Orthopaedic fracture diagnosis and surgical planning are being revolutionised by artificial intelligence (AI) and machine learning (ML). These technologies enable the automated detection and classification of fractures and provide surgeons with AI-assisted pre-operative decision-making support at multiple body sites. The objective of this review is to synthesise current evidence (from 2016 to 2026) related to the application of AI/ML technologies in orthopaedic fracture diagnosis and surgical planning, to summarise the primary clinical findings from these studies, and to describe the challenges associated with translating these technological advancements into clinical practice. A MEDLINE/PubMed literature search was conducted utilising a combination of Medical Subject Headings (MeSH) terms for AI, deep learning, fracture detection, and orthopaedic surgical planning. A total of 24 peer-reviewed publications that met the inclusion criteria and were published between 2016 and 2026 were identified and then synthesised through a narrative approach. Convolutional neural network (CNN)-based and Vision Transformer (ViT)-based deep learning models demonstrated area under receiver operating curves (AUROC) values ranging from 0.85 to 0.98 for detecting fractures in the hips, femurs, spines, scaphoids, tibias, pelvises, and ribs on radiographs and computed tomography (CT). The AI-assisted pre-operative planning tools have also been validated for use with pelvic and proximal humeral fractures. Additionally, ML has predicted the outcomes of surgery with AUROC values ranging from 0.78 to 0.87. There are several key limitations to the adoption of these technologies in clinical practice, including the majority of studies having been designed as retrospective analyses, the relative lack of diversity within their datasets, and the evolution of regulatory frameworks governing the application of AI/ML in medical imaging. Overall, AI technology has been shown to be highly accurate in both diagnosing fractures and assisting surgeons with planning for orthopaedic surgery. Before widespread clinical use can begin, prospective multi-centre validation and regulatory clarity will need to be achieved.
Badejo et al. (Tue,) studied this question.