Deep neural networks in the estimation of chronological age using the development of third molars in panoramic radiographs

Abstract

This project aims to develop, evaluate and make available a platform based on artificial intelligence, focused on machine learning methods such as deep neural networks (deep learning) for the estimation of the mineralization stage of third molars and chronological age of patients using panoramic radiographs. This approach will be essential to allow speed, practicality, robustness and accuracy in this analysis compared to the traditional methods, which involve manual operations of specialists in the area. To build this platform, 12,000 panoramic radiographs of patients of both sexes, aged 6 to 22 years, will be selected from the image dataset of the radiologic clinics of UFPE and UNICAMP, following the inclusion and exclusion criteria suggested in the literature. The images will be labeled by four specialists from the two universities, regarding the dental development stages of third molar teeth, according to the classification used in forensic dentistry. The data will be analyzed following the experimental methodology of the area and will serve as input for training computer vision models for the automation of this activity. Among the results, it is expected that this is the first version of an intelligent platform that will aim to expand specialized services in dentistry using artificial intelligence, in order to subsidize new clinical studies in dentistry based on machine learning. This technology will be essential to improve people's quality of life, since it will be able to provide greater optimization of time, accuracy of diagnosis, and the decision for an appropriate treatment. (AU)