Radiology Spotters Collection

Pediatric Bone Age Calculator [Includes AI-based Automated Tools]

Bone age is skeletal maturity read from a radiograph of the left hand and wrist and reported as the chronological age whose reference standard the film matches. It is matched against the Greulich-Pyle atlas or scored by the Tanner-Whitehouse method. The number means nothing on its own: it only becomes useful once you compare it with the normal spread for that child’s age and sex.

How big a difference from chronological age is normal?

There is no validated cut-off that makes a bone age abnormal. Two separate spreads govern how much a difference is worth, the biological spread between healthy children and the spread between radiologists reading the same film, and both are close to a year.

What was measuredResultSource
Greulich-Pyle age against chronological age, 257 healthy UK girls aged 10 to 17Mean absolute difference 10.0 months; individual differences from 40 months below to 31 months aboveSan-Millรกn 2025 [1]
Radiologists rating the same film, 1,285 radiographs, 22 readersWithin-image SD 8.7 months under 12 years, 4.2 months at 12 and overLong 2026 [2]
One reader against a three-reader consensusMean absolute error 6.5 months; 3.4% of readings out by over 1.8 yearsLong 2026 [2]
Why a difference of about a year between bone age and chronological age may mean nothing at all.

The widely repeated “plus or minus 2 years” approximates 2 standard deviations; it is not a published threshold. Treat a difference under roughly a year as uninformative by itself, and read anything larger against growth velocity, pubertal stage and the clinical question.

The bone age X-ray

  • Projection. A single posteroanterior (dorsopalmar) view of the left hand and wrist, hand flat on the detector, fingers extended and slightly separated, centred on the third metacarpal head.
  • Why the left hand. The left hand and wrist is the standard film the atlases were built on [10], by convention because most children are right dominant and the left side is less likely to be deformed or injured.
  • Coverage. Distal radius and ulna, all carpals, metacarpals and phalanges. The epiphyses and growth plates are what you are reading, so a film clipping the distal radius is not reportable.
  • Dose. Negligible, and worth saying so to parents. Monte Carlo modelling put the effective dose of a hand radiograph at around 0.01 microsieverts in a 3 to 6 year old [11].

Greulich-Pyle versus Tanner-Whitehouse

Greulich-Pyle matches the film as a whole to the nearest sex-specific atlas plate. Tanner-Whitehouse 3 scores bones individually and sums them, with separate RUS (radius, ulna, short bones) and Carpal versions [3].

MethodWhat it readsWhere it earns its place
Greulich-PyleWhole hand and wrist matched to the closest atlas standardThe default, and the more reproducible of the two in a large series: intra-rater and inter-rater correlation 0.993 and 0.992, against 0.985 and 0.984 for TW3-RUS [3]
Tanner-Whitehouse 3Grades 20 regions of interest, summed to a maturity scoreWhen an auditable numeric score is wanted. More accurate than Greulich-Pyle for age estimation in a meta-analysis [4]
Automated software and AIRates the film without a readerRemoves reader drift, see below
Bone age methods, and what each one is actually better at.

Which is “more accurate” is not settled: the meta-analysis favours TW3 [4], the largest reproducibility comparison favours Greulich-Pyle [3]. Whether the reference population fits your patient matters more.

PopulationFinding against the standardSource
Hispanic and Asian children over 12About 1 year advanced against the standardThodberg 2010 [5]
Zimbabwean peripubertal boysGreulich-Pyle read 0.76 years below chronological age, TW3 0.43 below; no difference in girlsKowo-Nyakoko 2023 [6]
Asian, Arab and African adolescents near adulthoodGreulich-Pyle overestimated; TW3 more accurate in African youthsMartรญn Pรฉrez 2023 [7]
The atlas reference images are of Caucasian children, and the fit to other populations is uneven.

Advanced bone age and delayed bone age

Tall children tend to an advanced bone age and short children to a delayed one [10], so the finding often just restates the growth chart. The value lies in the discrepancy that does not fit the child’s height and pubertal stage.

Advanced bone ageDelayed bone age
Central precocious pubertyConstitutional delay of growth and puberty
Congenital adrenal hyperplasia and other causes of androgen or oestrogen excessGrowth hormone deficiency
HyperthyroidismHypothyroidism
ObesityHypogonadism and Turner syndrome
Familial tall stature, and overgrowth syndromes such as Sotos and Beckwith-WiedemannChronic illness, coeliac disease, malnutrition, chronic renal failure, glucocorticoid excess
The clinical categories a referrer works through. The radiograph says advanced or delayed; it does not name the cause.

Can bone age predict adult height?

Only loosely, and the error is larger than families expect. The Bayley-Pinneau method converts a Greulich-Pyle bone age into the percentage of adult height already attained. Compared head to head, every prediction model overestimated adult height, by 0.7 to 6.7 cm in boys and 0.9 to 3.7 cm in girls, and up to 43% of Bayley-Pinneau predictions differed by more than 5 cm when two readers rated the same film [9]. Khamis-Roche needs no radiograph at all, using parental height and the child’s own measurements. Give a range, never a single number.

Automated and AI bone age tools

Automated bone age is now at least as good as a single radiologist. The 2017 RSNA Pediatric Bone Age Machine Learning Challenge ran 14,236 radiographs and the best model reached a mean absolute difference of 4.2 months [8]. In the multi-reader benchmark above, BoneXpert’s mean absolute error against a three-reader consensus was 4.8 months, beating the 6.5 months of a single human reader, and it was out by over 1.8 years in 0.7% of cases against 3.4% for single readings [2]. That study was funded by BoneXpert’s manufacturer and its chief executive is a co-author, so read the margin over humans with that in mind.

Stanford hand bone age AI tool with a left hand radiograph uploaded for bone age estimation
Stanford hand bone age AI: upload the left hand radiograph.
Stanford bone age AI returning an estimated skeletal age for the uploaded hand radiograph
The tool returns an estimated bone age, which you still check against the atlas before reporting.

Bone age atlas images online, and a video walkthrough

Reporting it

  • Name the method used: a Greulich-Pyle age and a Tanner-Whitehouse age are not interchangeable.
  • Give chronological age, bone age and the difference, then say whether it is within the normal spread for age and sex.
  • Do not call a bone age abnormal on a difference of about a year: that is inside both biological and reader variation [1,2].
  • Say so when carpal and phalangeal maturity disagree.
  • Leave the cause to the clinician, and flag the reference population if the child is unlikely to match the atlas [7].

Frequently asked questions

What is a normal bone age for a child’s chronological age?

There is no published cut-off. In 257 healthy UK girls aged 10 to 17, Greulich-Pyle skeletal age differed from chronological age by 10.0 months on average, and by anything from 40 months below to 31 months above [1]. A difference of about a year is expected in a normal child, not a finding.

How is bone age measured on a hand X-ray?

From one posteroanterior radiograph of the left hand and wrist. Greulich-Pyle matches the whole film to the closest sex-specific atlas standard. Tanner-Whitehouse 3 grades 20 regions of interest and sums them into a maturity score [3]. The distal radius, ulna and all carpals must be on the image.

What does an advanced or delayed bone age mean?

Advanced points to early or excess sex steroid exposure (central precocious puberty, congenital adrenal hyperplasia), hyperthyroidism, obesity or an overgrowth syndrome. Delayed points to constitutional delay, growth hormone deficiency, hypothyroidism, hypogonadism, glucocorticoid excess or chronic illness. Since tall children tend to be advanced and short children delayed [10], the informative result is the one that contradicts the growth chart.

Can a bone age X-ray tell how much taller a child will grow?

Only as a range. Compared head to head, every height prediction model overestimated adult height, by 0.7 to 6.7 cm in boys and 0.9 to 3.7 cm in girls, and up to 43% of Bayley-Pinneau predictions differed by more than 5 cm between two readers of the same film [9].

Can bone age prove someone’s real age?

No. A meta-analysis of the three main left hand and wrist methods found that age estimation in living individuals cannot be considered accurate from these alone [4], and using skeletal maturity to establish chronological age is explicitly not recommended [10]. Bone age says how mature the skeleton is, not how old the person is.

References

  1. San-Millรกn M, Holmes T, De Luca S, et al. Assessing skeletal maturity in a UK modern female population. Forensic Sci Med Pathol. 2025;21(4):1676-1688. PMID: 40767910
  2. Long J, Larson DB, Thodberg HH, et al. Quantifying human variability in pediatric bone age assessment: a multi-institutional multi-reader study and benchmark for AI evaluation. Pediatr Radiol. 2026 Sep 1 (online ahead of print). PMID: 42678523
  3. Yuh YS, Chou TY, Tung TH. Bone age assessment: large-scale comparison of Greulich-Pyle method and Tanner-Whitehouse 3 method for Taiwanese children. J Chin Med Assoc. 2023;86(2):246-253. PMID: 36652571
  4. Serinelli S, Panetta V, Pasqualetti P, et al. Accuracy of three age determination X-ray methods on the left hand-wrist: a systematic review and meta-analysis. Leg Med (Tokyo). 2011;13(3):120-133. PMID: 21377398
  5. Thodberg HH, Sรคvendahl L. Validation and reference values of automated bone age determination for four ethnicities. Acad Radiol. 2010;17(11):1425-1432. PMID: 20691616
  6. Kowo-Nyakoko F, Gregson CL, Madanhire T, et al. Evaluation of two methods of bone age assessment in peripubertal children in Zimbabwe. Bone. 2023;170:116725. PMID: 36871897
  7. Martรญn Pรฉrez SE, Martรญn Pรฉrez IM, Vega Gonzรกlez JM, et al. Precision and accuracy of radiological bone age assessment in children among different ethnic groups: a systematic review. Diagnostics (Basel). 2023;13(19):3124. PMID: 37835867
  8. Halabi SS, Prevedello LM, Kalpathy-Cramer J, et al. The RSNA Pediatric Bone Age Machine Learning Challenge. Radiology. 2019;290(2):498-503. PMID: 30480490
  9. Chรกvez-Vรกzquez AG, Klรผnder-Klรผnder M, Garibay-Nieto NG, et al. Evaluation of height prediction models: from traditional methods to artificial intelligence. Pediatr Res. 2024;95(1):308-315. PMID: 37735232
  10. Martin DD, Wit JM, Hochberg Z, et al. The use of bone age in clinical practice, part 2. Horm Res Paediatr. 2011;76(1):10-16. PMID: 21691055
  11. Beaulieu D, Lee C, Kim DS, et al. Radiation dose estimation of pediatric upper extremity radiographs. Pediatr Radiol. 2025;55(8):1719-1727. PMID: 40526117

4 thoughts on “Pediatric Bone Age Calculator [Includes AI-based Automated Tools]”

  1. Ngov Songngin

    Corrected Version (Recommended):

    Dear Sir,
    Could you please grant us free full access to the AI-based bone age calculation tool?
    Thank you very much for your kind support for a resource-limited country.

    Best regards,
    Songngin

  2. maha boukheir

    xray result for a boy
    A.I. estimate: 16 years [ 0.756 ] yrs +/-12 mo, male
    how much he will be tall

    1. Dr. Amar Udare, MD

      Unfortunately we will not be able to help with that. Please talk to your physician.

Leave a Comment

Your email address will not be published. Required fields are marked *


Wish to be a BETTER Radiologist? Join 15000 Radiology Colleagues !

Enter your email address below to access HIGH YIELD radiology content, updates, and resources.

Email Newsletter Subscription Pop Up

No spam, only VALUE! Unsubscribe anytime with a single click.

Scroll to Top