Agreement between three state-of-the-art deep learning bone age estimation models and chronological age in a…

AI / informaticsPediatric radiology· Jun 2026

Agreement between three state-of-the-art deep learning bone age estimation models and chronological age in a large contemporary pediatric cohort.

Elanchezhian Somasundaram, Rama S Ayyala, William Tepe … Jonathan R Dillman

Pediatric RadiologyAI / informaticsDiagnosisQ1IF 2.2

Best deep learning model still overestimated chronological bone age by 3.20 months

Agreement between three state-of-the-art deep learning bone age estimation models and chronological age in a large contemporary pediatric cohort.

Elanchezhian Somasundaram … Jonathan R Dillman

Pediatric radiology · Jun 2026 · Q1 in Radiology

Background

Radiographic bone age estimation is routinely performed in children to evaluate short stature, early or late puberty, and endocrine disorders, as well as for surgical planning.

Purpose

To evaluate agreement between three state-of-the-art deep learning bone age estimation models and chronological age in a large contemporary pediatric cohort.

Methods

We retrospectively identified a large contemporary cohort of children (n=7,189; 3,669 females and 3,520 males) aged 24-216 months (mean 143.6±46.2 months) with consecutive radiologically normal hand (including the wrist) radiographs that were evaluated for trauma between November 1, 2010, and October 31, 2020.

n = 7,189

Results

AI still read children's bones as older than their true age

CCHMC

3.2months

MedImageInsight

4.6months

Stanford

7.03months

More results

All three models systematically overestimated chronological age, although the magnitude of bias varied by model and sex.

Mean differences compared with chronological age were statistically significant for all models (P<0.0001).

Evidence of proportional bias was observed in most models and subgroups.

More results

For all three models, the difference between predicted bone age and chronological age was greater for Black compared to White and Hispanic compared to non-Hispanic children.

“

Conclusion

Greulich and Pyle Atlas-based bone age models systematically overestimate chronological age in a large contemporary cohort of children undergoing hand radiography for trauma, with biases related to age, sex, race, and ethnicity.

Agreement between three state-of-the-art deep learning bone age estimation models and chronological age in a large contemporary pediatric cohort.

Agreement between three state-of-the-art deep learning bone age estimation models and chronological age in a large contemporary pediatric cohort.

Related papers