Background and Objectives: Knee Osteoarthritis (KOA) is a leading cause of global disability, yet current management remains reactive due to the limitations of the Kellgren-Lawrence (KL) grading system. This review aims to synthesize technological advancements from 2018–2026 to provide a framework for AI-driven precision orthopedics. Methods: A comprehensive narrative review was conducted by searching major electronic databases for peer-reviewed research focusing on AI-driven KOA diagnostics and prognosis. Evidence was categorized into seven thematic domains, ranging from automated grading to omics-based molecular discovery. Findings: Technical analysis reveals that multi-stage pipelines utilizing YOLOv2 and Faster R-CNN isolate joint spaces with over 98% accuracy, reducing diagnostic errors by 5–7%. Natural Language Processing (NLP) models, specifically BiLSTM architectures, identify KOA risk 24 to 36 months before radiographic confirmation with an AUC of 0.911. Multi-modal frameworks like LBTRBC-M achieve a predictive AUC of 0.913 for structural progression and improve the prognostic accuracy of resident physicians from approximately 45% to over 66%. Furthermore, Generative AI (GPT-4) delivers personalized patient guidance 14 times faster than clinicians. In molecular research, AI has identified PDK1 as a critical regulatory gene for chondrocyte autophagy. Conclusions: AI is shifting KOA management from reactive diagnosis to proactive health forecasting. However, the wide-scale adoption of these technologies depends on robust socio-ethical governance, addressing algorithmic bias, and fostering trust through Explainable AI. Future frontiers include the development of "digital twins" and integrated omics-imaging models.
Tryczyński et al. (Wed,) studied this question.