What is bone age?
Bone age (skeletal age) is the biological maturity level the skeleton has reached. After birth, bones mature in a specific and predictable order: cartilage ossifies, ossification centers appear and grow, and growth plates eventually fuse. That sequence can run faster or slower from person to person, and bone age measures that pace.
Chronological age is calculated from the date of birth and advances at the same rate for everyone. Bone age, by contrast, can run ahead of or behind chronological age depending on hormonal status, nutrition, genetics and chronic illness. The gap between the two is what carries clinical information.
Why a hand and wrist radiograph?
Bone age assessment is almost always performed on a single posteroanterior (PA) radiograph of the left hand and wrist. The reason is practical: the hand and wrist contain many small bones and ossification centers together, so dozens of maturity indicators can be assessed in one image.
The hand radiograph also carries a very low radiation dose and is easy to acquire. Using the left hand is a historical standard — the reference atlases were built on left-hand images.
When is bone age requested?
Bone age is never a diagnosis on its own; it supplements the clinical picture. It is most often requested for:
- Short stature or falling noticeably behind peers
- Faster-than-expected growth or signs of precocious puberty
- Delayed onset of puberty
- Evaluation and treatment follow-up of endocrine disorders such as growth hormone or thyroid problems
- Situations requiring adult height prediction
- Age estimation in undocumented individuals (only ever as one part of a forensic assessment)
How is it assessed? Greulich-Pyle and Tanner-Whitehouse
In the classical approach a radiologist compares the radiograph against a reference atlas. The two most widespread methods are:
- Greulich-Pyle (GP): the radiograph is compared as a whole against standard images organised by age and sex; the age of the closest-matching standard is reported. It is fast, which is why it dominates day-to-day practice.
- Tanner-Whitehouse (TW2/TW3): specific bones are individually scored by maturity stage and the scores are summed and converted to a bone age. It is more detailed and more reproducible, but takes considerably longer.
Where does AI fit in?
Atlas matching is a visual, experience-driven task, and two radiologists can read the same radiograph differently (inter-observer variability). Deep-learning models are used to reduce that variability and shorten reporting time: the model produces an estimate in seconds from patterns learned across thousands of radiographs.
The crucial caveat: AI does not replace the radiologist. A model cannot recognise poor image quality, an atypical skeletal disorder or incorrect positioning as well as you can — and it will still output a number when it is uncertain. Interpreting the result in clinical context remains the physician's job.