Artificial intelligence in knee osteoarthritis imaging and total knee arthroplasty: advances, challenges, and segmentation methods - A review.

Authors

Abstract

Knee osteoarthritis (KOA) is one of the most prevalent degenerative joint diseases and a significant cause of disability. Total Knee Arthroplasty (TKA) is the standard treatment for advanced cases, but its success depends on accurate imaging, precise preoperative planning, and intraoperative execution. Imaging modalities such as radiographs, computed tomography (CT), magnetic resonance imaging (MRI), and ultrasound are widely used in TKA workflows. Still, they are often limited by anatomical variability, image artifacts, and poor soft tissue contrast. This review surveys automated segmentation and analysis approaches for knee imaging, covering classical methods including region-, boundary-, atlas-, and model-based techniques, as well as recent artificial intelligence (AI)-driven approaches, particularly deep learning models. Their capabilities, limitations, and clinical relevance in TKA planning and evaluation are discussed. Classical segmentation methods provide foundational tools but often exhibit limited generalization and robustness in heterogeneous clinical datasets. AI-based methods, particularly deep learning, enable automated feature extraction, improved segmentation accuracy, and enhanced classification and outcome prediction. These methods address many limitations of traditional approaches, although their dependence on large annotated datasets and variability in imaging protocols remains a significant challenge. This review presents a comprehensive overview of knee anatomy, imaging modalities, and segmentation techniques relevant to TKA. Comparing classical and AI-based approaches highlights their strengths, limitations, and persistent challenges, while also identifying opportunities for integrating AI into clinical workflows to achieve more precise, reliable, and patient-specific knee arthroplasties.

Tags

Topics