Olecranon bone age assessment in puberty using a lateral elbow radiograph and a deep-learning model

Area of Science:

• Pediatric Radiology
• Artificial Intelligence in Medicine
• Skeletal Development

Background:

• Accurate pubertal bone age (BA) assessment is crucial for evaluating growth and endocrine disorders.
• Conventional methods for BA evaluation, particularly those using hand radiographs, have limitations in precision and applicability.
• Elbow radiography offers a potential alternative for BA assessment, but standardized methods are lacking.

Purpose of the Study:

• To develop a precise and practical elbow BA classification system utilizing the olecranon ossification center.
• To create a deep-learning artificial intelligence (AI) model for automated olecranon BA assessment.
• To evaluate the performance and reliability of the novel olecranon BA method and AI model compared to established techniques.

Main Methods:

• Retrospective analysis of 3508 lateral elbow radiographs from children under 18 years.
• Development of a novel olecranon BA classification based on morphological changes during puberty.
• Comparison of olecranon BA with other elbow and hand BA methods using intraclass correlation coefficients (ICCs).
• Development and external validation of a deep-learning AI model (EfficientDet-b4) for olecranon BA determination.

Main Results:

• The olecranon BA classification demonstrated 100% applicability and excellent interobserver agreement (ICC 0.993).
• High reliability was observed between olecranon BA and established methods like Sauvegrain, Dimeglio, and Korean Standard (KS) hand BA.
• The AI model achieved high accuracy (0.96) and specificity (0.98) for olecranon BA, with external validation showing 0.86 accuracy and 0.91 specificity.

Conclusions:

• The olecranon BA evaluation, using a single lateral elbow radiograph, offers superior applicability and interobserver agreement.
• This method exhibits excellent reliability when compared to existing elbow and hand-based BA assessment techniques.
• The developed AI model shows high performance, making olecranon BA a practical and accurate tool for pubertal BA assessment.