Histopathology stain-color normalization using deep generative models

Farhad Ghazvinian Zanjani, Svitlana Zinger, P.H.N. de With, Babak E. Bejnordi, Jeroen A.W.M. van der Laak

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

245 Downloads (Pure)

Abstract

Performance of designed CAD algorithms for histopathology image analysis is affected by the amount of variations in the samples such as color and intensity of stained images. Stain-color normalization is a well-studied technique for compensating such effects at the input of CAD systems. In this paper, we introduce unsupervised generative neural networks for performing stain-color normalization. For color normalization in stained hematoxylin and eosin (H&E) images, we present three methods based on three frameworks for deep generative models: variational auto-encoder (VAE), generative adversarial networks (GAN) and deep convolutional Gaussian mixture models (DCGMM). Our contribution is defining the color normalization as a learning generative model that is able to generate various color copies of the input image through a nonlinear parametric transformation. In contrast to earlier generative models proposed for stain-color normalization, our approach does not need any labels for data or any other assumptions about the H&E image content. Furthermore, our models learn a parametric transformation during training and can convert the color information of an input image to resemble any arbitrary reference image. This property is essential in time-critical CAD systems in case of changing the reference image, since our approach does not need retraining in contrast to other proposed generative models for stain-color normalization. Experiments on histopathological H&E images with high staining variations, collected from different laboratories, show that our proposed models outperform quantitatively state-of-the-art methods in the measure of color constancy with at least 10-15%, while the converted images are visually in agreement with this performance improvement.
Original languageEnglish
Title of host publication1st Conference on Medical Imaging with Deep Learning (MIDL 2018)
Pages1-11
Number of pages11
Publication statusPublished - 4 Jul 2018
Event1st Conference on Medical Imaging with Deep Learning (MIDL 2018) - Amsterdam, Netherlands
Duration: 4 Jul 20186 Jul 2018
https://midl.amsterdam

Conference

Conference1st Conference on Medical Imaging with Deep Learning (MIDL 2018)
Abbreviated titleMIDL 2018
CountryNetherlands
CityAmsterdam
Period4/07/186/07/18
Internet address

Fingerprint

Color
Computer aided design
Image analysis
Labels
Neural networks
Experiments

Cite this

Ghazvinian Zanjani, F., Zinger, S., de With, P. H. N., E. Bejnordi, B., & van der Laak, J. A. W. M. (2018). Histopathology stain-color normalization using deep generative models. In 1st Conference on Medical Imaging with Deep Learning (MIDL 2018) (pp. 1-11)
Ghazvinian Zanjani, Farhad ; Zinger, Svitlana ; de With, P.H.N. ; E. Bejnordi, Babak ; van der Laak, Jeroen A.W.M. / Histopathology stain-color normalization using deep generative models. 1st Conference on Medical Imaging with Deep Learning (MIDL 2018). 2018. pp. 1-11
@inproceedings{8ba7339df0cc4f49901a728fe23aa1b3,
title = "Histopathology stain-color normalization using deep generative models",
abstract = "Performance of designed CAD algorithms for histopathology image analysis is affected by the amount of variations in the samples such as color and intensity of stained images. Stain-color normalization is a well-studied technique for compensating such effects at the input of CAD systems. In this paper, we introduce unsupervised generative neural networks for performing stain-color normalization. For color normalization in stained hematoxylin and eosin (H&E) images, we present three methods based on three frameworks for deep generative models: variational auto-encoder (VAE), generative adversarial networks (GAN) and deep convolutional Gaussian mixture models (DCGMM). Our contribution is defining the color normalization as a learning generative model that is able to generate various color copies of the input image through a nonlinear parametric transformation. In contrast to earlier generative models proposed for stain-color normalization, our approach does not need any labels for data or any other assumptions about the H&E image content. Furthermore, our models learn a parametric transformation during training and can convert the color information of an input image to resemble any arbitrary reference image. This property is essential in time-critical CAD systems in case of changing the reference image, since our approach does not need retraining in contrast to other proposed generative models for stain-color normalization. Experiments on histopathological H&E images with high staining variations, collected from different laboratories, show that our proposed models outperform quantitatively state-of-the-art methods in the measure of color constancy with at least 10-15{\%}, while the converted images are visually in agreement with this performance improvement.",
author = "{Ghazvinian Zanjani}, Farhad and Svitlana Zinger and {de With}, P.H.N. and {E. Bejnordi}, Babak and {van der Laak}, {Jeroen A.W.M.}",
year = "2018",
month = "7",
day = "4",
language = "English",
pages = "1--11",
booktitle = "1st Conference on Medical Imaging with Deep Learning (MIDL 2018)",

}

Ghazvinian Zanjani, F, Zinger, S, de With, PHN, E. Bejnordi, B & van der Laak, JAWM 2018, Histopathology stain-color normalization using deep generative models. in 1st Conference on Medical Imaging with Deep Learning (MIDL 2018). pp. 1-11, 1st Conference on Medical Imaging with Deep Learning (MIDL 2018), Amsterdam, Netherlands, 4/07/18.

Histopathology stain-color normalization using deep generative models. / Ghazvinian Zanjani, Farhad; Zinger, Svitlana; de With, P.H.N.; E. Bejnordi, Babak; van der Laak, Jeroen A.W.M.

1st Conference on Medical Imaging with Deep Learning (MIDL 2018). 2018. p. 1-11.

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

TY - GEN

T1 - Histopathology stain-color normalization using deep generative models

AU - Ghazvinian Zanjani, Farhad

AU - Zinger, Svitlana

AU - de With, P.H.N.

AU - E. Bejnordi, Babak

AU - van der Laak, Jeroen A.W.M.

PY - 2018/7/4

Y1 - 2018/7/4

N2 - Performance of designed CAD algorithms for histopathology image analysis is affected by the amount of variations in the samples such as color and intensity of stained images. Stain-color normalization is a well-studied technique for compensating such effects at the input of CAD systems. In this paper, we introduce unsupervised generative neural networks for performing stain-color normalization. For color normalization in stained hematoxylin and eosin (H&E) images, we present three methods based on three frameworks for deep generative models: variational auto-encoder (VAE), generative adversarial networks (GAN) and deep convolutional Gaussian mixture models (DCGMM). Our contribution is defining the color normalization as a learning generative model that is able to generate various color copies of the input image through a nonlinear parametric transformation. In contrast to earlier generative models proposed for stain-color normalization, our approach does not need any labels for data or any other assumptions about the H&E image content. Furthermore, our models learn a parametric transformation during training and can convert the color information of an input image to resemble any arbitrary reference image. This property is essential in time-critical CAD systems in case of changing the reference image, since our approach does not need retraining in contrast to other proposed generative models for stain-color normalization. Experiments on histopathological H&E images with high staining variations, collected from different laboratories, show that our proposed models outperform quantitatively state-of-the-art methods in the measure of color constancy with at least 10-15%, while the converted images are visually in agreement with this performance improvement.

AB - Performance of designed CAD algorithms for histopathology image analysis is affected by the amount of variations in the samples such as color and intensity of stained images. Stain-color normalization is a well-studied technique for compensating such effects at the input of CAD systems. In this paper, we introduce unsupervised generative neural networks for performing stain-color normalization. For color normalization in stained hematoxylin and eosin (H&E) images, we present three methods based on three frameworks for deep generative models: variational auto-encoder (VAE), generative adversarial networks (GAN) and deep convolutional Gaussian mixture models (DCGMM). Our contribution is defining the color normalization as a learning generative model that is able to generate various color copies of the input image through a nonlinear parametric transformation. In contrast to earlier generative models proposed for stain-color normalization, our approach does not need any labels for data or any other assumptions about the H&E image content. Furthermore, our models learn a parametric transformation during training and can convert the color information of an input image to resemble any arbitrary reference image. This property is essential in time-critical CAD systems in case of changing the reference image, since our approach does not need retraining in contrast to other proposed generative models for stain-color normalization. Experiments on histopathological H&E images with high staining variations, collected from different laboratories, show that our proposed models outperform quantitatively state-of-the-art methods in the measure of color constancy with at least 10-15%, while the converted images are visually in agreement with this performance improvement.

M3 - Conference contribution

SP - 1

EP - 11

BT - 1st Conference on Medical Imaging with Deep Learning (MIDL 2018)

ER -

Ghazvinian Zanjani F, Zinger S, de With PHN, E. Bejnordi B, van der Laak JAWM. Histopathology stain-color normalization using deep generative models. In 1st Conference on Medical Imaging with Deep Learning (MIDL 2018). 2018. p. 1-11