Software development using artificial intelligence as a detection aid for caries and gingivitis, for the organization of demand for dental services.

• Grant number: 25/10293-2
• Support Opportunities: Regular Research Grants
• Start date: January 01, 2026
• End date: December 31, 2027
• Field of knowledge: Health Sciences - Dentistry - Social and Preventive Dentistry
• Principal Investigator: Antonio Carlos Pereira
• Grantee: Antonio Carlos Pereira
• Host Institution: Faculdade de Odontologia de Piracicaba (FOP). Universidade Estadual de Campinas (UNICAMP). Piracicaba , SP, Brazil
• City of the host institution: Piracicaba
• Associated researchers: Danilo Rodrigues Pereira ; Karine Laura Cortellazzi Mendes ; Vanessa Gallego Arias Pecorari

Abstract

This project aims to develop algorithms structured in convolutional neural networks and ViTs for the creation of software using artificial intelligence (AI) to assist in the diagnosis of oral health problems through the analysis of dental data obtained from image databases and subsequent use in health services and large-scale epidemiological surveys. For the development of the five research projects: In Project 1 - A systematic literature review will be conducted to evaluate diagnostic tests in oral health. This review will identify and categorize the different deep learning methods employed, the neural network architectures adopted, the hyperparameters used, as well as the accuracies obtained. Project 2 - This aims to develop an application compatible with the main smartphone operating systems to facilitate the capture of standardized images (extra and intraoral), using the Flutter framework and the Flutter1000 programming language (Google). The application will have as main features: authentication, user registration, standardized photo capture, and submission of radiographic images for validation. In Project 3, the neural network will be created based on images from a diagnostic test study. Six hundred patients over six years of age will be selected, from UFPB, UFMS, and UNICAMP, after approval by the Ethics Committee. Clinical and radiographic examinations will be carried out, in addition to photographic captures obtained through the application developed in Project 2. Experienced, previously calibrated specialist dentists will analyze and classify the images for the presence or absence of caries and gingivitis. Radiographic images and clinical examination will be used to validate the classification of the photographic captures. Next, image preprocessing will be performed with randomization, splitting the database into training (80%) and validation (20%), using a convolutional neural network (CNN) and Vision Transformers (ViTs). The training data will be divided into batches and epochs with a learning rate of 0.01. The algorithm results will be calculated using metrics such as accuracy, sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and area under the ROC curve (AUC). In Project 4, the objective will be to create software integrated with the application for the detection and diagnosis of caries and gingivitis. The project focuses on the development of an online platform that will use standard images already collected by the mobile application in Project 3. The system will be based on technologies from the Flutter framework and the Flutter1000 programming language to ensure compatibility across different devices. In addition, the platform will strictly follow the guidelines of the General Data Protection Law (LGPD) to ensure user privacy and data security. Finally, in Project 5, strategies for implementing the APP will be developed. The implementation strategies will consider the RE-AIM model, which will allow the evaluation of essential dimensions such as reach, effectiveness, adoption, fidelity of use, and sustainability of the technology in the context of public services. In addition, ethical and technical guidelines from the WHO/ITU AI4H standard will be adopted to ensure that the use of artificial intelligence meets the principles of safety, equity, and transparency. Oral health surveillance actions supported by AI will follow the principles of the CDC's Guidelines for Public Health Surveillance. These frameworks will strengthen large-scale adoption in the SUS. This architecture will include the development of both front-end and back-end, the integration of a database system with the implementation of security protocols for data protection. (AU)