Development of artificial intelligence algorithms for the diagnosis and assessment of dental caries in children in clinical and epidemiological settings

• Grant number: 24/09855-3
• Support Opportunities: Regular Research Grants
• Start date: November 01, 2024
• End date: October 31, 2026
• Field of knowledge: Health Sciences - Dentistry - Pediatric Dentistry
• Principal Investigator: Fausto Medeiros Mendes
• Grantee: Fausto Medeiros Mendes
• Host Institution: Faculdade de Odontologia (FO). Universidade de São Paulo (USP). São Paulo , SP, Brazil
• Associated researchers: Daniela Prócida Raggio ; Flávio Fernando Demarco ; Luiz Alexandre Chisini ; Mariana Minatel Braga Fraga ; Maximiliano Sergio Cenci ; Tamara Kerber Tedesco
• Associated scholarship(s): 25/09998-1 - Comprehensive pediatric dental treatment of children participating in the study aimed at developing automated tools for detecting caries lesions, BP.TT; 25/09999-8 - Clinical, radiographic and intraoral scanning evaluation in progress the development of automated tools for detecting caries lesions in children, BP.TT

Abstract

This research project aims to develop some artificial intelligence algorithms for diagnosis and assessment of dental caries in children, considering the epidemiological and clinical settings. The protocol is composed by two main arms: 1) one research developed in the epidemiological context with populational basis, and 2) another conducted in the context of clinical practice, and it has 5 specific aims: (i) to develop deep learning algorithms for detecting caries lesions in intraoral scans of children in the context of epidemiological studies; (ii) to develop machine learning prediction models for dental caries in children at the same setting; (iii) to develop deep learning algorithms for detecting caries in children in a clinical setting using intraoral scans, and (iv) bitewings, and (v) to validate the algorithms developed in a prospective study in a clinical environment and simulating the scenario of epidemiological studies. The studies will occur simultaneously and coordinated by researchers from the Schools of Dentistry at the Federal University of Pelotas and the University of São Paulo. The epidemiological study will be conducted on the 2015 Pelotas/RS birth cohort. The 4,275 children included in the study will be contacted in order to take part in a new survey at the age of 11, in the mixed dentition, which will be submitted to a full clinical examination. For dental caries, all tooth surfaces of primary and permanent teeth present will be examined, following the combined ICDAS criteria (healthy surface, with initial lesion, moderate lesion or severe lesion). Next, all participants will have their dental arches imaged using an intraoral. The 3D images obtained by the scanner in different formats will be used to develop algorithms for caries detection in epidemiological studies, using deep learning approaches in order to optimize the validity of the models (objective i). Machine learning models using the main algorithms approaches will be created to identify predictors of the occurrence of dental caries in children in this same context (objective ii). Detecting dental caries in the context of clinical practice involves not only assessing the presence of caries lesions but also staging and classifying the activity of these lesions. The clinical study will, therefore, be performed with children seeking dental treatment at FOUSP who are in the primary or mixed dentition stage. The children will be assessed for the presence of caries using the same simplified ICDAS criteria and intraoral scans will be obtained. Deep learning models aimed at detecting caries at all stages of severity will be developed (objective iii). In parallel, bitewings of children in the mixed dentition stage will be classified and annotated in relation to dental caries, and another deep learning-based model will be developed (objective iv). The accuracy of these models will first be tested on the database on samples randomly selected as test dataset, which were not part of the development of the models. Once the models have been developed, all the algorithms will be assessed for their validity in a prospective study with consecutive inclusion of children seeking dental treatment (objective v). (AU)