Comprehensive Summary
This article provides a systematic review of recent literature on the use of artificial intelligence (AI) algorithms to detect bone injuries in paediatric patients. The review included peer-reviewed studies published between 2011 and 2024 containing the keywords ‘child’, ‘AI’, ‘fracture’, and ‘imaging’. Twenty-six articles were included, seventeen of which were published within the past two years. Reported model accuracy ranged from 85-100%, and two papers found a significant improvement in human diagnostic accuracy when assisted by AI. Recent research shows growing interest in this field, with many models exhibiting strong predictive performance and a stronger shift towards evaluating human accuracy when aided by AI, a more clinically relevant metric.
Outcomes and Implications
Childhood fractures are a common injury and often present subtly, leading to high rates of missed diagnosis. Insufficient treatment may result in growth-plate disturbances and long-term disability, highlighting the need for accurate diagnostic tools. The authors noted the potential application of AI in identifying cases of suspected child abuse, especially as there is a decreasing number of radiologists available for expert testimony. Ultimately, clinical implementation and cost-effectiveness remain the key barriers in clinical adoption of AI in detecting paediatric fractures.