Deep Neural Networks for Assessing the Legal Age from Panoramic Dental X-ray Images.

Forensic anthropologists and odontologists frequently need to establish the age of a living or deceased individual. In particular, a relevant problem consists of determining the legal age for persons not possessing identity documents or for whom it is not possible to have trustworthy information on their date of birth. An accurate technique to perform this measurement involves the analysis of panoramic dental X-ray images (orthopantomography), conducted by experienced forensic scientists. Although studies in the literature on automatic methods for age estimation based on dental images present promising results, none of them specifically focus on establishing the legal age. Instead, we propose a method based on deep learning for assessing the legal age. Specifically, we propose a method that mimics the analysis performed by human experts by examining the third molars of the inferior arch with Deep Neural Networks (DNNs). Our method can assess the legal age from a single or multiple molar images by using a score-level fusion strategy. To evaluate the performance of our method, we collected a dataset composed of 440 samples from individuals ranging from 15 to 22 years old. The obtained results proved the applicability of our method as a decision support tool for forensic scientists.