Radiographic skeletal maturity assessment of the left hand and wrist — Greulich-Pyle atlas matching, TW3 RUS scoring, and Bayley-Pinneau adult height prediction
Bone age assessment begins with a single, deceptively simple radiograph: a posteroanterior view of the left hand and wrist. This one image captures roughly 30 discrete ossification centers — carpal bones, metacarpal and phalangeal epiphyses, and the distal radius and ulna — each maturing on its own predictable schedule. Decades of standardization make this the most reproducible windows into a child's biological, as opposed to calendar, age.
The convention of using the left hand traces back to the Brush Foundation growth study of the 1930s-40s, on which the Greulich-Pyle atlas was built — investigators simply chose the left hand as their standard and every subsequent atlas and scoring system followed suit for comparability.
Multiple validation studies have since confirmed there is no clinically meaningful left-right asymmetry in skeletal maturation, even in children who are strongly hand-dominant. Using one side consistently:
• Minimizes radiation exposure to a single limb • Ensures every film is directly comparable to the reference atlas, which was built exclusively from left-hand films • Removes hand-dominance as a confounding variable across serial studies used to track growth over time
Technique: • Patient's palm is placed flat against the image receptor, fingers slightly abducted (not touching) to prevent overlapping bone margins • The third metacarpal is aligned with the central X-ray beam, which is angled perpendicular to the plate • Low-dose pediatric exposure factors are used (low kVp, fast digital detector or screen-film combination) — bone age is a low-yield, high-frequency exam and dose minimization (ALARA) is a priority • A single PA exposure suffices; no oblique or lateral views are needed
What is visible on the resulting film: • Epiphyses (secondary ossification centers) at the distal radius and ulna, each metacarpal head/base, and each phalangeal base • A radiolucent growth plate (physis) separating each epiphysis from its metaphysis — the structure whose progressive narrowing and eventual disappearance is the core signal read by every bone age method • Carpal bones, which appear sequentially between roughly 3 months and 7-8 years of age (capitate and hamate first, pisiform last) and whose count is itself a coarse maturity indicator in very young children
For over 65 years the Greulich-Pyle method has remained the most widely used bone age technique worldwide because of its speed: a trained reader compares the entire hand, as a gestalt pattern, against a sequence of reference radiographs and picks the single best match. The matched plate's labeled age becomes the assigned bone age — a single visual judgment standing in for a formal measurement.
The Greulich-Pyle atlas is a direct descendant of the Brush Foundation study, a longitudinal project that serially X-rayed a cohort of predominantly white, middle-class children in Cleveland, Ohio between 1931 and 1942. Researchers selected the radiograph that best represented "typical" skeletal maturity at each age and sex, assembling them into a sequential atlas of standard plates spanning birth to skeletal maturity.
This origin is also the method's most cited limitation: the reference population is demographically narrow and nearly a century old, and secular trends (children today tend to mature somewhat earlier, and population body composition has shifted substantially since the 1940s) mean the original standards may not perfectly represent a contemporary, diverse patient population.
The reader scans the whole hand and locates the standard plate that most closely resembles the overall pattern of ossification — not any single bone in isolation. When the whole-hand match is ambiguous (asymmetric maturation across bones, which is common), readers fall back to a bone-specific comparison, checking individual centers against the atlas and averaging or selecting the most representative structures (commonly the carpals plus a subset of phalangeal/metacarpal epiphyses).
This whole-pattern approach is fast — an experienced pediatric radiologist can assign a bone age in roughly a minute — which is precisely why it remains the default clinical method, despite being less granular than numeric scoring systems.
• Subjectivity: the "best match" judgment varies between readers, and even the same reader re-reading the same film weeks later (intra-observer variability) • Population specificity: the reference standards were derived from a narrow mid-20th-century American cohort and may generalize imperfectly to other populations • Coarse resolution: matching to a discrete set of ~30 reference plates per sex cannot express bone age more precisely than the spacing between adjacent standards • No formal uncertainty estimate: the method returns a single point value with no confidence interval, unlike numeric scoring systems
Despite its limitations, Greulich-Pyle remains the most used method globally because it requires no specialized software and takes under two minutes per case — a critical advantage in high-volume pediatric radiology practice.
Where Greulich-Pyle asks "which reference plate does this hand resemble," the Tanner-Whitehouse method asks a more rigorous question of every bone individually: which of eight or nine defined developmental stages does this specific bone match? Summing calibrated points across a fixed panel of bones yields a numeric maturity score that converts to bone age through sex-specific tables — a method built for reproducibility rather than speed.
TW3 scores the Radius, Ulna, and Short bones (RUS) — thirteen structures in total: the distal radius, distal ulna, and the metacarpals and phalanges (proximal, middle, distal) of the first, third, and fifth digits. A parallel Carpal score covering seven carpal bones is available but is now used less often, as carpal maturation plateaus and loses discriminatory value after mid-childhood.
Each of the thirteen RUS bones is independently staged from a defined checklist of radiographic criteria — specific features of epiphyseal shape, width relative to the metaphysis, and the degree of epiphyseal-metaphyseal fusion — assigned to discrete stages labeled A through H or I. Each stage has an explicit, illustrated definition, unlike the gestalt "does this look similar" judgment of atlas matching.
Every stage letter for every bone carries a pre-calibrated point value (these weights were themselves derived statistically, not assigned arbitrarily, so that a given stage transition contributes an amount of "maturity points" proportional to how much developmental information it actually conveys). The point values for all thirteen RUS bones are summed into a single RUS maturity score, typically expressed on a roughly 0-1000 scale that saturates as skeletal maturity is reached.
This maturity score is then converted to a numeric bone age using sex-specific reference tables built from the TW3 standardization sample — a population more contemporary than the original Greulich-Pyle cohort. Because scoring is bone-by-bone and criterion-based rather than whole-pattern matching, two independent readers scoring the same film converge far more closely than they would under Greulich-Pyle.
TW3's reproducibility advantage comes at the cost of time — bone-by-bone scoring of thirteen structures takes considerably longer by hand than a single atlas match, which has historically limited its routine clinical adoption relative to Greulich-Pyle.
This tradeoff has been substantially erased by automated software (most notably BoneXpert, CE-marked and used widely in Europe), which performs TW3-equivalent bone-by-bone analysis via automated image segmentation and shape-model matching in seconds, with reported reproducibility approaching the theoretical floor of the method itself. Automation has renewed interest in TW3-style scoring for both routine endocrinology follow-up and clinical trial endpoints, where its superior reproducibility over successive scans is especially valuable.
Because TW3 scores each bone against explicit criteria rather than matching a whole-hand gestalt, it is far better suited to automation — which is precisely why the modern software tools used to monitor children on growth hormone therapy are built on TW3 principles rather than Greulich-Pyle.
A bone age alone is just a number; its clinical value comes entirely from comparing it to chronological age. A discrepancy of roughly two years or more (outside the normal ±1-2 SD band expected from ordinary biological variation) is a signal, not a diagnosis — it points the clinician toward a specific, tractable differential depending on whether the skeleton is running ahead of or behind the calendar.
Sex steroids — principally estrogen, even in boys via peripheral aromatization of testosterone — are the dominant driver of epiphyseal maturation and eventual growth plate fusion. Any process that raises sex-steroid exposure ahead of schedule accelerates bone age:
• Precocious puberty (central or peripheral): early gonadarche drives early epiphyseal advancement • Congenital adrenal hyperplasia (CAH): chronic excess adrenal androgen exposure can advance bone age by 3-5 years if untreated, ultimately compromising final height despite tall stature in childhood • Obesity: increased peripheral aromatization of androgens to estrogen, plus hyperinsulinemia, modestly advances bone age in many children • Hyperthyroidism: excess thyroid hormone directly accelerates the ossification process • McCune-Albright syndrome and other causes of autonomous sex-steroid production
The clinical stakes are real: advanced bone age means less remaining growth potential than chronological age alone would suggest — the growth plates are closer to fusion than the calendar implies.
The delayed side of the differential is broader and includes both the most common overall explanation and the most concerning ones:
• Constitutional delay of growth and puberty (CDGP): by far the most common cause — a normal variant, often familial, in which both bone age and puberty are delayed in step; these children eventually catch up and reach a normal adult height • Growth hormone deficiency: reduced GH/IGF-1 signaling slows both linear growth and skeletal maturation • Hypothyroidism: among the most dramatic causes of bone age delay — untreated congenital or acquired hypothyroidism can markedly suppress ossification • Chronic systemic illness: celiac disease, inflammatory bowel disease, chronic kidney disease, and other conditions that divert energy away from growth • Malnutrition and psychosocial/emotional deprivation ("psychosocial dwarfism") • Turner syndrome and other conditions affecting growth plate biology directly
Distinguishing benign constitutional delay from a pathologic cause is precisely the diagnostic question bone age is used to help answer.
A bone age result rarely stands alone — it is interpreted alongside height velocity, pubertal staging, and targeted labs. In practice, the discrepancy directs the workup:
• Isolated delay with a strong family history of "late bloomers" and otherwise normal growth velocity → favors constitutional delay; reassurance and observation • Delay with poor growth velocity or other red flags → prompts GH stimulation testing, thyroid function tests, celiac serologies, or karyotype (in girls with unexplained short stature) • Advancement with early pubertal signs → prompts LH/FSH and estradiol/testosterone testing to distinguish central from peripheral precocious puberty, and consideration of GnRH agonist ("puberty blocker") therapy to preserve height potential • Advancement without puberty → prompts evaluation for CAH, exogenous hormone exposure, or thyroid disease
The bone age result feeds directly into the next stage of assessment: quantifying exactly how much growth potential remains.
The clinical payoff of bone age assessment is a number families and clinicians can act on: a predicted adult height. The Bayley-Pinneau method, published in 1952 and still in routine use, combines a child's current height with their bone age to estimate what percentage of adult stature has already been attained — then simply divides backward to project the final result.
Bayley and Pinneau built sex-specific tables of "percent of mature height attained" indexed by bone age, further stratified into average, accelerated, and retarded maturational tempo categories (comparing bone age to chronological age identifies which tempo category applies). Percent-attained rises steeply through early-to-mid childhood and then flattens as the epiphyses approach fusion, asymptoting toward 100% once the growth plates have closed.
The prediction itself is simple arithmetic once the table lookup is done:
Predicted adult height = current height ÷ (percent of adult height attained at this bone age)
A child whose bone age indicates 80% of adult height has been attained, standing at 128 cm today, projects to roughly 128 / 0.80 ≈ 160 cm as an adult — with real uncertainty attached, since the tables reflect population averages rather than that individual's exact trajectory.
Predicted adult height is compared against two other benchmarks:
• Mid-parental (genetic target) height, estimated as (father's height + mother's height + 13 cm)/2 for boys or (father's height + mother's height − 13 cm)/2 for girls, ± roughly 8.5 cm • Population height percentile norms for age and sex
A predicted adult height well below the genetic target, combined with delayed bone age and a demonstrated GH deficiency, supports GH therapy candidacy. Conversely, in precocious puberty with advanced bone age eroding predicted height, a GnRH agonist ("puberty blocker") can pause further epiphyseal advancement, preserving remaining growth potential until a more typical pubertal timeline resumes. Serial bone age studies, typically every 6-12 months, then track whether therapy is achieving its intended effect on the growth trajectory.
Bayley-Pinneau predictions are most reliable in children maturing at an average tempo; the further bone age diverges from chronological age (either direction), the wider the true error margin around the prediction becomes — a caveat clinicians weigh heavily before basing major treatment decisions on a single estimate.
Bayley-Pinneau's accuracy degrades in children with marked BA/CA discrepancy, in very short or very tall children at the tails of the reference distributions, and in any condition altering the normal relationship between bone maturation and linear growth (e.g., skeletal dysplasias). It also does not incorporate individual growth velocity trends directly.
Newer approaches — TW3-based height prediction tables, and increasingly machine-learning models trained on large longitudinal growth datasets that incorporate bone age, growth velocity, mid-parental height, and pubertal stage simultaneously — are gradually improving on the 70-year-old Bayley-Pinneau tables, though the original method remains the most widely taught and used baseline in pediatric endocrinology practice today.