Effect of decay in magnetic resonance imaging on deep neural networks
Autor(a) principal: | |
---|---|
Data de Publicação: | 2022 |
Outros Autores: | , , , , |
Tipo de documento: | Artigo |
Idioma: | por |
Título da fonte: | Research, Society and Development |
Texto Completo: | https://rsdjournal.org/index.php/rsd/article/view/31868 |
Resumo: | In the last decades, tasks of classification and segmentation of clinical findings using convolutional neural networkshave grown significantly in the sphere of radiology, and more precisely in the modality of magnetic resonance imaging. However, little is known about the behavior of the proposed architectures when faced with factors that degrade the spatial resolution and contrast resolution, since most models are trained with high quality images, which is not consistent with the general daily life. Therefore, it is necessary to analyze the performance of pre-trained neural networks under conditions in which there is deterioration of the input image. In this work, the effects of degradation of the resolutions were evaluated, both in classification and segmentation tasks of brain tumors, for three architectures: Mobilenet, Vgg16 and SEResNeXt50. The results obtained showed that the tasks performed are greatly affected by image quality distortions, especially in cases where the deteriorations become more intense. |
id |
UNIFEI_1ad27aa7b45c9bb3b8e9ff0c4f4b9fa9 |
---|---|
oai_identifier_str |
oai:ojs.pkp.sfu.ca:article/31868 |
network_acronym_str |
UNIFEI |
network_name_str |
Research, Society and Development |
repository_id_str |
|
spelling |
Effect of decay in magnetic resonance imaging on deep neural networksEfecto de la decadencia en las imágenes de resonancia magnética en las redes neuronales profundasEfeito da deterioração em imagens por ressonância magnética sobre redes neurais profundasMagnetic ResonanceBrain TumorsDeep learning.Resonancia magnéticaTumores del CerebroAprendizaje profundo.Imageamento por Ressonância MagnéticaTumores CerebraisAprendizado profundo.In the last decades, tasks of classification and segmentation of clinical findings using convolutional neural networkshave grown significantly in the sphere of radiology, and more precisely in the modality of magnetic resonance imaging. However, little is known about the behavior of the proposed architectures when faced with factors that degrade the spatial resolution and contrast resolution, since most models are trained with high quality images, which is not consistent with the general daily life. Therefore, it is necessary to analyze the performance of pre-trained neural networks under conditions in which there is deterioration of the input image. In this work, the effects of degradation of the resolutions were evaluated, both in classification and segmentation tasks of brain tumors, for three architectures: Mobilenet, Vgg16 and SEResNeXt50. The results obtained showed that the tasks performed are greatly affected by image quality distortions, especially in cases where the deteriorations become more intense.En las últimas décadas, las tareas de clasificación y segmentación de hallazgos clínicos mediante redes neuronales convolucionales han crecido significativamente en el ámbito del diagnóstico por imagen y, más precisamente, en la modalidad de resonancia magnética. Sin embargo, poco se sabe sobre el comportamiento de estas arquitecturas ante factores que degradan la resolución espacial y la resolución de contraste, ya que la mayoría de los modelos son entrenados con imágenes de alta calidad, lo que no es acorde con la vida cotidiana general. Por lo tanto, es necesario analizar el desempeño de las redes neuronales pre-entrenadas, bajo condiciones en las que existe deterioro de la imagen de entrada. En este trabajo se evaluaron los efectos de la degradación de ambas resoluciones, tanto en tareas de clasificación como de segmentación de tumores cerebrales, para tres arquitecturas: Mobilenet, Vgg16 y SEResNeXt50. Los resultados obtenidos mostraron que las tareas realizadas se ven muy afectadas por distorsiones en la calidad de las imágenes, especialmente en los casos en que los deterioros se vuelven más intensos.Nas últimas décadas, tarefas de classificação e segmentação de achados clínicos com uso de redes neurais convolucionais cresceram bastante na esfera do diagnóstico por imagem e, mais precisamente, na modalidade de imagem por ressonância magnética. Porém, pouco se sabe a respeito do comportamento dessas arquiteturas quando confrontadas com fatores que degradam a resolução espacial e a resolução de contraste, uma vez que a maioria dos modelos é treinada com imagens de alta qualidade, o que não é condizente com o cotidiano geral. Por isso, faz-se necessário analisar a performance das redes neurais pré-treinadas, sob condições em que haja deterioração da imagem de entrada. Neste trabalho, foram avaliados os efeitos da degradação de ambas as resoluções, tanto em tarefas de classificação quanto de segmentação de tumores cerebrais, para três arquiteturas: Mobilenet, Vgg16 e SEResNeXt50. Os resultados obtidos demonstraram que as tarefas executadas são muito afetadas pelas distorções na qualidade das imagens, em especial nos casos em que as deteriorações se tornam mais intensas.Research, Society and Development2022-07-10info:eu-repo/semantics/articleinfo:eu-repo/semantics/publishedVersionapplication/pdfhttps://rsdjournal.org/index.php/rsd/article/view/3186810.33448/rsd-v11i9.31868Research, Society and Development; Vol. 11 No. 9; e31411931868Research, Society and Development; Vol. 11 Núm. 9; e31411931868Research, Society and Development; v. 11 n. 9; e314119318682525-3409reponame:Research, Society and Developmentinstname:Universidade Federal de Itajubá (UNIFEI)instacron:UNIFEIporhttps://rsdjournal.org/index.php/rsd/article/view/31868/27162Copyright (c) 2022 Carlos Leandro Silva dos Prazeres; Perseu Lúcio Alexander Helene de Paula; Mozart Nolasco Monte; Marcela Costa Alcântara Estácio ; Esdras Adriano Barbosa dos Santos; Laelia Camposhttps://creativecommons.org/licenses/by/4.0info:eu-repo/semantics/openAccessPrazeres, Carlos Leandro Silva dosPaula, Perseu Lúcio Alexander Helene deMonte, Mozart NolascoEstácio , Marcela Costa AlcântaraSantos, Esdras Adriano Barbosa dos Campos, Laelia2022-07-21T12:36:16Zoai:ojs.pkp.sfu.ca:article/31868Revistahttps://rsdjournal.org/index.php/rsd/indexPUBhttps://rsdjournal.org/index.php/rsd/oairsd.articles@gmail.com2525-34092525-3409opendoar:2024-01-17T09:48:03.707960Research, Society and Development - Universidade Federal de Itajubá (UNIFEI)false |
dc.title.none.fl_str_mv |
Effect of decay in magnetic resonance imaging on deep neural networks Efecto de la decadencia en las imágenes de resonancia magnética en las redes neuronales profundas Efeito da deterioração em imagens por ressonância magnética sobre redes neurais profundas |
title |
Effect of decay in magnetic resonance imaging on deep neural networks |
spellingShingle |
Effect of decay in magnetic resonance imaging on deep neural networks Prazeres, Carlos Leandro Silva dos Magnetic Resonance Brain Tumors Deep learning. Resonancia magnética Tumores del Cerebro Aprendizaje profundo. Imageamento por Ressonância Magnética Tumores Cerebrais Aprendizado profundo. |
title_short |
Effect of decay in magnetic resonance imaging on deep neural networks |
title_full |
Effect of decay in magnetic resonance imaging on deep neural networks |
title_fullStr |
Effect of decay in magnetic resonance imaging on deep neural networks |
title_full_unstemmed |
Effect of decay in magnetic resonance imaging on deep neural networks |
title_sort |
Effect of decay in magnetic resonance imaging on deep neural networks |
author |
Prazeres, Carlos Leandro Silva dos |
author_facet |
Prazeres, Carlos Leandro Silva dos Paula, Perseu Lúcio Alexander Helene de Monte, Mozart Nolasco Estácio , Marcela Costa Alcântara Santos, Esdras Adriano Barbosa dos Campos, Laelia |
author_role |
author |
author2 |
Paula, Perseu Lúcio Alexander Helene de Monte, Mozart Nolasco Estácio , Marcela Costa Alcântara Santos, Esdras Adriano Barbosa dos Campos, Laelia |
author2_role |
author author author author author |
dc.contributor.author.fl_str_mv |
Prazeres, Carlos Leandro Silva dos Paula, Perseu Lúcio Alexander Helene de Monte, Mozart Nolasco Estácio , Marcela Costa Alcântara Santos, Esdras Adriano Barbosa dos Campos, Laelia |
dc.subject.por.fl_str_mv |
Magnetic Resonance Brain Tumors Deep learning. Resonancia magnética Tumores del Cerebro Aprendizaje profundo. Imageamento por Ressonância Magnética Tumores Cerebrais Aprendizado profundo. |
topic |
Magnetic Resonance Brain Tumors Deep learning. Resonancia magnética Tumores del Cerebro Aprendizaje profundo. Imageamento por Ressonância Magnética Tumores Cerebrais Aprendizado profundo. |
description |
In the last decades, tasks of classification and segmentation of clinical findings using convolutional neural networkshave grown significantly in the sphere of radiology, and more precisely in the modality of magnetic resonance imaging. However, little is known about the behavior of the proposed architectures when faced with factors that degrade the spatial resolution and contrast resolution, since most models are trained with high quality images, which is not consistent with the general daily life. Therefore, it is necessary to analyze the performance of pre-trained neural networks under conditions in which there is deterioration of the input image. In this work, the effects of degradation of the resolutions were evaluated, both in classification and segmentation tasks of brain tumors, for three architectures: Mobilenet, Vgg16 and SEResNeXt50. The results obtained showed that the tasks performed are greatly affected by image quality distortions, especially in cases where the deteriorations become more intense. |
publishDate |
2022 |
dc.date.none.fl_str_mv |
2022-07-10 |
dc.type.driver.fl_str_mv |
info:eu-repo/semantics/article info:eu-repo/semantics/publishedVersion |
format |
article |
status_str |
publishedVersion |
dc.identifier.uri.fl_str_mv |
https://rsdjournal.org/index.php/rsd/article/view/31868 10.33448/rsd-v11i9.31868 |
url |
https://rsdjournal.org/index.php/rsd/article/view/31868 |
identifier_str_mv |
10.33448/rsd-v11i9.31868 |
dc.language.iso.fl_str_mv |
por |
language |
por |
dc.relation.none.fl_str_mv |
https://rsdjournal.org/index.php/rsd/article/view/31868/27162 |
dc.rights.driver.fl_str_mv |
https://creativecommons.org/licenses/by/4.0 info:eu-repo/semantics/openAccess |
rights_invalid_str_mv |
https://creativecommons.org/licenses/by/4.0 |
eu_rights_str_mv |
openAccess |
dc.format.none.fl_str_mv |
application/pdf |
dc.publisher.none.fl_str_mv |
Research, Society and Development |
publisher.none.fl_str_mv |
Research, Society and Development |
dc.source.none.fl_str_mv |
Research, Society and Development; Vol. 11 No. 9; e31411931868 Research, Society and Development; Vol. 11 Núm. 9; e31411931868 Research, Society and Development; v. 11 n. 9; e31411931868 2525-3409 reponame:Research, Society and Development instname:Universidade Federal de Itajubá (UNIFEI) instacron:UNIFEI |
instname_str |
Universidade Federal de Itajubá (UNIFEI) |
instacron_str |
UNIFEI |
institution |
UNIFEI |
reponame_str |
Research, Society and Development |
collection |
Research, Society and Development |
repository.name.fl_str_mv |
Research, Society and Development - Universidade Federal de Itajubá (UNIFEI) |
repository.mail.fl_str_mv |
rsd.articles@gmail.com |
_version_ |
1797052768649740288 |