MMI-GAN: multi medical imaging translation using generative adversarial network
Autor(a) principal: | |
---|---|
Data de Publicação: | 2020 |
Tipo de documento: | Dissertação |
Idioma: | por |
Título da fonte: | Repositório Institucional da Universidade Federal de Alagoas (UFAL) |
Texto Completo: | http://www.repositorio.ufal.br/handle/riufal/7471 |
Resumo: | Medical image translation is considered a new frontier in the field of medical image analysis, with great potential for application. However, existing approaches have limited scalability and robustness in handling more than two image domains, since different models must be created independently for each pair of domains. To address these limitations, we developed MMI-GAN, a new approach for translation between multiple image domains, capable of translating intermodal (CT and RM) and intramodal (PD, T1 and T2) images using only a single generator and a discriminator, trained with image data from all domains. We propose a GAN architecture that can be easily extended to other translation tasks for the benefit of the medical imaging community. MMI-GAN is based on recent advances in the area of GANs (Generative Adversarial Network), using an adversary structure with a new combination of non-adversarial losses, which allows the simultaneous training of several data sets with different domains in the same network, as well as the innovative capacity to translate with flexibility between and inter/intra modalities. The images translated by MMI-GAN managed to obtain MAE of 5.792, PSNR of 27.398, MI of 1.430 and SSIM of 0.900. Its results were shown, often statically comparable or superior to Pix2pix and in almost all translations it was superior to Cyclegan. |
id |
UFAL_a01e82272f9db401a5091093c31f99b4 |
---|---|
oai_identifier_str |
oai:www.repositorio.ufal.br:riufal/7471 |
network_acronym_str |
UFAL |
network_name_str |
Repositório Institucional da Universidade Federal de Alagoas (UFAL) |
repository_id_str |
|
spelling |
MMI-GAN: multi medical imaging translation using generative adversarial networkMMI-GAN: tradução multi-imagem médica usando rede adversarial generativaGenerative adversarial networkTradução de imagensImagem por ressonância magnéticaTomografia computadorizadaGenerative adversarial networksImage translationMulti-domainMagnetic resonanceComputed tomographyCNPQ::CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAOMedical image translation is considered a new frontier in the field of medical image analysis, with great potential for application. However, existing approaches have limited scalability and robustness in handling more than two image domains, since different models must be created independently for each pair of domains. To address these limitations, we developed MMI-GAN, a new approach for translation between multiple image domains, capable of translating intermodal (CT and RM) and intramodal (PD, T1 and T2) images using only a single generator and a discriminator, trained with image data from all domains. We propose a GAN architecture that can be easily extended to other translation tasks for the benefit of the medical imaging community. MMI-GAN is based on recent advances in the area of GANs (Generative Adversarial Network), using an adversary structure with a new combination of non-adversarial losses, which allows the simultaneous training of several data sets with different domains in the same network, as well as the innovative capacity to translate with flexibility between and inter/intra modalities. The images translated by MMI-GAN managed to obtain MAE of 5.792, PSNR of 27.398, MI of 1.430 and SSIM of 0.900. Its results were shown, often statically comparable or superior to Pix2pix and in almost all translations it was superior to Cyclegan.CAPES - Coordenação de Aperfeiçoamento de Pessoal de Nível SuperiorA tradução de imagens médicas é considerada uma nova fronteira no campo da análise de imagens médicas, com grande potencial de aplicação. No entanto, as abordagens existentes têm escalabilidade e robustez limitadas no manuseio de mais de dois domínios de imagens, uma vez que diferentes modelos devem ser criados independentemente para cada par de domínios. Para resolver essas limitações, desenvolvemos a MMI-GAN, uma nova abordagem para tradução entre múltiplos domínios de imagem, capaz de traduzir imagens intermodais (TC e RM) e intramodais (PD, T1 e T2) usando apenas um único gerador e um discriminador, treinados com dados de imagens de todos os domínios. Propomos uma arquitetura GAN que pode ser facilmente estendida a outras tarefas de tradução para o benefício da comunidade de imagens médicas. A MMI-GAN baseia-se nos avanços recentes na área das GANs (Generative Adversarial Network), utilizando uma estrutura adversária com uma nova combinação de perdas não adversárias, que permite o treino simultâneo de vários conjuntos de dados com diferentes domínios numa mesma rede, bem como a capacidade inovadora de traduzir com flexibilidade entre e intra/inter modalidades. As imagens traduzidas pelo MMI-GAN conseguiram obter MAE de 5.792, PSNR de 27.398, MI de 1.430 e SSIM de 0.900. Os seus resultados se mostraram, por muitas vezes estaticamente equiparáveis ou superiores a Pix2pix e em quase todas as traduções foi superior a Cyclegan.Universidade Federal de AlagoasBrasilPrograma de Pós-Graduação em InformáticaUFALOliveira, Marcelo Costahttp://lattes.cnpq.br/9562890319093965Vieira, Tiago Figueiredohttp://lattes.cnpq.br/8601011832053651Marques, Paulo Mazzoncini de Azevedohttp://lattes.cnpq.br/7119886675051877Souza, Eduardo Felipe de2021-01-22T04:13:31Z2021-01-212021-01-22T04:13:31Z2020-11-27info:eu-repo/semantics/publishedVersioninfo:eu-repo/semantics/masterThesisapplication/pdfSOUZA, Eduardo Felipe de. MMI-GAN: multi medical imaging translation using generative adversarial network. 2020. 57 f. Dissertação (Mestrado em Informática) - Instituto de Computação, Programa de Pós Graduação em Informática, Universidade Federal de Alagoas, Maceió, 2021.http://www.repositorio.ufal.br/handle/riufal/7471porinfo:eu-repo/semantics/openAccessreponame:Repositório Institucional da Universidade Federal de Alagoas (UFAL)instname:Universidade Federal de Alagoas (UFAL)instacron:UFAL2021-01-22T04:13:31Zoai:www.repositorio.ufal.br:riufal/7471Repositório InstitucionalPUBhttp://www.repositorio.ufal.br/oai/requestri@sibi.ufal.bropendoar:2021-01-22T04:13:31Repositório Institucional da Universidade Federal de Alagoas (UFAL) - Universidade Federal de Alagoas (UFAL)false |
dc.title.none.fl_str_mv |
MMI-GAN: multi medical imaging translation using generative adversarial network MMI-GAN: tradução multi-imagem médica usando rede adversarial generativa |
title |
MMI-GAN: multi medical imaging translation using generative adversarial network |
spellingShingle |
MMI-GAN: multi medical imaging translation using generative adversarial network Souza, Eduardo Felipe de Generative adversarial network Tradução de imagens Imagem por ressonância magnética Tomografia computadorizada Generative adversarial networks Image translation Multi-domain Magnetic resonance Computed tomography CNPQ::CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO |
title_short |
MMI-GAN: multi medical imaging translation using generative adversarial network |
title_full |
MMI-GAN: multi medical imaging translation using generative adversarial network |
title_fullStr |
MMI-GAN: multi medical imaging translation using generative adversarial network |
title_full_unstemmed |
MMI-GAN: multi medical imaging translation using generative adversarial network |
title_sort |
MMI-GAN: multi medical imaging translation using generative adversarial network |
author |
Souza, Eduardo Felipe de |
author_facet |
Souza, Eduardo Felipe de |
author_role |
author |
dc.contributor.none.fl_str_mv |
Oliveira, Marcelo Costa http://lattes.cnpq.br/9562890319093965 Vieira, Tiago Figueiredo http://lattes.cnpq.br/8601011832053651 Marques, Paulo Mazzoncini de Azevedo http://lattes.cnpq.br/7119886675051877 |
dc.contributor.author.fl_str_mv |
Souza, Eduardo Felipe de |
dc.subject.por.fl_str_mv |
Generative adversarial network Tradução de imagens Imagem por ressonância magnética Tomografia computadorizada Generative adversarial networks Image translation Multi-domain Magnetic resonance Computed tomography CNPQ::CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO |
topic |
Generative adversarial network Tradução de imagens Imagem por ressonância magnética Tomografia computadorizada Generative adversarial networks Image translation Multi-domain Magnetic resonance Computed tomography CNPQ::CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO |
description |
Medical image translation is considered a new frontier in the field of medical image analysis, with great potential for application. However, existing approaches have limited scalability and robustness in handling more than two image domains, since different models must be created independently for each pair of domains. To address these limitations, we developed MMI-GAN, a new approach for translation between multiple image domains, capable of translating intermodal (CT and RM) and intramodal (PD, T1 and T2) images using only a single generator and a discriminator, trained with image data from all domains. We propose a GAN architecture that can be easily extended to other translation tasks for the benefit of the medical imaging community. MMI-GAN is based on recent advances in the area of GANs (Generative Adversarial Network), using an adversary structure with a new combination of non-adversarial losses, which allows the simultaneous training of several data sets with different domains in the same network, as well as the innovative capacity to translate with flexibility between and inter/intra modalities. The images translated by MMI-GAN managed to obtain MAE of 5.792, PSNR of 27.398, MI of 1.430 and SSIM of 0.900. Its results were shown, often statically comparable or superior to Pix2pix and in almost all translations it was superior to Cyclegan. |
publishDate |
2020 |
dc.date.none.fl_str_mv |
2020-11-27 2021-01-22T04:13:31Z 2021-01-21 2021-01-22T04:13:31Z |
dc.type.status.fl_str_mv |
info:eu-repo/semantics/publishedVersion |
dc.type.driver.fl_str_mv |
info:eu-repo/semantics/masterThesis |
format |
masterThesis |
status_str |
publishedVersion |
dc.identifier.uri.fl_str_mv |
SOUZA, Eduardo Felipe de. MMI-GAN: multi medical imaging translation using generative adversarial network. 2020. 57 f. Dissertação (Mestrado em Informática) - Instituto de Computação, Programa de Pós Graduação em Informática, Universidade Federal de Alagoas, Maceió, 2021. http://www.repositorio.ufal.br/handle/riufal/7471 |
identifier_str_mv |
SOUZA, Eduardo Felipe de. MMI-GAN: multi medical imaging translation using generative adversarial network. 2020. 57 f. Dissertação (Mestrado em Informática) - Instituto de Computação, Programa de Pós Graduação em Informática, Universidade Federal de Alagoas, Maceió, 2021. |
url |
http://www.repositorio.ufal.br/handle/riufal/7471 |
dc.language.iso.fl_str_mv |
por |
language |
por |
dc.rights.driver.fl_str_mv |
info:eu-repo/semantics/openAccess |
eu_rights_str_mv |
openAccess |
dc.format.none.fl_str_mv |
application/pdf |
dc.publisher.none.fl_str_mv |
Universidade Federal de Alagoas Brasil Programa de Pós-Graduação em Informática UFAL |
publisher.none.fl_str_mv |
Universidade Federal de Alagoas Brasil Programa de Pós-Graduação em Informática UFAL |
dc.source.none.fl_str_mv |
reponame:Repositório Institucional da Universidade Federal de Alagoas (UFAL) instname:Universidade Federal de Alagoas (UFAL) instacron:UFAL |
instname_str |
Universidade Federal de Alagoas (UFAL) |
instacron_str |
UFAL |
institution |
UFAL |
reponame_str |
Repositório Institucional da Universidade Federal de Alagoas (UFAL) |
collection |
Repositório Institucional da Universidade Federal de Alagoas (UFAL) |
repository.name.fl_str_mv |
Repositório Institucional da Universidade Federal de Alagoas (UFAL) - Universidade Federal de Alagoas (UFAL) |
repository.mail.fl_str_mv |
ri@sibi.ufal.br |
_version_ |
1748233748538720256 |