2018 · 134 citations · 21 references
ComputingComputer SciencePathologyComputer VisionHistopathologyMachine LearningDeep LearningGenerative ModelingHistological StainingUnsupervised LearningColor NormalizationDeconvolutionGenerative Adversarial NetworksNonlinear Source TransformationNormalizationPrior Distribution SpecificationStain DeconvolutionTemplate-based Matching
Computational histopathology relies on computer‑aided diagnosis of stained slides, and stain color normalization—typically performed via stain deconvolution or template matching—is a critical preprocessing step. This study proposes a fully unsupervised, parametric generative model to solve the stain‑normalization problem. The model, built on a generative adversarial network, learns a nonlinear mapping of latent variables with a Dirichlet prior, enabling a source image’s latent code to be replaced by that of a template to produce a color‑normalized copy that preserves tissue structure and can be applied instantly to unseen images. Experiments on H&E images from multiple laboratories show that the proposed method outperforms most state‑of‑the‑art stain‑normalization techniques.
Computational histopathology involves CAD for microscopic analysis of stained histopathological slides to study presence, localization or grading of disease. An important stage in a CAD system, stain color normalization, has been broadly studied. The existing approaches are mainly defined in the context of stain deconvolution and template matching. In this paper, we propose a novel approach to this problem by introducing a parametric, fully unsupervised generative model. Our model is based on end-to-end machine learning in the framework of generative adversarial networks. It can learn a nonlinear transformation of a set of latent variables, which are forced to have a prior Dirichlet distribution and control the color of staining hematoxylin and eosin (H&E) images. By replacing the latent variables of a source image with those extracted from a template image in the trained model, it can generate a new color copy of the source image while preserving the important tissue structures resembling the chromatic information of the template image. Our proposed method can instantly be applied to new unseen images, which is different from previous methods that need to compute some statistical properties on input test data. This is potentially problematic when the test sample sizes are limited. Experiments on H&E images from different laboratories show that the proposed model outperforms most state-of-the-art methods.
21
Fundamentals of digital image processing
Computer Vision Graphics and Image Processing · 1989 · 4.7K citations
Erik Reinhard, M. Adhikhmin, Bryan N S Gooch et al. · IEEE Computer Graphics and Applications · 2001 · 3K citations
Arts, Vision Science, Color Theory +10
Quantification of histochemical staining by color deconvolution.
Arnout Ruifrok, D. A. Johnston · PubMed · 2001 · 2.1K citations
Medicine, Cell Biology, Pathology +17
A method for normalizing histology slides for quantitative analysis
Marc Macenko, Marc Niethammer, J. S. Marron et al. · 2009 · 1.3K citations
Medicine, Biostatistics, Pathology +13