The Rise of Artificial Intelligence in Age Estimation

Resumo

Introduction: The combination of multiple information of interest and fully automated assessment are optimization concepts that are increasingly applied to age estimation. Machine learning can be a valuable tool to obtain additional features that were not initially detected by imaging. Neural networks that analyze large amounts of skeletal and dental data are designed to surpass manual performance, reduce time spent, improve reliability, and provide more accurate predictive models.

Objective: Clarify the growing application of artificial intelligence to age estimation, namely by interpreting the potential for using multiple features of interest through deep learning, and its applicability and accuracy across adult age groups.

Materials and Methods: A bibliographic search was performed on the application of artificial intelligence in age estimation, using the PubMed, Web of Science, and Scopus databases. The adopted keywords were: forensic age assessment; multifactorial information; algorithms. The inclusion criteria were articles written in English, studies on humans, and participants aged 14 or older.

Results and Discussion: Aging is a multilayered and nonlinear phenomenon. Models generated by convolutional neural networks can use large data sets, establishing a complex and comprehensive correlation between multiple attributes and age (Guo et al., 2021). The backpropagation algorithm minimizes estimation error (Farhadian et al., 2019). Stern et al. conducted multifactorial studies, using magnetic resonance imaging, to clarify the relative contribution of various structures (third molars, left wrist, and clavicles) in the assessment of legal majority. They demonstrated that the contribution of the wrist decreased while the contribution of the clavicles increased with age. The combination of the three structures is especially helpful around the age of 18 (Stern et al., 2019). The matching of relevant information from the teeth and surrounding bones in panoramic X-rays, by convolutional neural networks, contains useful indicators for age prediction (Milosevic et al., 2022). Bone density decreases consistently with age. For elderly age estimation, modeling bone mineral density data from different bone structures has been a trend. Machine learning models have demonstrated the ability to assess cadavers´ age, even in putrefaction cases (Curate et al., 2022).

Conclusions: Artificial intelligence, particularly through neural networks, has been gaining prominence for age estimation. However, improvements should be made to include studies in different populations, broader age ranges, and with multiple alternative skeletal regions. The goal is to provide an easy-to-use online platform.