A Comprehensive Review of Artificial Intelligence Methods in Bone Age Assessment.

Area of Science:

• Radiology and Artificial Intelligence in Paediatric Healthcare

Background:

• Bone age assessment is crucial for monitoring child growth and development, traditionally relying on radiological methods like bone age atlases.
• Current methods face limitations including restricted case diversity, small sample sizes, and lack of multi-expert validation, hindering generalizability.
• Existing models often overlook patient comorbidities and detailed health histories, impacting clinical relevance.

Purpose of the Study:

• To review and analyze emerging artificial intelligence (AI) based approaches for paediatric bone age assessment.
• To identify challenges and opportunities for enhancing AI models in clinical practice for more accurate skeletal maturity evaluation.

Main Methods:

• Systematic review of AI-based methods for paediatric bone age assessment, primarily utilizing hand and wrist radiographs.
• Analysis of AI models, including RCNN, evaluating their potential using metrics like mean absolute error.
• Examination of radiology reports integrated with AI assessments.

Main Results:

• AI models, particularly RCNN, demonstrate significant potential for accurate bone age determination.
• The study identified key limitations in current AI models, such as data diversity and validation.
• Opportunities exist for integrating comprehensive health histories and expert validation into future AI models.

Conclusions:

• AI holds substantial promise for transforming paediatric bone age assessment, offering greater accuracy and clinical utility.
• Future AI models must incorporate diverse datasets, thorough health histories, and expert validation for improved performance.
• Addressing identified gaps will enhance the clinical application of AI in evaluating skeletal maturity.