AI revolutionises bone age assessment in paediatric orthopaedic surgery
A new study reveals that AI offers promising advancements to enhance the accuracy, efficiency and consistency in bone age evaluation.
Bone age assessment is crucial in paediatric orthopaedic surgery, assisting in the diagnosis, treatment and management of growth-related disorders.
Traditional assessment methods, such as the Greulich-Pyle and Tanner-Whitehouse techniques, rely heavily on manual interpretation of hand and wrist radiographs, making the process time-consuming and susceptible to inter-observer variability.
A new study has explored the integration of AI for paediatric bone age prediction, utilising the Radiological Society of North America (RSNA) 2017 Paediatric Bone Age Challenge dataset.
A deep learning model, leveraging the ResNet-50 architecture (developed by Microsoft Research, Redmond, Washington, USA), was trained on 12,611 hand and wrist radiographs, validated on 1,425 images, and tested on 200 images.
The model’s predictive accuracy was measured using metrics such as root mean square error (RMSE), mean absolute error (MAE) and the coefficient of determination (R²).
The AI model demonstrated high accuracy, achieving an RMSE of 11.07 months, an MAE of 8.54 months, and an R² of 0.929.
The Pearson correlation coefficient (0.963) and Spearman's rank correlation (0.955) also reinforced the model’s reliability in predicting skeletal maturity.
Compared to traditional methods, which have shown a variability range of six to 18 months, the AI model exhibited a notable reduction in inter-observer differences, significantly enhancing diagnostic consistency.
Integrating AI into paediatric bone age assessment offers a standardised, efficient and highly accurate alternative to conventional manual approaches.
By minimising human subjectivity and reducing assessment time, AI significantly enhances clinical decision-making, enabling optimised treatment timing for growth modulation, limb lengthening, and scoliosis correction.
Generating rapid and precise bone age predictions can substantially improve patient outcomes by ensuring that surgical interventions align with critical growth phases.
However, broader validation across diverse patient populations remains essential to establish the reliability of AI applications in clinical practice.
AI-powered bone age assessment represents a paradigm shift in paediatric orthopaedic surgery, promising transformative benefits by reducing variability, enhancing diagnostic accuracy, and streamlining treatment planning.
As this technology continues to evolve, its integration into routine clinical practice can potentially improve accessibility, accelerate diagnosis and optimise surgical outcomes.
