Artificial intelligence (AI) has become an indispensable tool in orthopedic surgery. It provides new methods to increase surgical precision, improve patient safety, and support personalized treatment plans. This review presents a comprehensive analysis of AI-assisted orthopedic surgery across three core domains. Based on 89 recent studies, this review organizes findings around a perception-decision-execution framework. It groups diverse AI applications into certain categories while highlighting the mutuality across domains. Perception systems have progressed from basic CNN-based segmentation models to advanced transformer architectures. They support multi-modal data fusion and enable uncertainty quantification. Decision systems have moved far beyond rigid rule-based methods and evolve into data-driven models that support surgical planning, accurate risk prediction and continuous outcome optimization. And execution systems have advanced from passive navigation tools to active robotic assistance systems with real-time adaptive capabilities. Beyond mapping technological advances, this review also identifies pivotal challenges that hinder clinical translation and concludes with a clear roadmap for future research, which marks closed-loop surgical assistance systems as the next key development direction. Building on these findings, this review illuminates the potential of AI-assisted orthopedic surgery and guides future research toward innovations that can be translated into clinical practice.
Keywords: artificial intelligence; deep learning; machine learning; orthopedic surgery; surgical navigation.