Development of a biometric method to estimate age on hand radiographs

- A biometric method for age estimation of minors from hand radiographs analyses.
- Virtual application of the osteometric board used in physical anthropology.
- Formulas developed for age groups identified by a Linear Discriminant Analysis.
- Age estimated at maximum ± 1 year (95% confidence interval).

Age estimation of living individuals aged less than 13, 18 or 21 years, which are some relevant legal ages in most European countries, is currently problematic in the forensic context. Thus, numerous methods are available for legal authorities, although their efficiency can be discussed. For those reasons, we aimed to propose a new method, based on the biometric analysis of hand bones. 451 hand radiographs of French individuals under the age of 21 were retrospectively analyzed. This total sample was divided into three subgroups bounded by the relevant legal ages previously mentioned: 0-13, 13-18 and 18-21 years. On these radiographs, we numerically applied the osteometric board method used in anthropology, by including each metacarpal and proximal phalange of the five hand rays in the smallest rectangle possible. In that we can access their length and width information thanks to a measurement protocol developed precisely for our treatment with the ORS Visual® software. Then, a statistical analysis was performed from these biometric data: a Linear Discriminant Analysis (LDA) evaluated the probability for an individual to belong to one of the age group (0-13, 13-18 or 18-21); and several multivariate regression models were tested for the establishment of age estimation formulas for each of these age groups. The mean Correlation Coefficient between chronological age and both lengths and widths of hand bones is equal to 0.90 for the total sample. Repeatability and reproducibility were assessed. The LDA could more easily predict the belonging to the 0-13 age group. Age can be estimated with a mean standard error which never exceeds 1 year for the 95% confidence interval. Finally, compared to the literature, we can conclude that estimating an age from the biometric information of metacarpals and proximal phalanges is promising.