Rhaiu-net: precise tumor segmentation in breast ultrasound images

Martínez Gárriz, Iñaki. (2022). Rhaiu-net: precise tumor segmentation in breast ultrasound images Master Thesis, Universidad Nacional de Educación a Distancia (España). Escuela Técnica Superior de Ingeniería Informática. Departamento de Inteligencia Artificial

Ficheros (Some files may be inaccessible until you login with your e-spacio credentials)
Nombre Descripción Tipo MIME Size
Martinez_Garriz_Inaki_TFM.pdf Martinez_Garriz_Inaki_TFM.pdf application/pdf 19.93MB

Título Rhaiu-net: precise tumor segmentation in breast ultrasound images
Autor(es) Martínez Gárriz, Iñaki
Abstract Breast cancer is the most common cancer among women. Moreover, according to the International Agency of Research on Cancer (IARC) breast cancer has already been the most diagnosed cancer in the world since 2021. The most determining factor when it comes to saving the life of a patient is its early detection. For this reason, computer-aided diagnosis (CAD) systems are being developed, so that breast cancer screening can be done faster and more accurately. Ultrasound is one of the most used screening techniques because it is inexpensive, fast, radiation-free, and painless. However, breast ultrasound image segmentation is a challenging task since they usually are speckle noisy and they suffer from low contrast. For this reason, the task is very time-consuming and needs to be performed by specialized doctors. The current trend in breast ultrasound image segmentation is building deep learning models, specifically; convolutional neural networks (CNNs), which are reporting very promising performance. In this work, we propose RHAIU-Net, a CNN-based novel architecture for breast ultrasound image segmentation. Specifically, it is based on the U-Net architecture, incorporating residual inception blocks, SiLU activation functions, attention gates, and Hartley pooling layers. We also propose a training framework which consists of data augmentation, image prepossessing, learning rate decay, and adversarial training. We have trained it along with other 11 models based on state-of-the-art architectures with the same dataset and framework, and then we have compared them. RHAIU-Net has outperformed nearly all of them in small tumors while performing as well as others in bigger ones.
Notas adicionales Trabajo de Fin de Máster Universitario en Investigación en Inteligencia Artificial. UNED
Materia(s) Ingeniería Informática
Palabra clave RHAIU-Net
Breast ultrasound (BUS)
Breast tumor
Tumor segmentation
Convolutional neural network
Editor(es) Universidad Nacional de Educación a Distancia (España). Escuela Técnica Superior de Ingeniería Informática. Departamento de Inteligencia Artificial
Director/Tutor Díez Vegas, Francisco Javier
Pérez Martín, Jorge
Rincón Zamorano, Mariano
Fecha 2022-06-01
Formato application/pdf
Identificador bibliuned:master-ETSInformatica-IIA-Imartinez
http://e-spacio.uned.es/fez/view/bibliuned:master-ETSInformatica-IIA-Imartinez
Idioma eng
Versión de la publicación acceptedVersion
Nivel de acceso y licencia http://creativecommons.org/licenses/by-nc-nd/4.0
info:eu-repo/semantics/openAccess
Tipo de recurso master Thesis
Tipo de acceso Acceso abierto

 
Versiones
Versión Tipo de filtro
Contador de citas: Google Scholar Search Google Scholar
Estadísticas de acceso: 142 Visitas, 105 Descargas  -  Estadísticas en detalle
Creado: Wed, 13 Sep 2023, 18:27:40 CET