Generative Adversarial Networks Based Data Augmentation for Ultrasound Fetal Brain Planes Classification

Generative adversarial networks have been recently applied to medical imaging on different modalities (MRI, CT, X-ray, etc). However there are not many applications on ultrasound modality as a data augmentation technique applied to downstream classification tasks. This experimental case study aims t...

Descripción completa

Detalles Bibliográficos
Autor: Montero Agudo, Alberto
Tipo de recurso: tesis de maestría
Fecha de publicación:2021
País:España
Institución:Universitat Oberta de Catalunya (UOC)
Repositorio:O2, repositorio institucional de la UOC
OAI Identifier:oai:openaccess.uoc.edu:10609/127107
Acceso en línea:https://hdl.handle.net/10609/127107
Access Level:acceso abierto
Palabra clave:generative adversarial networks
data augmentation
ultrasound fetal brain images classification
xarxes contradictòries generatives
augment de dades
classificació d'imatges cerebrals fetals per ultrasons
redes generativas adversarias
aumento de datos
clasificación de imágenes de ultrasonido del cerebro fetal
Databases -- TFM
Bases de dades -- TFM
Bases de datos -- TFM
Descripción
Sumario:Generative adversarial networks have been recently applied to medical imaging on different modalities (MRI, CT, X-ray, etc). However there are not many applications on ultrasound modality as a data augmentation technique applied to downstream classification tasks. This experimental case study aims to explore and evaluate the generation of synthetic ultrasound fetal brain images via generative adversarial networks and apply to ultrasound fetal brain plane classification tasks. State of the art Generative Adversarial Networks stylegan2-ada was applied to fetal brain image generation and GAN-based data augmentation classifiers were compared with baseline classifiers. Our experimental results show that GAN-Based data augmentation combined with classical data augmentation outperforms classifiers with only classical data augmentation by 2% in both accuracy and area under the curve score.