performance compared to conventional machine learning and deep learning baselines. Experimental results show significant improvements in classification accuracy, precision, recall, and F1-score, while also reducing training time. The combination of high predictive performance and enhanced interpretability makes the proposed system a reliable decision-support tool for healthcare professionals.

This work contributes toward the development of transparent, efficient, and accessible AI-driven diagnostic systems, with strong potential for real-world deployment in clinical and low-resource environments.

Keywords: Bone cancer; Deep learning; Convolutional Neural Network (CNN); X-ray classification; Grad-CAM; Large Language Model (LLM).