Deep learning models for bone fracture detection from current trends and challenges to future directions

Abstract Bone fractures pose a significant global health concern. Accurate and prompt diagnosis of these injuries is necessary to prevent serious complications. Recent advancements in deep learning enable automated fracture detection, which can assist radiologists. This review provides an overview of deep learning applications in this domain, focusing on its importance, current technologies, challenges, and future directions. We review representative models and approaches from studies published between 2020 and 2025, providing detailed information on each model, including its architecture, technical innovations, and performance across various anatomical regions and imaging modalities. Among the reviewed studies, a predominance of radiographs (X-rays) and a focus on hip, wrist, and spinal fractures were observed. Key challenges, including limited data availability, model generalizability, interpretability, and clinical integration, are discussed. Emerging directions such as self-supervised learning, foundation models, generative data augmentation, multimodal fusion, federated learning, and explainable artificial intelligence are also highlighted. Overall, this review provides an overview of current approaches, emerging trends, and practical considerations in AI-assisted fracture detection.