Robust Histopathology Image Analysis: To Label or to Synthesize?

Detection, segmentation and classification of nuclei are fundamental analysis operations in digital pathology. Existing state-of-the-art approaches demand extensive amount of supervised training data from pathologists and may still perform poorly in images from unseen tissue types. We propose an uns...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:Proceedings (IEEE Computer Society Conference on Computer Vision and Pattern Recognition. Online) Ročník 2019; s. 8525 - 8534
Hlavní autoři: Hou, Le, Agarwal, Ayush, Samaras, Dimitris, Kurc, Tahsin M., Gupta, Rajarsi R., Saltz, Joel H.
Médium: Konferenční příspěvek Journal Article
Jazyk:angličtina
Vydáno: United States IEEE 01.06.2019
Témata:
ISSN:1063-6919, 1063-6919
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Abstract Detection, segmentation and classification of nuclei are fundamental analysis operations in digital pathology. Existing state-of-the-art approaches demand extensive amount of supervised training data from pathologists and may still perform poorly in images from unseen tissue types. We propose an unsupervised approach for histopathology image segmentation that synthesizes heterogeneous sets of training image patches, of every tissue type. Although our synthetic patches are not always of high quality, we harness the motley crew of generated samples through a generally applicable importance sampling method. This proposed approach, for the first time, re-weighs the training loss over synthetic data so that the ideal (unbiased) generalization loss over the true data distribution is minimized. This enables us to use a random polygon generator to synthesize approximate cellular structures (i.e., nuclear masks) for which no real examples are given in many tissue types, and hence, GAN-based methods are not suited. In addition, we propose a hybrid synthesis pipeline that utilizes textures in real histopathology patches and GAN models, to tackle heterogeneity in tissue textures. Compared with existing state-of-the-art supervised models, our approach generalizes significantly better on cancer types without training data. Even in cancer types with training data, our approach achieves the same performance without supervision cost. We release code and segmentation results on over 5000 Whole Slide Images (WSI) in The Cancer Genome Atlas (TCGA) repository, a dataset that would be orders of magnitude larger than what is available today.
AbstractList Detection, segmentation and classification of nuclei are fundamental analysis operations in digital pathology. Existing state-of-the-art approaches demand extensive amount of supervised training data from pathologists and may still perform poorly in images from unseen tissue types. We propose an unsupervised approach for histopathology image segmentation that synthesizes heterogeneous sets of training image patches, of every tissue type. Although our synthetic patches are not always of high quality, we harness the motley crew of generated samples through a generally applicable importance sampling method. This proposed approach, for the first time, re-weighs the training loss over synthetic data so that the ideal (unbiased) generalization loss over the true data distribution is minimized. This enables us to use a random polygon generator to synthesize approximate cellular structures (i.e., nuclear masks) for which no real examples are given in many tissue types, and hence, GAN-based methods are not suited. In addition, we propose a hybrid synthesis pipeline that utilizes textures in real histopathology patches and GAN models, to tackle heterogeneity in tissue textures. Compared with existing state-of-the-art supervised models, our approach generalizes significantly better on cancer types without training data. Even in cancer types with training data, our approach achieves the same performance without supervision cost. We release code and segmentation results on over 5000 Whole Slide Images (WSI) in The Cancer Genome Atlas (TCGA) repository, a dataset that would be orders of magnitude larger than what is available today.
Detection, segmentation and classification of nuclei are fundamental analysis operations in digital pathology. Existing state-of-the-art approaches demand extensive amount of supervised training data from pathologists and may still perform poorly in images from unseen tissue types. We propose an unsupervised approach for histopathology image segmentation that synthesizes heterogeneous sets of training image patches, of every tissue type. Although our synthetic patches are not always of high quality, we harness the motley crew of generated samples through a generally applicable importance sampling method. This proposed approach, for the first time, re-weighs the training loss over synthetic data so that the ideal (unbiased) generalization loss over the true data distribution is minimized. This enables us to use a random polygon generator to synthesize approximate cellular structures (i.e., nuclear masks) for which no real examples are given in many tissue types, and hence, GAN-based methods are not suited. In addition, we propose a hybrid synthesis pipeline that utilizes textures in real histopathology patches and GAN models, to tackle heterogeneity in tissue textures. Compared with existing state-of-the-art supervised models, our approach generalizes significantly better on cancer types without training data. Even in cancer types with training data, our approach achieves the same performance without supervision cost. We release code and segmentation results on over 5000 Whole Slide Images (WSI) in The Cancer Genome Atlas (TCGA) repository, a dataset that would be orders of magnitude larger than what is available today.Detection, segmentation and classification of nuclei are fundamental analysis operations in digital pathology. Existing state-of-the-art approaches demand extensive amount of supervised training data from pathologists and may still perform poorly in images from unseen tissue types. We propose an unsupervised approach for histopathology image segmentation that synthesizes heterogeneous sets of training image patches, of every tissue type. Although our synthetic patches are not always of high quality, we harness the motley crew of generated samples through a generally applicable importance sampling method. This proposed approach, for the first time, re-weighs the training loss over synthetic data so that the ideal (unbiased) generalization loss over the true data distribution is minimized. This enables us to use a random polygon generator to synthesize approximate cellular structures (i.e., nuclear masks) for which no real examples are given in many tissue types, and hence, GAN-based methods are not suited. In addition, we propose a hybrid synthesis pipeline that utilizes textures in real histopathology patches and GAN models, to tackle heterogeneity in tissue textures. Compared with existing state-of-the-art supervised models, our approach generalizes significantly better on cancer types without training data. Even in cancer types with training data, our approach achieves the same performance without supervision cost. We release code and segmentation results on over 5000 Whole Slide Images (WSI) in The Cancer Genome Atlas (TCGA) repository, a dataset that would be orders of magnitude larger than what is available today.
Detection, segmentation and classification of nuclei are fundamental analysis operations in digital pathology. Existing state-of-the-art approaches demand extensive amount of supervised training data from pathologists and may still perform poorly in images from unseen tissue types. We propose an unsupervised approach for histopathology image segmentation that synthesizes heterogeneous sets of training image patches, of every tissue type. Although our synthetic patches are not always of high quality, we harness the motley crew of generated samples through a generally applicable importance sampling method. This proposed approach, for the first time, re-weighs the training loss over synthetic data so that the ideal (unbiased) generalization loss over the true data distribution is minimized. This enables us to use a random polygon generator to synthesize approximate cellular structures (i.e., nuclear masks) for which no real examples are given in many tissue types, and hence, GAN-based methods are not suited. In addition, we propose a hybrid synthesis pipeline that utilizes textures in real histopathology patches and GAN models, to tackle heterogeneity in tissue textures. Compared with existing state-of-the-art supervised models, our approach generalizes significantly better on cancer types without training data. Even in cancer types with training data, our approach achieves the same performance without supervision cost. We release code and segmentation results1 on over 5000 Whole Slide Images (WSI) in The Cancer Genome Atlas (TCGA) repository, a dataset that would be orders of magnitude larger than what is available today.
Author Agarwal, Ayush
Samaras, Dimitris
Saltz, Joel H.
Hou, Le
Kurc, Tahsin M.
Gupta, Rajarsi R.
AuthorAffiliation 2 Stanford University, California
1 Stony Brook University
AuthorAffiliation_xml – name: 2 Stanford University, California
– name: 1 Stony Brook University
Author_xml – sequence: 1
  givenname: Le
  surname: Hou
  fullname: Hou, Le
  organization: Stony Brook Univ
– sequence: 2
  givenname: Ayush
  surname: Agarwal
  fullname: Agarwal, Ayush
  organization: Stanford Univ
– sequence: 3
  givenname: Dimitris
  surname: Samaras
  fullname: Samaras, Dimitris
  organization: Stony Brook Univ
– sequence: 4
  givenname: Tahsin M.
  surname: Kurc
  fullname: Kurc, Tahsin M.
  organization: Stony Brook Univ
– sequence: 5
  givenname: Rajarsi R.
  surname: Gupta
  fullname: Gupta, Rajarsi R.
  organization: Stony Brook Univ
– sequence: 6
  givenname: Joel H.
  surname: Saltz
  fullname: Saltz, Joel H.
  organization: Stony Brook Univ
BackLink https://www.ncbi.nlm.nih.gov/pubmed/34025103$$D View this record in MEDLINE/PubMed
BookMark eNpVkEFLw0AUhFep2Fp79iBIjl5S9-1LNrse1FLUFgpKrV7DJtm0K2m2ZlOh_noDraWe3vDmYwbmjLRKW2pCLoD2Aai8GX68TvuMguxTKiI8Ij0ZCYiYAGQSxTHpAOXocwmydaDbpOfcJ6UUGQCX4pS0MaAsBIod8jC1ydrV3si42q5UvbCFnW-88VLNtTcoVbFxxt16M-tNVKILz1Zebb23TVkvtDM_-v6cnOSqcLq3u13y_vQ4G478ycvzeDiY-AbDsPZTiLjKIg4Z6oDrDCWgFBKESNKMyVyqTKicI00jEUiRiiyPWPNnmHPKgGOX3G1zV-tkqbNUl3WlinhVmaWqNrFVJv7vlGYRz-133KwjA4pNwPUuoLJfa-3qeGlcqotCldquXcxChDBgggcNenXYtS_5m60BLreA0VrvbSFDDCDEXyt0fP0
ContentType Conference Proceeding
Journal Article
DBID 6IE
6IH
CBEJK
RIE
RIO
NPM
7X8
5PM
DOI 10.1109/CVPR.2019.00873
DatabaseName IEEE Electronic Library (IEL) Conference Proceedings
IEEE Proceedings Order Plan (POP) 1998-present by volume
IEEE Xplore All Conference Proceedings
IEEE Xplore
IEEE Proceedings Order Plans (POP) 1998-present
PubMed
MEDLINE - Academic
PubMed Central (Full Participant titles)
DatabaseTitle PubMed
MEDLINE - Academic
DatabaseTitleList
MEDLINE - Academic

PubMed
Database_xml – sequence: 1
  dbid: NPM
  name: PubMed
  url: http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed
  sourceTypes: Index Database
– sequence: 2
  dbid: RIE
  name: IEEE Xplore
  url: https://ieeexplore.ieee.org/
  sourceTypes: Publisher
– sequence: 3
  dbid: 7X8
  name: MEDLINE - Academic
  url: https://search.proquest.com/medline
  sourceTypes: Aggregation Database
DeliveryMethod fulltext_linktorsrc
Discipline Applied Sciences
Computer Science
EISBN 9781728132938
1728132932
EISSN 1063-6919
EndPage 8534
ExternalDocumentID PMC8139403
34025103
8953415
Genre orig-research
Journal Article
GrantInformation_xml – fundername: NLM NIH HHS
  grantid: R01 LM011119
– fundername: NLM NIH HHS
  grantid: R01 LM009239
– fundername: NCI NIH HHS
  grantid: U24 CA180924
– fundername: NCI NIH HHS
  grantid: U24 CA215109
– fundername: NCI NIH HHS
  grantid: UG3 CA225021
GroupedDBID 6IE
6IH
6IL
6IN
AAWTH
ABLEC
ADZIZ
ALMA_UNASSIGNED_HOLDINGS
BEFXN
BFFAM
BGNUA
BKEBE
BPEOZ
CBEJK
CHZPO
IEGSK
IJVOP
OCL
RIE
RIL
RIO
23M
29F
29O
6IK
ABDPE
ACGFS
IPLJI
M43
NPM
RIG
RNS
7X8
5PM
ID FETCH-LOGICAL-i355t-c176ad761d3e46ed3913989188bcd29f9ad8af630c78498c8df7229f23f602163
IEDL.DBID RIE
ISICitedReferencesCount 95
ISICitedReferencesURI http://www.webofscience.com/api/gateway?GWVersion=2&SrcApp=Summon&SrcAuth=ProQuest&DestLinkType=CitingArticles&DestApp=WOS_CPL&KeyUT=000542649302015&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D
ISSN 1063-6919
IngestDate Tue Sep 30 16:36:22 EDT 2025
Fri Sep 05 06:02:37 EDT 2025
Wed Feb 19 02:06:12 EST 2025
Wed Aug 27 02:30:20 EDT 2025
IsDoiOpenAccess false
IsOpenAccess true
IsPeerReviewed false
IsScholarly true
Language English
LinkModel DirectLink
MergedId FETCHMERGED-LOGICAL-i355t-c176ad761d3e46ed3913989188bcd29f9ad8af630c78498c8df7229f23f602163
Notes ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 23
OpenAccessLink https://www.ncbi.nlm.nih.gov/pmc/articles/8139403
PMID 34025103
PQID 2531542864
PQPubID 23479
PageCount 10
ParticipantIDs ieee_primary_8953415
proquest_miscellaneous_2531542864
pubmedcentral_primary_oai_pubmedcentral_nih_gov_8139403
pubmed_primary_34025103
PublicationCentury 2000
PublicationDate 2019-06-01
PublicationDateYYYYMMDD 2019-06-01
PublicationDate_xml – month: 06
  year: 2019
  text: 2019-06-01
  day: 01
PublicationDecade 2010
PublicationPlace United States
PublicationPlace_xml – name: United States
PublicationTitle Proceedings (IEEE Computer Society Conference on Computer Vision and Pattern Recognition. Online)
PublicationTitleAbbrev CVPR
PublicationTitleAlternate Proc IEEE Comput Soc Conf Comput Vis Pattern Recognit
PublicationYear 2019
Publisher IEEE
Publisher_xml – name: IEEE
SSID ssj0003211698
ssj0023720
Score 2.5747204
Snippet Detection, segmentation and classification of nuclei are fundamental analysis operations in digital pathology. Existing state-of-the-art approaches demand...
SourceID pubmedcentral
proquest
pubmed
ieee
SourceType Open Access Repository
Aggregation Database
Index Database
Publisher
StartPage 8525
SubjectTerms Biological and Cell Microscopy
Grouping and Shape; Vision Applications and Systems
Medical
Segmentation
Title Robust Histopathology Image Analysis: To Label or to Synthesize?
URI https://ieeexplore.ieee.org/document/8953415
https://www.ncbi.nlm.nih.gov/pubmed/34025103
https://www.proquest.com/docview/2531542864
https://pubmed.ncbi.nlm.nih.gov/PMC8139403
Volume 2019
WOSCitedRecordID wos000542649302015&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D
hasFullText 1
inHoldings 1
isFullTextHit
isPrint
link http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwlV1LS8NAEB60ePDkW-ujrODRaJNN9uFFQSwKIkWr9BY2m1ksaCJNKuivdzdN44NevIW8WGZmd2ZnZ74P4Cjhyrd-XXpSJXaDQo2dc44yTKtQI2OBDGqyCX53J4ZD2V-A46YXBhGr4jM8cZfVWX6a64lLlZ0KGdGqo3yRcz7t1WryKdT-lElRo_f4XXl6-dS_d7VbDpBScEeYQ0MXUTt-rIpJZV5Q-bc28oez6a38b5irsPndtUf6jT9agwXM1mGlDjNJPYmLDbi4z5NJUZIKIcQxEleZdXLzapcWMgMpOSODnNyqBF9IPiZlTh4-MhsrFqNPPN-Ex97V4PLaq4kUvJENJ0pP-5yplDM_pRgyTKnDAhXSFyLRaSCNVKlQhtGu5iKUQovU8MDeD6hhNgZgdAtaWZ7hDhCNKrJqF12GIlShEREmSWBUF400JsI2bDiBxG9TrIy4lkUbDmeijq39ukMJlWE-KeLALgKR3QOxsA3bU9E3H8_01Qb-SynNCw4b-_eTbPRcYWQL31G-0935w9mDZWcY05KvfWiV4wkewJJ-L0fFuGPNayg6lXl9AUAZz3g
linkProvider IEEE
linkToHtml http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwlV1LS8QwEB5EBT2tb9dnBI-ubps0TbwoiKK4Louu4q2k6QQXtJVtV9Bfb9Lt1gdevJW-CDOTzGQy830A-3GoPOvXZUuq2G5QqLFzzlGGacU0cu5LvyKbCLtd8fgoe1NwUPfCIGJZfIaH7rI8y08yPXKpsiMhA1p2lM8EjPneuFurzqhQ-1suRYXf47Xl0dlD79ZVbzlIShE6yhzKXEztGLJKLpW_wsrf1ZHf3M1F438DXYCVr7490qs90iJMYboEjSrQJNU0zpfh9DaLR3lBSowQx0lc5tbJ1YtdXMgEpuSY9DPSUTE-k2xIiozcvac2WswHH3iyAvcX5_2zy1ZFpdAa2ICiaGkv5CoJuZdQZBwT6tBAhfSEiHXiSyNVIpThtK1DwaTQIjGhb-_71HAbBXC6CtNpluI6EI0qsIoXbY6CKWZEgHHsG9VGI40JsAnLTiDR6xgtI6pk0YS9iagja8HuWEKlmI3yyLfLQGB3QZw1YW0s-vrjib6aEP5QSv2CQ8f--SQdPJUo2cJzpO904-_h7MLcZf-mE3WuutebMO-MZFwAtgXTxXCE2zCr34pBPtwpjewTSTPR1w
openUrl ctx_ver=Z39.88-2004&ctx_enc=info%3Aofi%2Fenc%3AUTF-8&rfr_id=info%3Asid%2Fsummon.serialssolutions.com&rft_val_fmt=info%3Aofi%2Ffmt%3Akev%3Amtx%3Abook&rft.genre=proceeding&rft.title=Proceedings+%28IEEE+Computer+Society+Conference+on+Computer+Vision+and+Pattern+Recognition.+Online%29&rft.atitle=Robust+Histopathology+Image+Analysis%3A+To+Label+or+to+Synthesize%3F&rft.au=Hou%2C+Le&rft.au=Agarwal%2C+Ayush&rft.au=Samaras%2C+Dimitris&rft.au=Kurc%2C+Tahsin+M.&rft.date=2019-06-01&rft.pub=IEEE&rft.eissn=1063-6919&rft.spage=8525&rft.epage=8534&rft_id=info:doi/10.1109%2FCVPR.2019.00873&rft_id=info%3Apmid%2F34025103&rft.externalDocID=8953415
thumbnail_l http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/lc.gif&issn=1063-6919&client=summon
thumbnail_m http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/mc.gif&issn=1063-6919&client=summon
thumbnail_s http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/sc.gif&issn=1063-6919&client=summon