Stain-based optimized compression of digital pathology slides

• US 8,077,959 B2
• Filed: 09/30/2009
• Issued: 12/13/2011
• Est. Priority Date: 09/30/2009
• Status: Active Grant

• Current Assignee: General Electric Company
• Original Assignee: General Electric Company
• Inventors: Dekel, Shai, Google, Leon
• Primary Examiner(s): STREGE, JOHN B
• Application Number:
  US12/570,290
• Publication Number:
  US 20110075897A1
• Time in Patent Office: 804 Days
• Field of Search: 382/128, 382/166, 382/133, 382/232
• US Class Current: 382/133
• CPC Class Codes: H04N 19/136 Incoming video signal characteristics or properties
• CPC Class Codes: H04N 19/186 the unit being a colour or a chrominance component
• CPC Class Codes: H04N 19/60 using transform coding
• CPC Class Codes: H04N 19/63 using sub-band based transform, e.g. wavelets
• CPC Class Codes: H04N 19/85 using pre-processing or post-processing specially adapted for video compression

Abstract

A novel and useful method and system of optimized image compression of digital pathology slide images. The optimized image compression mechanism exploits the special color properties of the stained tissue represented by the digital pathology slides and provides an image compression algorithm having improved rate-distortion performance. Optimized color transforms are pre-computed using training sets of pathology slide image scan data for each stain type. The optimized color transforms are used to compress input slide image scans resulting in more efficient image streaming enabling users to review extremely large digital slide scans from any connected location, such as in a hospital, satellite center, home or on a mobile telephone.

Claims

1. A method of image compression, said method comprising the steps of:

pre-computing a plurality of color transforms, each color transform computed in accordance with a set of training images comprising sample specimen pathology slide scan images; and

compressing an input digital image slide scan utilizing one of said color transforms.

2. The method according to claim 1, wherein said color transforms are computed using principle component analysis (PCA).

3. The method according to claim 1, wherein said plurality of color transforms are stored in a database.

4. The method according to claim 1, wherein each set of training images corresponds to a different pathology slide stain type.

5. The method according to claim 1, further comprising selecting quantization steps for a time-frequency transformation in accordance with the principle components of each color transform.

6. The method according to claim 5, wherein a smallest quantization step is applied to a first principle component, a next larger quantization step is applied to a second principle component, and a largest quantization step is applied to a third principle component.

7. A method of compressing digital pathology stained images, said method comprising the steps of:

pre-computing a plurality of color transforms, each color transform computed in accordance with a set of training slide images corresponding to a particular stain and comprising sample specimen pathology slide scan images;

mapping an input digital image slide scan to a particular stain; and

compressing said input digital image scan utilizing a pre-computed color transform corresponding to said mapped stain.

8. The method according to claim 7, wherein said color transforms are computed using principle component analysis (PCA).

9. The method according to claim 7, wherein mapping of an input digital image to a particular stain is based on information read from a laboratory information system (LIS).

10. The method according to claim 7, further comprising selecting quantization steps for a time-frequency transformation in accordance with the principle components of each color transform.

11. The method according to claim 10, wherein a smallest quantization step is applied to a first principle component, a next larger quantization step is applied to a second principle component, and a largest quantization step is applied to a third principle component.

12. A server computer for performing image compression of digital pathology slide scans, comprising:

an image storage device adapted to store a plurality of pathology slide scans; and

an image compression module for mapping an input slide scan digital image to a particular stain and compressing said input digital image utilizing a pre-computed color transform corresponding to said mapped stain, wherein color transforms are pre-computed, each color transform computed in accordance with a set of training images corresponding to a particular stain and comprising sample specimen pathology slide images.

13. The server according to claim 12, wherein said color transforms are computed using principle component analysis (PCA).

14. The server according to claim 12, said image compression module further comprising means for selecting quantization steps for a time-frequency transformation in accordance with the principle components of each color transform.

15. The server according to claim 14, wherein a smallest quantization step is applied to a first principle component, a next larger quantization step is applied to a second principle component, and a largest quantization step is applied to a third principle component.

16. A method of computing optimized color transforms for use in compressing digital pathology slide scans, said method comprising the steps of:

receiving sample stain data from a set of training slide scans comprising sample pathology specimens representative of a particular histochemistry staining method of a particular tissue type;

forming an input vector from said sample stain data; and

for each training slide set, calculating a color transform utilizing said input vector and storing the resulting matrix coefficients in a database.

17. The method according to claim 16, wherein said color transforms are computed using principle component analysis (PCA).

18. The method according to claim 16, further comprising the steps of:

determining a stain type for an input digital pathology image;

retrieving a color transform corresponding to said stain type from said database; and

utilizing said color transform in compressing said input digital pathology image.

19. The method according to claim 16, further comprising the steps of:

determining a stain type for an input digital pathology image;

retrieving a color transform corresponding to said stain type from said database; and

applying a smallest quantization step to a first principle component of said color transform, applying a next larger quantization step to a second principle component of said color transform, and applying a largest quantization step to a third principle component of said color transform.

20. The method according to claim 16, wherein said step of calculating a color transform from a training slide set comprises the steps of:

calculating a covariance matrix of said training slide set;

calculating eigenvectors and eigenvalues of said covariance matrix;

normalizing said eigenvectors by said eigenvalues; and

sorting said normalized eigenvectors by said eigenvalues.

21. A computer program product characterized by that upon loading it into computer memory a digital pathology stained image compression process is executed, the computer program product comprising:

a non-transitory computer usable medium having computer usable program code embodied therewith, the computer usable program code comprising;

computer usable code configured to pre-compute a plurality of color transforms, each color transform computed in accordance with a set of training slide images corresponding to a particular stain;

computer usable code configured to map an input digital image to a particular stain; and

computer usable code configured to compress said input digital image utilizing a pre-computed color transform corresponding to said mapped stain.