Uma ferramenta baseada em redes morfológicas profundas para auxiliar no diagnóstico de hipertrofia adenoideana

• Ano de defesa: 2023
• Autor(a) principal: Martins, Guilherme Bueno
• Orientador(a): Alves, Wonder Alexandre Luz
• Banca de defesa: Alves, Wonder Alexandre Luz , Dantas, Daniel Oliveira , Araújo, Sidnei Alves de
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Nove de Julho
• Programa de Pós-Graduação: Programa de Pós-Graduação em Informática e Gestão do Conhecimento
• Departamento: Informática
• País: Brasil
• Palavras-chave em Português: redes morfológicas profundas; redes neurais artificiais; visão computacional; hipertrofia adenoideana; otorrinolaringologia
• Palavras-chave em Inglês: deep morphological networks; artificial neural networks; computational vision; adenoid hypertrophy; otolaryngology
• Área do conhecimento CNPq: CIENCIA DA COMPUTACAO::SISTEMAS DE COMPUTACAO
• Link de acesso: http://bibliotecatede.uninove.br/handle/tede/3512

Machine learning has revolutionized various fields of knowledge, including the medical field, significantly improving the state of the art in numerous applications. One prominent application is the processing of medical images using convolutional neural networks. However, the scarcity of available medical samples or data poses challenges to the effective application of this technique. This work proposes a method based on deep morphological networks for the development of an open-source software-based tool that aids in the diagnosis of adenoid hypertrophy. The tool employs machine learning techniques to recognizes whose frames (images) in a video of nasopharyngolaryngoscopy exam are ideals to the specialist medical formulate a adenoid hypertrophy diagnosis. The construction process of this tool involves the following steps: (1) Creation of a nasopharyngolaryngoscopy exam database. (2) Creation of a dataset labeled by a specialist. (3) Development of a model based on deep morphological networks. (4) Creation of a plugin for the open-source software ImageJ, utilizing the proposed model, which has been appropriately trained and validated. The proposed model combines the LeNet architecture with morphological operators, its accuracy is superior to 90% and precision above 90%. Through this work, we aim to contribute to the advancement of adenoid hypertrophy diagnosis, providing medical professionals with an effective and reliable machine learning-based tool.