Please use this identifier to cite or link to this item: https://doi.org/10.15480/882.3822
DC FieldValueLanguage
dc.contributor.authorBengs, Marcel-
dc.contributor.authorPant, Satish-
dc.contributor.authorBockmayr, Michael-
dc.contributor.authorSchüller, Ulrich-
dc.contributor.authorSchlaefer, Alexander-
dc.date.accessioned2021-10-14T04:49:54Z-
dc.date.available2021-10-14T04:49:54Z-
dc.date.issued2021-08-01-
dc.identifier.citationCurrent Directions in Biomedical Engineering 7 (1): 20211115, 63-66 (2021-08-01)de_DE
dc.identifier.issn2364-5504de_DE
dc.identifier.urihttp://hdl.handle.net/11420/10491-
dc.description.abstractMedulloblastoma (MB) is a primary central nervous system tumor and the most common malignant brain cancer among children. Neuropathologists perform microscopic inspection of histopathological tissue slides under a microscope to assess the severity of the tumor. This is a timeconsuming task and often infused with observer variability. Recently, pre-trained convolutional neural networks (CNN) have shown promising results for MB subtype classification. Typically, high-resolution images are divided into smaller tiles for classification, while the size of the tiles has not been systematically evaluated. We study the impact of tile size and input strategy and classify the two major histopathological subtypes-Classic and Desmoplastic/Nodular. To this end, we use recently proposed EfficientNets and evaluate tiles with increasing size combined with various downsampling scales. Our results demonstrate using large input tiles pixels followed by intermediate downsampling and patch cropping significantly improves MB classification performance. Our top-performing method achieves the AUC-ROC value of 90.90% compared to 84.53% using the previous approach with smaller input tiles.en
dc.language.isoende_DE
dc.publisherDe Gruyterde_DE
dc.relation.ispartofCurrent directions in biomedical engineeringde_DE
dc.rights.urihttps://creativecommons.org/licenses/by/4.0/de_DE
dc.subjectconvolutional neural networksde_DE
dc.subjectdigital pathologyde_DE
dc.subjecthistopathologyde_DE
dc.subjectmedulloblastomade_DE
dc.subjectTransfer learningde_DE
dc.subject.ddc600: Technikde_DE
dc.subject.ddc610: Medizinde_DE
dc.titleMulti-scale input strategies for medulloblastoma tumor classification using deep transfer learningde_DE
dc.typeArticlede_DE
dc.identifier.doi10.15480/882.3822-
dc.type.diniarticle-
dcterms.DCMITypeText-
tuhh.identifier.urnurn:nbn:de:gbv:830-882.0147320-
tuhh.oai.showtruede_DE
tuhh.abstract.englishMedulloblastoma (MB) is a primary central nervous system tumor and the most common malignant brain cancer among children. Neuropathologists perform microscopic inspection of histopathological tissue slides under a microscope to assess the severity of the tumor. This is a timeconsuming task and often infused with observer variability. Recently, pre-trained convolutional neural networks (CNN) have shown promising results for MB subtype classification. Typically, high-resolution images are divided into smaller tiles for classification, while the size of the tiles has not been systematically evaluated. We study the impact of tile size and input strategy and classify the two major histopathological subtypes-Classic and Desmoplastic/Nodular. To this end, we use recently proposed EfficientNets and evaluate tiles with increasing size combined with various downsampling scales. Our results demonstrate using large input tiles pixels followed by intermediate downsampling and patch cropping significantly improves MB classification performance. Our top-performing method achieves the AUC-ROC value of 90.90% compared to 84.53% using the previous approach with smaller input tiles.de_DE
tuhh.publisher.doi10.1515/cdbme-2021-1014-
tuhh.publication.instituteMedizintechnische und Intelligente Systeme E-1de_DE
tuhh.identifier.doi10.15480/882.3822-
tuhh.type.opus(wissenschaftlicher) Artikel-
dc.type.driverarticle-
dc.type.casraiJournal Article-
tuhh.container.issue1de_DE
tuhh.container.volume7de_DE
tuhh.container.startpage63de_DE
tuhh.container.endpage66de_DE
dc.rights.nationallicensefalsede_DE
dc.identifier.scopus2-s2.0-85114439798de_DE
tuhh.container.articlenumber20211115de_DE
local.status.inpressfalsede_DE
local.type.versionpublishedVersionde_DE
local.funding.infoThis work was partially supported by the Hamburg University of Technology i3 initiative.de_DE
item.openairecristypehttp://purl.org/coar/resource_type/c_6501-
item.creatorOrcidBengs, Marcel-
item.creatorOrcidPant, Satish-
item.creatorOrcidBockmayr, Michael-
item.creatorOrcidSchüller, Ulrich-
item.creatorOrcidSchlaefer, Alexander-
item.cerifentitytypePublications-
item.mappedtypeArticle-
item.openairetypeArticle-
item.fulltextWith Fulltext-
item.grantfulltextopen-
item.creatorGNDBengs, Marcel-
item.creatorGNDPant, Satish-
item.creatorGNDBockmayr, Michael-
item.creatorGNDSchüller, Ulrich-
item.creatorGNDSchlaefer, Alexander-
item.languageiso639-1en-
crisitem.author.deptMedizintechnische und Intelligente Systeme E-1-
crisitem.author.deptMedizintechnische und Intelligente Systeme E-1-
crisitem.author.orcid0000-0002-2229-9547-
crisitem.author.parentorgStudiendekanat Elektrotechnik, Informatik und Mathematik-
crisitem.author.parentorgStudiendekanat Elektrotechnik, Informatik und Mathematik-
Appears in Collections:Publications with fulltext
Files in This Item:
File Description SizeFormat
10.1515_cdbme-2021-1014.pdfVerlags-PDF566,18 kBAdobe PDFView/Open
Thumbnail
Show simple item record

Page view(s)

21
Last Week
0
Last month
1
checked on Dec 5, 2021

Download(s)

8
checked on Dec 5, 2021

Google ScholarTM

Check

Note about this record

Cite this record

Export

This item is licensed under a Creative Commons License Creative Commons