Current applications and challenges of artificial intelligence applied to diagnostics in pediatric musculoskeletal imaging.

The use of artificial intelligence (AI) in pediatric musculoskeletal imaging has undergone significant expansion over the past few years. Until recently, the use of AI was limited to evaluating bone age and opportunistically assessing the bone health index. Currently, validated AI software for commercial use includes the detection of appendicular fractures, automated measurement of scoliosis, assessment of lower limb length discrepancy, and assessment of developing hip dysplasia. For other applications, further work is needed. Diagnostic accuracy for detecting rib and vertebral fractures in children using AI is currently not satisfactory; however, future research using enhanced deep learning is projected to address these limitations. The implementation of other applications of diagnostic AI in pediatric musculoskeletal imaging for non-accidental trauma, bone dysplasia, and tumor assessment is hindered by the lack of large pediatric datasets, which would require multicenter collaborations. This paper aims to succinctly outline the present clinical applications of AI in the pediatric musculoskeletal field, while elucidating existing possibilities, limitations, and future needs and prospects.