Ultrasound-based area ossification ratios for pediatric radiographic bone age: Development and internal validation of a radiation-free nomogram.

Purpose

Radiographic bone age (BA) assessment involves ionizing radiation, limiting its utility in frequent longitudinal follow-up. This study developed and internally validated a radiation-free nomogram based on multi-site two-dimensional ultrasound area ossification ratios (ORs) to assess Zhonghua-05 (RUS-CHN) radiographic bone age and predicted adult height (PAH), evaluating its incremental value over chronological age.

Methods

In this single-center prospective study (August 2024 ∼ December 2025), 1,053 children aged 3-15 years were randomized into training and validation sets (7:3). Ultrasound area ORs of the left distal radius, ulna, and femur were calculated. Using radiographic RUS-CHN BA as the reference standard, predictor variables were selected via LASSO regression to construct the nomogram. Model performance was evaluated across the cohort and within developmental discordance subgroups (advanced/delayed maturation).

Results

Ultrasound OR measurements showed excellent reproducibility (ICC > 0.88). In the validation set, the nomogram predicted RUS-CHN BA with a mean absolute error (MAE) of 2.94 months (RMSE = 3.67 months, R2 = 0.989). Crucially, in children with advanced or delayed maturation, the nomogram reduced BA prediction MAE by over 60 % compared to a baseline clinical feature model. The secondary model for PAH achieved an MAE of 0.63 cm.

Conclusions

Multi-site ultrasound area ORs provide a highly reproducible, radiation-free index of skeletal maturity. As an internally validated, standard-specific research tool, the nomogram estimated RUS-CHN radiographic BA and corrected age-anchoring bias in children with discordant maturation. Because its numeric calibration is specific to the RUS-CHN standard, external multicenter and longitudinal validation is required before clinical use. ABBREVIATIONS BA, Bone age; BMI, Body mass index; CI, Confidence interval; GP, Greulich Pyle; ICC, Intraclass correlation coefficient; LASSO, Least absolute shrinkage and selection operator; LoA, Limits of agreement; MAE, Mean absolute error; MPH, Mid‑parental height; OR, Area ossification ratio; PAH, Predicted adult height; QC, Quality control; RMSE, Root mean square error; RUS‑CHN, Radius, ulna, and short bones RUS skeletal‑maturity score, Chinese standard; SOP, Standard operating procedure; TRIPOD, Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis.