Error-Correcting Mean-Teacher: Corrections instead of consistency-targets applied to semi-supervised medical image segmentation

Detalhes bibliográficos
Autor(a) principal: Mendel, Robert
Data de Publicação: 2023
Outros Autores: Rauber, David, de Souza, Luis A., Papa, João P. [UNESP], Palm, Christoph
Tipo de documento: Artigo
Idioma: eng
Título da fonte: Repositório Institucional da UNESP
Texto Completo: http://dx.doi.org/10.1016/j.compbiomed.2023.106585
http://hdl.handle.net/11449/249677
Resumo: Semantic segmentation is an essential task in medical imaging research. Many powerful deep-learning-based approaches can be employed for this problem, but they are dependent on the availability of an expansive labeled dataset. In this work, we augment such supervised segmentation models to be suitable for learning from unlabeled data. Our semi-supervised approach, termed Error-Correcting Mean-Teacher, uses an exponential moving average model like the original Mean Teacher but introduces our new paradigm of error correction. The original segmentation network is augmented to handle this secondary correction task. Both tasks build upon the core feature extraction layers of the model. For the correction task, features detected in the input image are fused with features detected in the predicted segmentation and further processed with task-specific decoder layers. The combination of image and segmentation features allows the model to correct present mistakes in the given input pair. The correction task is trained jointly on the labeled data. On unlabeled data, the exponential moving average of the original network corrects the student's prediction. The combined outputs of the students’ prediction with the teachers’ correction form the basis for the semi-supervised update. We evaluate our method with the 2017 and 2018 Robotic Scene Segmentation data, the ISIC 2017 and the BraTS 2020 Challenges, a proprietary Endoscopic Submucosal Dissection dataset, Cityscapes, and Pascal VOC 2012. Additionally, we analyze the impact of the individual components and examine the behavior when the amount of labeled data varies, with experiments performed on two distinct segmentation architectures. Our method shows improvements in terms of the mean Intersection over Union over the supervised baseline and competing methods. Code is available at https://github.com/CloneRob/ECMT.
id UNSP_ace0ff8cfb1cbb034b4f25e30cd7fe9a
oai_identifier_str oai:repositorio.unesp.br:11449/249677
network_acronym_str UNSP
network_name_str Repositório Institucional da UNESP
repository_id_str 2946
spelling Error-Correcting Mean-Teacher: Corrections instead of consistency-targets applied to semi-supervised medical image segmentationMean-TeacherMedical imagingPseudo-labelsSegmentationSemi-supervisedSemantic segmentation is an essential task in medical imaging research. Many powerful deep-learning-based approaches can be employed for this problem, but they are dependent on the availability of an expansive labeled dataset. In this work, we augment such supervised segmentation models to be suitable for learning from unlabeled data. Our semi-supervised approach, termed Error-Correcting Mean-Teacher, uses an exponential moving average model like the original Mean Teacher but introduces our new paradigm of error correction. The original segmentation network is augmented to handle this secondary correction task. Both tasks build upon the core feature extraction layers of the model. For the correction task, features detected in the input image are fused with features detected in the predicted segmentation and further processed with task-specific decoder layers. The combination of image and segmentation features allows the model to correct present mistakes in the given input pair. The correction task is trained jointly on the labeled data. On unlabeled data, the exponential moving average of the original network corrects the student's prediction. The combined outputs of the students’ prediction with the teachers’ correction form the basis for the semi-supervised update. We evaluate our method with the 2017 and 2018 Robotic Scene Segmentation data, the ISIC 2017 and the BraTS 2020 Challenges, a proprietary Endoscopic Submucosal Dissection dataset, Cityscapes, and Pascal VOC 2012. Additionally, we analyze the impact of the individual components and examine the behavior when the amount of labeled data varies, with experiments performed on two distinct segmentation architectures. Our method shows improvements in terms of the mean Intersection over Union over the supervised baseline and competing methods. Code is available at https://github.com/CloneRob/ECMT.Regensburg Medical Image Computing (ReMIC) Ostbayerische Technische Hochschule Regensburg (OTH Regensburg)Regensburg Center of Health Sciences and Technology (RCHST) OTH RegensburgComputer Science Department Federal University of São CarlosDepartment of Computing São Paulo State UniversityDepartment of Computing São Paulo State UniversityOstbayerische Technische Hochschule Regensburg (OTH Regensburg)OTH RegensburgUniversidade Federal de São Carlos (UFSCar)Universidade Estadual Paulista (UNESP)Mendel, RobertRauber, Davidde Souza, Luis A.Papa, João P. [UNESP]Palm, Christoph2023-07-29T16:06:12Z2023-07-29T16:06:12Z2023-03-01info:eu-repo/semantics/publishedVersioninfo:eu-repo/semantics/articlehttp://dx.doi.org/10.1016/j.compbiomed.2023.106585Computers in Biology and Medicine, v. 154.1879-05340010-4825http://hdl.handle.net/11449/24967710.1016/j.compbiomed.2023.1065852-s2.0-85148480011Scopusreponame:Repositório Institucional da UNESPinstname:Universidade Estadual Paulista (UNESP)instacron:UNESPengComputers in Biology and Medicineinfo:eu-repo/semantics/openAccess2024-04-23T16:11:00Zoai:repositorio.unesp.br:11449/249677Repositório InstitucionalPUBhttp://repositorio.unesp.br/oai/requestopendoar:29462024-08-05T21:10:43.919619Repositório Institucional da UNESP - Universidade Estadual Paulista (UNESP)false
dc.title.none.fl_str_mv Error-Correcting Mean-Teacher: Corrections instead of consistency-targets applied to semi-supervised medical image segmentation
title Error-Correcting Mean-Teacher: Corrections instead of consistency-targets applied to semi-supervised medical image segmentation
spellingShingle Error-Correcting Mean-Teacher: Corrections instead of consistency-targets applied to semi-supervised medical image segmentation
Mendel, Robert
Mean-Teacher
Medical imaging
Pseudo-labels
Segmentation
Semi-supervised
title_short Error-Correcting Mean-Teacher: Corrections instead of consistency-targets applied to semi-supervised medical image segmentation
title_full Error-Correcting Mean-Teacher: Corrections instead of consistency-targets applied to semi-supervised medical image segmentation
title_fullStr Error-Correcting Mean-Teacher: Corrections instead of consistency-targets applied to semi-supervised medical image segmentation
title_full_unstemmed Error-Correcting Mean-Teacher: Corrections instead of consistency-targets applied to semi-supervised medical image segmentation
title_sort Error-Correcting Mean-Teacher: Corrections instead of consistency-targets applied to semi-supervised medical image segmentation
author Mendel, Robert
author_facet Mendel, Robert
Rauber, David
de Souza, Luis A.
Papa, João P. [UNESP]
Palm, Christoph
author_role author
author2 Rauber, David
de Souza, Luis A.
Papa, João P. [UNESP]
Palm, Christoph
author2_role author
author
author
author
dc.contributor.none.fl_str_mv Ostbayerische Technische Hochschule Regensburg (OTH Regensburg)
OTH Regensburg
Universidade Federal de São Carlos (UFSCar)
Universidade Estadual Paulista (UNESP)
dc.contributor.author.fl_str_mv Mendel, Robert
Rauber, David
de Souza, Luis A.
Papa, João P. [UNESP]
Palm, Christoph
dc.subject.por.fl_str_mv Mean-Teacher
Medical imaging
Pseudo-labels
Segmentation
Semi-supervised
topic Mean-Teacher
Medical imaging
Pseudo-labels
Segmentation
Semi-supervised
description Semantic segmentation is an essential task in medical imaging research. Many powerful deep-learning-based approaches can be employed for this problem, but they are dependent on the availability of an expansive labeled dataset. In this work, we augment such supervised segmentation models to be suitable for learning from unlabeled data. Our semi-supervised approach, termed Error-Correcting Mean-Teacher, uses an exponential moving average model like the original Mean Teacher but introduces our new paradigm of error correction. The original segmentation network is augmented to handle this secondary correction task. Both tasks build upon the core feature extraction layers of the model. For the correction task, features detected in the input image are fused with features detected in the predicted segmentation and further processed with task-specific decoder layers. The combination of image and segmentation features allows the model to correct present mistakes in the given input pair. The correction task is trained jointly on the labeled data. On unlabeled data, the exponential moving average of the original network corrects the student's prediction. The combined outputs of the students’ prediction with the teachers’ correction form the basis for the semi-supervised update. We evaluate our method with the 2017 and 2018 Robotic Scene Segmentation data, the ISIC 2017 and the BraTS 2020 Challenges, a proprietary Endoscopic Submucosal Dissection dataset, Cityscapes, and Pascal VOC 2012. Additionally, we analyze the impact of the individual components and examine the behavior when the amount of labeled data varies, with experiments performed on two distinct segmentation architectures. Our method shows improvements in terms of the mean Intersection over Union over the supervised baseline and competing methods. Code is available at https://github.com/CloneRob/ECMT.
publishDate 2023
dc.date.none.fl_str_mv 2023-07-29T16:06:12Z
2023-07-29T16:06:12Z
2023-03-01
dc.type.status.fl_str_mv info:eu-repo/semantics/publishedVersion
dc.type.driver.fl_str_mv info:eu-repo/semantics/article
format article
status_str publishedVersion
dc.identifier.uri.fl_str_mv http://dx.doi.org/10.1016/j.compbiomed.2023.106585
Computers in Biology and Medicine, v. 154.
1879-0534
0010-4825
http://hdl.handle.net/11449/249677
10.1016/j.compbiomed.2023.106585
2-s2.0-85148480011
url http://dx.doi.org/10.1016/j.compbiomed.2023.106585
http://hdl.handle.net/11449/249677
identifier_str_mv Computers in Biology and Medicine, v. 154.
1879-0534
0010-4825
10.1016/j.compbiomed.2023.106585
2-s2.0-85148480011
dc.language.iso.fl_str_mv eng
language eng
dc.relation.none.fl_str_mv Computers in Biology and Medicine
dc.rights.driver.fl_str_mv info:eu-repo/semantics/openAccess
eu_rights_str_mv openAccess
dc.source.none.fl_str_mv Scopus
reponame:Repositório Institucional da UNESP
instname:Universidade Estadual Paulista (UNESP)
instacron:UNESP
instname_str Universidade Estadual Paulista (UNESP)
instacron_str UNESP
institution UNESP
reponame_str Repositório Institucional da UNESP
collection Repositório Institucional da UNESP
repository.name.fl_str_mv Repositório Institucional da UNESP - Universidade Estadual Paulista (UNESP)
repository.mail.fl_str_mv
_version_ 1808129294189199360