Automated epithelial nuclei segmentation for computational disease detection algorithms

Patent metadata

Field	Value
Publication number	US-9626583-B2
Application number	US-201414568900-A
Country	US
Kind code	B2
Filing date	Dec 12, 2014
Priority date	Dec 12, 2013
Publication date	Apr 18, 2017
Grant date	Apr 18, 2017

Abstract

In aspects, the subject innovation can comprise systems and methods capable of automatically labeling cell nuclei (e.g., epithelial nuclei) in tissue images containing multiple cell types. The enhancements to standard nuclei segmentation algorithms of the subject innovation can enable cell type specific analysis of nuclei, which has recently been shown to reveal novel disease biomarkers and improve diagnostic accuracy of computational disease classification models.

First claim

What is claimed is:

1. A method, comprising:
receiving in a computing device an image comprising a plurality of cells;
fitting in the computing device a Gaussian mixture model to an intensity distribution of the image, wherein the model comprises one or more Gaussian distributions and a background distribution;
identifying in the computing device a first Gaussian component G g associated with a first Gaussian distribution of the one or more Gaussian distributions corresponding to nuclei of the plurality of cells; and
defining in the computing device a nuclei mask as a binary matrix based at least in part on the first Gaussian component G g , wherein the nuclei mask comprises one or more putative nuclei associated with the plurality of cells.

2. The method of claim 1 , further comprising cleaning the nuclei mask by removing one or more of holes, isolated pixels, or bridge pixels.

3. The method of claim 1 , further comprising contrast normalizing the nuclei mask.

4. The method of claim 1 , further comprising removing thin lines of pixels included in the nuclei mask.

5. The method of claim 1 , further comprising breaking one or more large regions of the nuclei mask into individual nuclei.

6. The method of claim 1 , further comprising removing one or more very small regions from the nuclei mask.

7. The method of claim 1 , further comprising expanding at least one of the one or more putative nuclei via a watershed.

8. A method, comprising:
receiving in a computing device an image comprising a plurality of cells;
fitting in the computing device a Gaussian mixture model to an intensity distribution of the image, wherein the model comprises one or more Gaussian distributions and a background distribution, and wherein the one or more Gaussian distributions correspond to nuclei, cytoplasm, and stroma/lumen of the plurality of cells;
identifying in the computing device a first Gaussian component G g associated with a first Gaussian distribution of the one or more Gaussian distributions using intensities and region sizes of pixels described by one or more Gaussian components; and
defining in the computing device a nuclei mask as a binary matrix based at least in part on the first Gaussian component G g , wherein the nuclei mask comprises one or more putative nuclei associated with the plurality of cells.

9. The method of claim 8 , further comprising cleaning the nuclei mask by removing one or more of holes, isolated pixels, or bridge pixels in the image.

10. The method of claim 8 , further comprising contrast normalizing the nuclei mask.

11. The method of claim 8 , further comprising removing thin lines of pixels included in the nuclei mask.

12. The method of claim 8 , further comprising breaking one or more large regions of the nuclei mask into individual nuclei.

13. The method of claim 8 , further comprising removing one or more very small regions from the nuclei mask.

14. The method of claim 8 , wherein the first Gaussian component G g corresponds to the nuclei of the plurality of cells.

15. A method, comprising:
receiving in a computing device an image comprising a plurality of cells;
fitting in the computing device a Gaussian mixture model to an intensity distribution of the image, wherein the model comprises one or more Gaussian distributions and a background distribution;
identifying in the computing device a first Gaussian component G g associated with a first Gaussian distribution of the one or more Gaussian distributions corresponding to nuclei of the plurality of cells using intensities and region sizes of pixels described by one or more Gaussian components; and
defining in the computing device a nuclei mask as a binary matrix based at least in part on the first Gaussian component G g , wherein the nuclei mask comprises one or more putative nuclei associated with the plurality of cells.

16. The method of claim 15 , further comprising cleaning the nuclei mask by removing one or more of holes, isolated pixels, or bridge pixels.

17. The method of claim 15 , further comprising contrast normalizing the nuclei mask.

18. The method of claim 15 , further comprising removing thin lines of pixels included in the nuclei mask.

19. The method of claim 15 , further comprising breaking one or more large regions of the nuclei mask into individual nuclei.

20. The method of claim 15 , further comprising removing one or more very small regions from the nuclei mask.

Assignees

• Univ Of Pittsburgh—Of The Commonwealth System Of Higher Education
• Univ Of Pittsburg—Of The Commonwealth System Of Higher Education

Inventors

• Burger Virginia M
• Chennubhotla Srinivas C

Classifications

• G06V10/457

  by analysing connectivity, e.g. edge linking, connected component analysis or slices · CPC title

• G06T7/11 Primary

  Region-based segmentation · CPC title

• G06F18/2148

  characterised by the process organisation or structure, e.g. boosting cascade · CPC title

• G06V10/44

  Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components · CPC title

• G06V10/143

  Sensing or illuminating at different wavelengths · CPC title