Enhancing Pediatric Bone Age Assessment Using Artificial Intelligence: Implications for Orthopedic Surgery.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Pediatric Orthopedics

Background:

• Bone age assessment is crucial for pediatric orthopedic surgery, guiding treatment for growth disorders.
• Traditional methods (Greulich-Pyle, Tanner-Whitehouse) are manual, time-consuming, and prone to variability.
• Artificial intelligence (AI) presents a potential solution for enhanced accuracy and standardization.

Purpose of the Study:

• To evaluate the efficacy of an AI deep learning model for pediatric bone age prediction.
• To compare AI performance against traditional methods in terms of accuracy and reliability.

Main Methods:

• A ResNet-50 deep learning model was developed and trained on 12,611 hand and wrist radiographs from the RSNA 2017 dataset.
• Model performance was assessed using Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and coefficient of determination (R²).
• Validation and testing were performed on 1,425 and 200 images, respectively.

Main Results:

• The AI model achieved an RMSE of 11.07 months, MAE of 8.54 months, and R² of 0.929.
• High correlation coefficients (Pearson: 0.963, Spearman: 0.955) confirmed predictive robustness.
• The AI model demonstrated reduced inter-operator variability compared to traditional methods (errors 6-18 months).

Conclusions:

• AI provides a standardized, rapid, and precise alternative for bone age assessment.
• Improved accuracy and efficiency have significant implications for pediatric orthopedic surgery timing and access.
• Further validation is necessary for broad clinical applicability across diverse populations.