• Deep learning models significantly improve diagnostic efficiency and accuracy in bone imaging, including fracture, bone tumor, osteoporosis, osteoarthritis, synovitis, and spinal conditions.
• Convolutional neural networks and derived models excel in classification, segmentation, and detection tasks, enabling automated feature extraction from skeletal images.
• Current challenges include limited model generalization across different medical centers and imaging devices, and the need for large annotated datasets.
• Future research should focus on enhancing model robustness, generalization, and interpretability to facilitate clinical adoption.