Color standardization for digitized histological images

Patent metadata

Field	Value
Publication number	US-2016307305-A1
Application number	US-201415030972-A
Country	US
Kind code	A1
Filing date	Oct 23, 2014
Priority date	Oct 23, 2013
Publication date	Oct 20, 2016
Grant date	—

Abstract

A system is provided for standardizing digital histological images so that the color space for a histological image correlates with the color space of a template image of the histological image. The image data for the image is segmented into a plurality of subsets that correspond to different tissue classes in the image. The image data for each subset is then compared with a corresponding subset in the template image. Based on the comparison, the color channels for the histological image subsets are shifted to create a series of standardized subsets, which are then combined to create a standardized image.

First claim

1 . A method for processing histological images to improve color consistency, comprising the steps of:
providing image data for a histological image; selecting a template image comprising image data corresponding to tissue in the histological image, wherein the template comprises a plurality of data subsets corresponding to different tissue classes in the template; segmenting the image data for the histological image into a plurality of subsets, wherein the subsets correspond to different tissue classes; constructing a histogram for each data subset of the template and constructing a histogram for the corresponding subset of the image data for the histological image; aligning the histogram for each subset of the image data with the histogram of corresponding data subset of the template to create a series of standardized subsets of the image data; and combining standardized subsets of the image data to create a standardized histological image.

2 . The method of claim 1 wherein each subset of image data is divided into a plurality of color channels, wherein the step of constructing a histogram for each data subset comprises constructing a histogram for each color channel of each data subset of the template and constructing a histogram for the corresponding color channel of each subset of the image data for the histological image.

3 . The method of claim 1 wherein the step of segmenting the image data for the histological image into a plurality of subsets comprises segmenting the image data using an expectation-maximization algorithm.

4 . The method of claim 1 comprising the step of automatically segmenting the template into the plurality of data subsets by training an autoencoder to identify a plurality of tissue classes in a histological image.

5 . (canceled)

6 . The method of claim 4 wherein the step of automatically segmenting the template comprises training unsupervised deep learning filters using randomly selected subsets of the template image data.

7 . The method of claim 6 wherein the step of training deep learning filters comprises training deep sparse autoencoders on the randomly selected subsets.

8 . The method of claim 4 comprising the step of randomly selecting a plurality of subsets of image data from the template and using the subsets of image data during the step of training.

9 - 12 . (canceled)

13 . The method of claim 1 wherein the step of segmenting the image data for the histological image comprises the step of employing a standard k-means approach to identify a plurality of clusters centers.

14 . The method of claim 13 wherein the step of segmenting comprises assigning image data into subsets based on the relation of the data to the cluster centers.

15 . The method of claim 1 wherein the image data for the histological image is a two-dimensional set of pixels having color values in the Red, Green, Blue color space.

16 . A method for processing histological images to improve color consistency, comprising the steps of:
providing image data for a histological image; selecting a template corresponding to the histological image, wherein the template comprises a plurality of data subsets corresponding to different tissue classes in the template and each data subset is divided into a plurality of color channels; segmenting the image data for the histological image into a plurality of subsets, wherein the subsets correspond to different tissue classes and each subset of image data is divided into a plurality of color channels; comparing the histological image data of each color channel in a subset with the corresponding data subset of the corresponding color channel for the template; selectively varying the histological image data of each color channel in a subset in response to the step of comparing to create a series of standardized subsets of the image data; and combining standardized subsets of the image data to create a standardized histological image.

17 . The method of claim 16 wherein each subset of image data is divided into a plurality of color channels, wherein the step of constructing a histogram for each data subset comprises constructing a histogram for each color channel of each data subset of the template and constructing a histogram for the corresponding color channel of each subset of the image data for the histological image.

18 . The method of claim 16 wherein the step of segmenting the image data for the histological image into a plurality of subsets comprises segmenting the image data using an expectation-maximization algorithm.

19 . The method of any of claims 16 comprising the step of automatically segmenting the template into the plurality of data subsets by training an autoencoder to identify a plurality of tissue classes in a histological image.

20 . (canceled)

21 . The method of claim 18 wherein the step of automatically segmenting the template comprises training unsupervised deep learning filters using randomly selected subsets of the template image data.

22 . The method of claim 21 wherein the step of training deep learning filters comprises training deep sparse autoencoders on the randomly selected subsets.

23 . The method of claim 21 comprising the step of randomly selecting a plurality of subsets of image data from the template and using the subsets of image data during the step of training.

24 - 27 . (canceled)

28 . The method of claim 16 wherein the step of segmenting the image data for the histological image comprises the step of employing a standard k-means approach to identify a plurality of clusters centers.

29 . The method of claim 28 wherein the step of segmenting comprises assigning image data into subsets based on the relation of the data to the cluster centers.

30 . (canceled)

31 . A method for processing histological images to improve color consistency, comprising the steps of:
selecting a template histological image, wherein the template comprises a plurality of data subsets corresponding to different tissue classes in the template and each data subset is divided into a plurality of color channels; randomly selecting a number of the data subsets; training unsupervised deep learning filters on the randomly selected subsets; applying the deep learning filters to a histological image to produce a set of filtered image data; segmenting the filtered image data into a plurality of subsets; comparing the filtered image data subsets with the corresponding data subset for the template; selectively varying the histological image data of each color channel in a subset in response to the step of comparing to create a series of standardized subsets of the image data; and combining standardized subsets of the image data to create a standardized histological image.

32 . The method of claim 31 wherein the step of segmenting comprises the step of employing a standard k-means approach to identify a plurality of clusters centers.

33 . The method of claim 32 wherein the step of segmenting comprises assigning image data into subsets based on the relation of the data to the cluster centers.

34 . (canceled)

35 . The method of claim 31 wherein the step of training deep learning filters comprises training deep sparse autoencoders on the randomly selected subsets.

36 . The method of claim 31 comprising the ste

Assignees

• Univ Rutgers

Inventors

• Madabhushi Anant
• Basavanhally Ajay
• Janowczyk Andrew

Classifications

• G06V10/80

  Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level (multimodal speaker identification or verification G10L17/10) · CPC title

• G06T5/40 Primary

  using histogram techniques · CPC title

• G06T11/10 Primary

  Texturing; Colouring; Generation of textures or colours (retouching, inpainting or scratch removal G06T5/77) · CPC title

• G06F18/23

  Clustering techniques · CPC title

• G06F18/25

  Fusion techniques · CPC title