Simulation of facial growth based on longitudinal data: Age progression and age regression between 7 and 17 years of age using 3D surface data
Abstract
Modelling of the development of facial morphology during childhood and adolescence is highly useful in forensic and biomedical practice. However, most studies in this area fail to capture the essence of the face as a three-dimensional structure.
The main aims of our present study were (1) to construct ageing trajectories for the female and male face between 7 and 17 years of age and (2) to propose a three-dimensional age progression (age-regression) system focused on real growth-related facial changes. Our approach was based on an assessment of a total of 522 three-dimensional (3D) facial scans of Czech children (39 boys, 48 girls) that were longitudinally studied between the ages of 7 to 12 and 12 to 17 years.
Facial surface scans were obtained using a Vectra-3D scanner and evaluated using geometric morphometric methods (CPD-DCA, PCA, Hotelling's T-2 tests). We observed very similar growth rates between 7 and 10 years in both sexes, followed by an increase in growth velocity in both sexes, with maxima between 11 and 12 years in girls and 11 to 13 years in boys, which are connected with the different timing of the onset of puberty.
Based on these partly different ageing trajectories for girls and boys, we simulated the effects of age progression (age regression) on facial scans. In girls, the mean error was 1.81 mm at 12 years and 1.7 mm at 17 years.
In boys, the prediction system was slightly less successful: 2.0 mm at 12 years and 1.94 mm at 17 years. The areas with the greatest deviations between predicted and real facial morphology were not important for facial recognition.
Changes of body mass index percentiles in children throughout the observation period had no significant influence on the accuracy of the age progression models for both sexes.