Age assessment by using facial photo-anthropometry in a Brazilian population

This study aimed to test the discriminant power of the photoanthropometric analysis of the human face for distinguishing females and males aged below or above 14 and 18. The sample consisted of 1354 photographs taken in frontal view of the Brazilian females and males aged between 10 and 22 years. Using SAFF-2D® software (Department of Federal Police, Brasília, Brazil), fourteen examiners positioned 35 landmarks in each of the photographs. The landmark positioning led to the quantification of 110 indices and 51 iridian ratios. These variables were tested into logistic regression models designed to distinguish females and males older or not than 14 and 18 years. Receiver Operating Characteristics (ROC) curves were used to assess the distinctive power of the models and the inherent Area Under the Curve (AUC) founded inferences about accuracy. The model was able to distinguish females that were not older than 14 in 129 cases (64.5 %) and females that were older than 14 in 359 cases (75.42 %). The females that were not older than 18 years were distinguished in 250 cases (60.83 %), while the females older than 18 years were distinguished in 199 cases (74.53 %). Males that were not older than 14 were distinguished in 175 cases (84.95 %) and males that were older than 14 were distinguished in 381 cases (81.06 %). Males that were not older than 18 years were distinguished in 280 cases (68.97 %), while males older than 18 years were distinguished in 224 cases (83.58 %). The female model reached 76 % and 73 % of distinctive accuracy for the thresholds of 14 and 18 years, respectively. The male model reached 90 % and 83 % for the same thresholds, respectively. Facial photoanthropometry is a useful tool for age estimation in criminal cases that involve the legal age thresholds of 14 and 18 years.