Domain-adversarial neural networks to address the appearance variability of histopathology images

Research output: Chapter in Book/Report/Conference proceedingConference contributionAcademicpeer-review

18 Citations (Scopus)

Abstract

Preparing and scanning histopathology slides consists of several steps, each with a multitude of parameters. The parameters can vary between pathology labs and within the same lab over time, resulting in significant variability of the tissue appearance that hampers the generalization of automatic image analysis methods. Typically, this is addressed with ad-hoc approaches such as staining normalization that aim to reduce the appearance variability. In this paper, we propose a systematic solution based on domain-adversarial neural networks. We hypothesize that removing the domain information from the model representation leads to better generalization. We tested our hypothesis for the problem of mitosis detection in breast cancer histopathology images and made a comparative analysis with two other approaches. We show that combining color augmentation with domain-adversarial training is a better alternative than standard approaches to improve the generalization of deep learning methods.

LanguageEnglish
Title of host publicationDeep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support - 3rd International Workshop, DLMIA 2017 and 7th International Workshop, ML-CDS 2017 Held in Conjunction with MICCAI 2017, Proceedings
PublisherSpringer
Pages83-91
Number of pages9
ISBN (Print)9783319675572
DOIs
StatePublished - 2017
Event3rd International Workshop on Deep Learning in Medical Image Analysis, DLMIA 2017 and 7th International Workshop on Multimodal Learning for Clinical Decision Support, ML-CDS 2017 held in Conjunction with 20th International Conference on Medical Image Computing and Computer Assisted Intervention, MICCAI 2017, 10-14 September Quebec, Canada - Quebec City, Canada
Duration: 10 Sep 201714 Sep 2017
http://www.miccai2017.org/

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume10553 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference3rd International Workshop on Deep Learning in Medical Image Analysis, DLMIA 2017 and 7th International Workshop on Multimodal Learning for Clinical Decision Support, ML-CDS 2017 held in Conjunction with 20th International Conference on Medical Image Computing and Computer Assisted Intervention, MICCAI 2017, 10-14 September Quebec, Canada
Abbreviated titleDLMIA2017
CountryCanada
CityQuebec City
Period10/09/1714/09/17
Internet address

Fingerprint

Pathology
Image analysis
Neural Networks
Tissue
Color
Neural networks
Scanning
Augmentation
Breast Cancer
Comparative Analysis
Image Analysis
Normalization
Vary
Alternatives
Deep learning
Generalization
Model

Keywords

  • Domain-adversarial training
  • Histopathology image analysis

Cite this

Lafarge, M. W., Pluim, J. P. W., Eppenhof, K. A. J., Moeskops, P., & Veta, M. (2017). Domain-adversarial neural networks to address the appearance variability of histopathology images. In Deep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support - 3rd International Workshop, DLMIA 2017 and 7th International Workshop, ML-CDS 2017 Held in Conjunction with MICCAI 2017, Proceedings (pp. 83-91). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 10553 LNCS). Springer. DOI: 10.1007/978-3-319-67558-9_10
Lafarge, M.W. ; Pluim, J.P.W. ; Eppenhof, K.A.J. ; Moeskops, P. ; Veta, M./ Domain-adversarial neural networks to address the appearance variability of histopathology images. Deep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support - 3rd International Workshop, DLMIA 2017 and 7th International Workshop, ML-CDS 2017 Held in Conjunction with MICCAI 2017, Proceedings. Springer, 2017. pp. 83-91 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{7293afeac1d148018c1ef2e4b35f3597,
title = "Domain-adversarial neural networks to address the appearance variability of histopathology images",
abstract = "Preparing and scanning histopathology slides consists of several steps, each with a multitude of parameters. The parameters can vary between pathology labs and within the same lab over time, resulting in significant variability of the tissue appearance that hampers the generalization of automatic image analysis methods. Typically, this is addressed with ad-hoc approaches such as staining normalization that aim to reduce the appearance variability. In this paper, we propose a systematic solution based on domain-adversarial neural networks. We hypothesize that removing the domain information from the model representation leads to better generalization. We tested our hypothesis for the problem of mitosis detection in breast cancer histopathology images and made a comparative analysis with two other approaches. We show that combining color augmentation with domain-adversarial training is a better alternative than standard approaches to improve the generalization of deep learning methods.",
keywords = "Domain-adversarial training, Histopathology image analysis",
author = "M.W. Lafarge and J.P.W. Pluim and K.A.J. Eppenhof and P. Moeskops and M. Veta",
year = "2017",
doi = "10.1007/978-3-319-67558-9_10",
language = "English",
isbn = "9783319675572",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
publisher = "Springer",
pages = "83--91",
booktitle = "Deep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support - 3rd International Workshop, DLMIA 2017 and 7th International Workshop, ML-CDS 2017 Held in Conjunction with MICCAI 2017, Proceedings",
address = "Germany",

}

Lafarge, MW, Pluim, JPW, Eppenhof, KAJ, Moeskops, P & Veta, M 2017, Domain-adversarial neural networks to address the appearance variability of histopathology images. in Deep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support - 3rd International Workshop, DLMIA 2017 and 7th International Workshop, ML-CDS 2017 Held in Conjunction with MICCAI 2017, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 10553 LNCS, Springer, pp. 83-91, 3rd International Workshop on Deep Learning in Medical Image Analysis, DLMIA 2017 and 7th International Workshop on Multimodal Learning for Clinical Decision Support, ML-CDS 2017 held in Conjunction with 20th International Conference on Medical Image Computing and Computer Assisted Intervention, MICCAI 2017, 10-14 September Quebec, Canada, Quebec City, Canada, 10/09/17. DOI: 10.1007/978-3-319-67558-9_10

Domain-adversarial neural networks to address the appearance variability of histopathology images. / Lafarge, M.W.; Pluim, J.P.W.; Eppenhof, K.A.J.; Moeskops, P.; Veta, M.

Deep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support - 3rd International Workshop, DLMIA 2017 and 7th International Workshop, ML-CDS 2017 Held in Conjunction with MICCAI 2017, Proceedings. Springer, 2017. p. 83-91 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 10553 LNCS).

Research output: Chapter in Book/Report/Conference proceedingConference contributionAcademicpeer-review

TY - GEN

T1 - Domain-adversarial neural networks to address the appearance variability of histopathology images

AU - Lafarge,M.W.

AU - Pluim,J.P.W.

AU - Eppenhof,K.A.J.

AU - Moeskops,P.

AU - Veta,M.

PY - 2017

Y1 - 2017

N2 - Preparing and scanning histopathology slides consists of several steps, each with a multitude of parameters. The parameters can vary between pathology labs and within the same lab over time, resulting in significant variability of the tissue appearance that hampers the generalization of automatic image analysis methods. Typically, this is addressed with ad-hoc approaches such as staining normalization that aim to reduce the appearance variability. In this paper, we propose a systematic solution based on domain-adversarial neural networks. We hypothesize that removing the domain information from the model representation leads to better generalization. We tested our hypothesis for the problem of mitosis detection in breast cancer histopathology images and made a comparative analysis with two other approaches. We show that combining color augmentation with domain-adversarial training is a better alternative than standard approaches to improve the generalization of deep learning methods.

AB - Preparing and scanning histopathology slides consists of several steps, each with a multitude of parameters. The parameters can vary between pathology labs and within the same lab over time, resulting in significant variability of the tissue appearance that hampers the generalization of automatic image analysis methods. Typically, this is addressed with ad-hoc approaches such as staining normalization that aim to reduce the appearance variability. In this paper, we propose a systematic solution based on domain-adversarial neural networks. We hypothesize that removing the domain information from the model representation leads to better generalization. We tested our hypothesis for the problem of mitosis detection in breast cancer histopathology images and made a comparative analysis with two other approaches. We show that combining color augmentation with domain-adversarial training is a better alternative than standard approaches to improve the generalization of deep learning methods.

KW - Domain-adversarial training

KW - Histopathology image analysis

UR - http://www.scopus.com/inward/record.url?scp=85029818275&partnerID=8YFLogxK

U2 - 10.1007/978-3-319-67558-9_10

DO - 10.1007/978-3-319-67558-9_10

M3 - Conference contribution

SN - 9783319675572

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 83

EP - 91

BT - Deep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support - 3rd International Workshop, DLMIA 2017 and 7th International Workshop, ML-CDS 2017 Held in Conjunction with MICCAI 2017, Proceedings

PB - Springer

ER -

Lafarge MW, Pluim JPW, Eppenhof KAJ, Moeskops P, Veta M. Domain-adversarial neural networks to address the appearance variability of histopathology images. In Deep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support - 3rd International Workshop, DLMIA 2017 and 7th International Workshop, ML-CDS 2017 Held in Conjunction with MICCAI 2017, Proceedings. Springer. 2017. p. 83-91. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). Available from, DOI: 10.1007/978-3-319-67558-9_10