Method and apparatus for image scoring and analysis

Patent metadata

Field	Value
Publication number	US-9852354-B2
Application number	US-201514701816-A
Country	US
Kind code	B2
Filing date	May 1, 2015
Priority date	May 5, 2014
Publication date	Dec 26, 2017
Grant date	Dec 26, 2017

Abstract

Methods and apparatuses for analyzing digital pathology images are provided. The methods and apparatuses may provide estimates of staining intensity and proportion score in regions of interest within an image. Digital pathology images may be scored according to these metrics. The methods and apparatuses disclosed herein utilize various predetermined thresholds, parameters, and models to increase efficiency and permit accurate estimation of characteristics of a stained tissue sample.

First claim

What is claimed is:

1. A method for analyzing a digital pathology image comprising:
selecting at least one analysis region in the digital pathology image;
separating the at least one analysis region into a tissue region and a background region using a predetermined separation threshold;
identifying a plurality of nuclear prospects within the tissue region, wherein identifying the plurality of nuclear prospects includes:
calculating a principal component projection of the tissue region,
segmenting the tissue region of the digital pathology image into a plurality of image layers via a spectral hashing function, and
determining the plurality of nuclear prospects from among the plurality of layers;

selecting nuclei from among the identified nuclear prospects, wherein selecting nuclei includes:
generating a hierarchical connected graph associating at least a first one of the plurality of nuclear prospects on a first segmented layer of the digital image with at least a second one of the plurality of nuclear prospects on a second segmented layer of the digital image,
comparing the first one of the nuclear prospects to the second one of the nuclear prospects to identify a superior nuclear prospect, and
designating the superior nuclear prospect as a selected nuclei; and

generating a diagnostic score of the at least one analysis region based on the selected nuclei.

2. The method according to claim 1 , wherein generating a diagnostic score includes scoring the analysis region based on color metrics of the selected nuclei.

3. The method according to claim 1 , wherein comparing the first one of the nuclear prospects and the second one of the nuclear prospects includes identifying the nuclear prospect having a higher convexity score.

4. The method according to claim 1 , wherein the at least one analysis region includes the entire digital pathology image.

5. The method according to claim 1 , wherein the at least one analysis region includes a plurality of analysis regions, the method further comprising:
generating a combined diagnostic score based on the diagnostic scores of the plurality of analysis regions.

6. A method for analyzing a digital pathology image comprising:
selecting an analysis region of the digital pathology image;
separating a tissue region and a background region in the analysis region of the digital pathology image using a predetermined separation threshold;
identifying a plurality of nuclear prospects within the tissue region;
selecting a portion of nuclei from among the plurality of identified nuclear prospects, wherein selecting the portion of nuclei includes:
generating a hierarchical connected graph associating at least a first one of the plurality of nuclear prospects on a first segmented layer of the digital image with at least a second one of the plurality of nuclear prospects on a second segmented layer of the digital image,
comparing the first one of the nuclear prospects to the second one of the nuclear prospects to identify a superior nuclear prospect, wherein comparing the first one of the nuclear prospects and the second one of the nuclear prospects includes identifying the nuclear prospect having a higher convexity score, and
designating the superior nuclear prospect as a selected nuclei;

generating a diagnostic score of the analysis region based on the selected nuclei, wherein generating the diagnostic score includes scoring the analysis region based on color metrics of the selected nuclei.

7. The method according to claim 6 , wherein identifying a plurality of nuclear prospects includes:
calculating a principal component projection of the tissue region;
segmenting the tissue region of the digital pathology image into a plurality of image layers via a spectral hashing function; and
determining the plurality of nuclear prospects from among the plurality of layers.

8. The method according to claim 6 , wherein the at least one analysis region includes the entire digital pathology image.

9. The method according to claim 6 , wherein the at least one analysis region includes a plurality of analysis regions, the method further comprising:
generating a combined diagnostic score based on the diagnostic scores of the plurality of analysis regions.

10. A system for analyzing a digital pathology image comprising:
a non-transitory computer readable medium comprising instructions;
at least one processor configured to carry out the instructions to:
select at least one analysis region in the digital pathology image;
separate the at least one analysis region into a tissue region and a background region using a predetermined separation threshold;
identify a plurality of nuclear prospects within the tissue region, wherein identifying the plurality of nuclear prospects includes:
calculating a principal component projection of the tissue region,
segmenting the tissue region of the digital pathology image into a plurality of image layers via a spectral hashing function, and
determining the plurality of nuclear prospects from among the plurality of layers;

select nuclei from among the identified nuclear prospects, wherein selecting the nuclei includes:
generating a hierarchical connected graph associating at least a first one of the plurality of nuclear prospects on a first segmented layer of the digital image with at least a second one of the plurality of nuclear prospects on a second segmented layer of the digital image,
comparing the first one of the nuclear prospects to the second one of the nuclear prospects to identify a superior nuclear prospect, and
designating the superior nuclear prospect as a selected nuclei; and

generate a diagnostic score of the at least one analysis region based on the selected nuclei.

11. The system according to claim 10 , wherein generating a diagnostic score includes scoring the analysis region based on color metrics of the selected nuclei.

12. The system according to claim 10 , wherein the instructions to compare the compare the first one of the nuclear prospects and the second one of the nuclear prospects further include instructions to identify the nuclear prospect having a higher convexity score.

13. The system according to claim 10 , wherein the at least one analysis region includes the entire digital pathology image.

14. The system according to claim 10 , wherein the at least one analysis region includes a plurality of analysis regions, and the instructions further comprise instructions to:
generate a combined diagnostic score based on the diagnostic scores of the plurality of analysis regions.

15. A system for analyzing a digital pathology image comprising:
a non-transitory computer readable medium comprising instructions;
at least one processor configured to carry out the instructions to:
select an analysis region of the digital pathology image;
separate a tissue region and a background region in the analysis region of the digital pathology image using a predetermined separation threshold;
identify a plurality of nuclear prospects within the tissue region;
select a portion of nuclei from among the plurality of identified nuclear prospects, wherein selecting the portion of nuclei includes:
generate a hierarchical connected graph associating at least a first one of the plurality of nuclear prospects on a first segmented layer of the digital image with at least a second one of the plurality of nuclear prospects on a second segmented layer of the digital image,
compare the first one of the nuclear prospects to the second one of the nuclear prospects to identify a superior nuclear prospect, wherein comparing the fi

Assignees

• Dako Denmark As

Inventors

• Jain Rohit
• Dhingra Kapil
• Hartford Edward

Classifications

• G06T2207/30024

  Cell structures in vitro; Tissue sections in vitro · CPC title

• G06T7/194

  involving foreground-background segmentation · CPC title

• G06T7/11

  Region-based segmentation · CPC title

• G06T7/0012

  Biomedical image inspection · CPC title

• G06V20/695 Primary

  Preprocessing, e.g. image segmentation · CPC title