Validation and reference values of automated bone age determination for four ethnicities

Area of Science:

• Radiology
• Pediatric Endocrinology
• Medical Imaging Analysis

Background:

• Bone age (BA) assessment is crucial in pediatric endocrinology.
• Manual BA ratings exhibit significant rater variability.
• Automated methods offer potential to standardize BA determination.

Purpose of the Study:

• To validate the BoneXpert automated method for bone age determination.
• To assess the accuracy of the automated method across diverse ethnicities in American children.
• To compare the automated method's performance against manual ratings.

Main Methods:

• Utilized a publicly available database of 1100 hand X-rays from healthy American children (0-18.99 years).
• Included diverse ethnicities: Caucasian, African American, Hispanic, and Asian.
• Compared automated BoneXpert ratings against two independent manual BA ratings.

Main Results:

• The automated method achieved an RMS deviation of 0.61 years, comparable to manual rater variability (0.63 years).
• Analysis included images for boys (BA >2.5 years) and girls (BA >2 years).
• Hispanic and Asian children over 12 showed a ~1-year advancement relative to the Greulich and Pyle standard.

Conclusions:

• The automated BoneXpert method effectively determines bone age across multiple ethnicities.
• This automated approach can eliminate rater variability in bone age assessment.
• The method is applicable for boys aged 2.5-17 years and girls aged 2-15 years.