skip to main content
Idioma:

Teeth numbering and identification of decayed teeth in panoramic radiographs through convolutional neural networks

Zancan, Breno Augusto Guerra

Biblioteca Digital de Teses e Dissertações da USP; Universidade de São Paulo; Faculdade de Filosofia, Ciências e Letras de Ribeirão Preto 2023-11-27

Acesso online. A biblioteca também possui exemplares impressos.

  • Título:
    Teeth numbering and identification of decayed teeth in panoramic radiographs through convolutional neural networks
  • Autor: Zancan, Breno Augusto Guerra
  • Orientador: Macedo, Alessandra Alaniz
  • Assuntos: Aprendizado Profundo; Cárie; Sistemas De Visão Computacional; Diagnóstico Oral; Numeração De Dentes; Radiografia Panorâmica; Teeth Numbering; Panoramic Radiography; Oral Diagnosis; Deep Learning; Computer Vision Systems; Tooth Decay
  • Notas: Dissertação (Mestrado)
  • Descrição: Dental radiographs play a crucial role in dentistry, allowing early detection of dental problems and a wide range of dental and oral health problems, that may not be visible during a clinical examination alone. Panoramic dental radiographs belong to the category of extraoral imaging, capturing a comprehensive view of all teeth and specific adjacent anatomical structures. Despite the invaluable diagnostic support offered by radiographs, challenges may arise in the accurate interpretation of patients\' oral conditions. Factors such as lack of experience, subjective judgments, stress, and fatigue among dental professionals can impact their assessments. In this context, Convolutional Neural Networks hold substantial promise in enhancing the analysis of patient exams. These tools use Deep Learning, a subarea of Machine Learning, and they possess the potential to automate aspects of dental image analysis, extracting crucial information from images and facilitating thorough, intricate assessments while reducing subjectivity and saving a significant amount of time. Taking these factors into account, this work aims to employ Convolutional Neural Networks for tooth numbering and the identification of decayed teeth in panoramic radiographs. These tasks contribute to the development of a modular system in collaboration with another master\'s student. Utilizing a dataset of 18,836 tooth images for the numbering task, the following results were achieved: precision of 0.979, recall of 0.978, accuracy of 0.998, and an F1-score of 0.978. In the task of identifying decayed teeth, a combination of two datasets yielded the following results: precision of 0.963, recall of 0.914, accuracy of 0.939, and an F1-score of 0.937. These results illustrate the significant promise of employing Convolutional Neural Networks in tasks involving tooth numbering and identifying decayed teeth, thus allowing dentists more time for clinical treatments. Furthermore, automating the analysis of panoramic radiographs can aid in generating reports and populating dental records while serving as a secondary assessment for tooth identification, ultimately reducing the likelihood of errors.
  • DOI: 10.11606/D.59.2023.tde-20022024-135415
  • Editor: Biblioteca Digital de Teses e Dissertações da USP; Universidade de São Paulo; Faculdade de Filosofia, Ciências e Letras de Ribeirão Preto
  • Data de criação/publicação: 2023-11-27
  • Formato: Adobe PDF
  • Idioma: Inglês
Links
Voltar para lista de resultados
Resultado  1 AvançarIr para próxima página

  • Realização: ABCD-USP USP   Logos de Redes Sociais: Freepik

Buscando em bases de dados remotas. Favor aguardar.