Misir Abdulhamit
Bahcesehir University School of Medicine, Department of Orthopedics and Traumatology, Turkiye.
Injury. 2025 Aug;56(8):112570. doi: 10.1016/j.injury.2025.112570. Epub 2025 Jul 1.
Artificial intelligence (AI) has emerged as a transformative technology in healthcare, with significant applications in orthopedic trauma. This comprehensive review analyzes 217 studies published between 2015 and 2025 to evaluate the current state, applications, and future directions of AI in orthopedic trauma. The field has experienced exponential growth, with 52.5 % of all studies published in 2024 alone. Deep learning approaches (43.3 %) and traditional machine learning methods (39.2 %) dominated the research landscape. Fracture detection (24.4 %) and classification (12.0 %) were the most common applications, followed by prediction (21.2 %) and segmentation (8.3 %). Hip/femur (19.4 %), spine (18.9 %), and wrist fractures (12.0 %) represented the most frequently studied anatomical sites. AI systems frequently matched or exceeded specialist performance in detection and classification tasks, with sensitivities and specificities above 90 % commonly reported. Predictive models for complications and mortality consistently outperformed traditional scoring systems, with improvements in AUC typically between 0.10-0.15. However, only 14.5 % of studies underwent external validation, and just 3.2 % reported prospective clinical validation. Despite remarkable progress in developing accurate AI systems for orthopedic trauma, significant challenges remain in clinical integration, data standardization, and validation across diverse populations. Future development should focus on multimodal approaches integrating diverse data sources, transparent algorithms providing rationales for predictions, and rigorous clinical validation. Point-of-care applications and integration with emerging technologies offer promising directions for clinical impact. As these challenges are addressed, AI has the potential to significantly enhance orthopedic trauma care by improving diagnostic accuracy, optimizing treatment selection, and identifying high-risk patients for targeted interventions.
人工智能（AI）已成为医疗保健领域的一项变革性技术，在骨科创伤方面有重要应用。这篇综述分析了2015年至2025年间发表的217项研究，以评估人工智能在骨科创伤中的现状、应用及未来方向。该领域呈指数级增长，仅2024年就发表了所有研究的52.5%。深度学习方法（43.3%）和传统机器学习方法（39.2%）主导了研究格局。骨折检测（24.4%）和分类（12.0%）是最常见的应用，其次是预测（21.2%）和分割（8.3%）。髋部/股骨（19.4%）、脊柱（18.9%）和腕部骨折（12.0%）是研究最频繁的解剖部位。人工智能系统在检测和分类任务中常常达到或超过专家表现，常见的灵敏度和特异度均高于90%。并发症和死亡率的预测模型始终优于传统评分系统，AUC通常提高0.10 - 0.15。然而，只有14.5%的研究进行了外部验证，仅有3.2%报告了前瞻性临床验证。尽管在开发用于骨科创伤的精确人工智能系统方面取得了显著进展，但在临床整合、数据标准化以及针对不同人群的验证方面仍存在重大挑战。未来的发展应侧重于整合多种数据源的多模态方法、提供预测依据的透明算法以及严格的临床验证。即时护理应用以及与新兴技术的整合为临床影响提供了有前景的方向。随着这些挑战得到解决，人工智能有潜力通过提高诊断准确性、优化治疗选择以及识别高风险患者进行针对性干预，显著提升骨科创伤护理水平。
