In recent years, deep learning has demonstrated outstanding performance in various artificial intelligence systems, particularly in object detection, with numerous practical applications. In the field of medical imaging, deep learning models have brought significant progress and can serve as a reliable second opinion for experts. However, the biggest challenge remains the lack of transparency, which makes it difficult to trust these models, especially in sensitive fields such as medicine. To address this problem, explainable artificial intelligence (XAI) was born to help better understand how models make decisions. In this study, we apply two popular XAI methods, LIME and Grad-CAM, to state-of-the-art object detection models, including YOLOv8 and Detectron2. Based on the explainability results, we propose a hybrid model to improve the accuracy. The method was evaluated on a bone tumor radiograph (BTXRD) dataset, demonstrating that combining the model with XAI analysis enhances both model interpretability and detection performance, increasing reliability in real-world medical scenarios.