Automated bone age assessment in a German pediatric cohort: agreement between an artificial intelligence software and the manual Greulich and Pyle method - Report - MDSpire
Evaluation of Automated Bone Age Determination vs Greulich and Pyle Method
Overview
This study validated the performance of the CE-certified AI software IB Lab PANDA for automated bone age (BA) assessment in a German pediatric population. The AI showed strong agreement with expert radiologists using the Greulich and Pyle (G&P) manual method across a broad pediatric age range. The findings support the AI tool as a reliable and efficient alternative for BA determination.
Background
Bone age assessment is crucial in pediatric endocrinology, orthodontics, and orthopedics to evaluate skeletal maturity relative to chronological age. The Greulich and Pyle method is the most commonly used manual approach but suffers from intra- and inter-observer variability. Advances in digital imaging and deep learning have enabled automated BA estimation, potentially improving consistency and efficiency. The IB Lab PANDA software is a novel CE-certified AI tool designed to automate BA assessment based on the G&P atlas, but prior to this study, it lacked independent clinical validation.
Data Highlights
Metric
Value
Sample size
306 patients (153 females, 153 males)
Age range (females)
36 to 192 months
Age range (males)
36 to 204 months
Mean Absolute Difference (MAD)
Reported in study (exact values not provided in excerpt)
Root Mean Square Error (RMSE)
Reported in study (exact values not provided in excerpt)
Intraclass Correlation Coefficient (ICC)
High agreement between AI and experts
Pearson Correlation
Strong correlation between AI and expert readers
Key Findings
The AI software IB Lab PANDA demonstrated strong agreement with expert radiologists using the G&P method for bone age assessment.
The study included 306 pediatric patients aged 1 to 18 years, with stratified sampling ensuring balanced sex and age distribution.
Automated BA estimates correlated highly with the averaged expert readings, with low mean absolute differences and root mean square errors.
Inter- and intra-observer variability inherent to manual methods was addressed by averaging three expert readers to establish ground truth.
The AI tool provided rapid, objective BA assessments consistent with clinical standards within its intended age range.
Clinical Implications
The validated AI software offers a reliable, efficient alternative to manual bone age determination, potentially reducing observer variability and workload in clinical practice. Its use can streamline skeletal maturity assessments in pediatric endocrinology and related fields, supporting timely and accurate clinical decision-making. Integration of such AI tools may enhance consistency across institutions and practitioners.
Conclusion
This independent validation confirms that the IB Lab PANDA AI software provides bone age assessments comparable to expert radiologists using the Greulich and Pyle method. The tool represents a promising advancement for automated, standardized pediatric skeletal maturity evaluation.
References
Greulich and Pyle -- Radiographic Atlas of Skeletal Development of the Hand and Wrist
Martin et al 2011 -- Preferred Bone Age Assessment Methods
