Detalhes bibliográficos
| Ano de defesa: |
2024 |
| Autor(a) principal: |
FERNANDES, Arthur Guilherme Santos
 |
| Orientador(a): |
BRAZ JUNIOR, Geraldo
|
| Banca de defesa: |
BRAZ JUNIOR, Geraldo
,
DINIZ, João Otávio Bandeira
,
ALMEIDA, João Dallyson Sousa de
,
CUNHA, António Manuel Trigueiros da Silva
|
| Tipo de documento: |
Dissertação
|
| Tipo de acesso: |
Acesso aberto |
| Idioma: |
por |
| Instituição de defesa: |
Universidade Federal do Maranhão
|
| Programa de Pós-Graduação: |
PROGRAMA DE PÓS-GRADUAÇÃO EM CIÊNCIA DA COMPUTAÇÃO/CCET
|
| Departamento: |
DEPARTAMENTO DE INFORMÁTICA/CCET
|
| País: |
Brasil
|
| Palavras-chave em Português: |
|
| Palavras-chave em Inglês: |
|
| Área do conhecimento CNPq: |
|
| Link de acesso: |
https://tedebc.ufma.br/jspui/handle/tede/5485
|
Resumo: |
Organs at Risk (OAR) segmentation is crucial in radiotherapy planning. Its objective is to demarcate healthy tissues so that ionizing radiation is directed only to cancer cells. To do this, doctors demarcate the organs manually, which makes the process very time consuming and prone to errors. Therefore, automatic segmentation methodologies using deep learning can accelerate organ delineation during radiotherapy planning. Existing methodologies have many parameters, which makes the model cumbersome and expensive to make available as a service. This work proposes the use of a convolutional neural network architecture called EfficientDeeplab, trained on computed tomography scans to perform trachea segmentation. The model differs from other architectures by having fewer parameters, which makes it ideal for applications in large-scale healthcare services. To obtain a low number of parameters and quality segmentation, atrous convolutions and the EfficientNet architecture were applied. Tests were carried out on the SEGTHOR dataset, which obtained a dice score of 82.21%. |
