Image quality assessment including comparison of overlapped margins

• US 8,396,269 B2
• Filed: 04/08/2010
• Issued: 03/12/2013
• Est. Priority Date: 04/08/2010
• Status: Expired due to Fees

• Current Assignee: Digital Pathco, LLC, General Electric Company, Omnyx, LLC
• Original Assignee: Digital Pathco, LLC
• Inventors: Henderson, David L., Kenny, Kevin B., Padfield, Dirk R., Gao, Dashan, McKay, Richard R., Baxi, Vipul A., Filkins, Robert J., Montalto, Michael C.
• Primary Examiner(s): CHANG, JON CARLTON
• Application Number:
  US12/756,877
• Publication Number:
  US 20110249910A1
• Time in Patent Office: 1,069 Days
• Field of Search: 382/128, 382/284, 382/190, 382/195
• US Class Current: 382/128
• CPC Class Codes: G06T 2207/10056 Microscopic image
• CPC Class Codes: G06T 2207/20021 Dividing image into blocks, subimages or windows
• CPC Class Codes: G06T 2207/30024 Cell structures in vitro; Tissue sections in vitro
• CPC Class Codes: G06T 2207/30168 Image quality inspection
• CPC Class Codes: G06T 7/0012 Biomedical image inspection
• CPC Class Codes: G06V 10/16 using multiple overlapping images; Image stitching
• CPC Class Codes: G06V 10/993 Evaluation of the quality of the acquired pattern
• CPC Class Codes: G06V 20/693 Acquisition

Abstract

Image quality is assessed for a digital image that is a composite of tiles or other image segments, especially focus accuracy for a microscopic pathology sample. An algorithm or combination of algorithms correlated to image quality is applied to pixel data at margins where adjacent image segments overlap and thus contain the same content in separately acquired images. The margins may be edges merged to join the image segments smoothly into a composite image, and typically occur on four sides of the image segments. The two versions of the same image content at each margin are processed by the quality algorithm, producing two assessment values. A sign and difference value are compared with other image segments, including by subsets selected for the orientation of the margins on sides on the image segments. The differences are mapped to displays. Selection criteria determine segments to be re-acquired.

Claims

1. An image processing method, comprising:

acquiring a first digital image encompassing a limited area of a subject being imaged;

acquiring at least one second digital image encompassing an area of the subject adjacent to the limited area, wherein the first and second digital images overlap along margins thereof, the first and second digital images providing pixel data sets wherein pixel data subsets corresponding to the margins encompass a same area of the subject;

applying a numeric image analysis algorithm to the first and second digital images at the pixel data subsets corresponding to the margins, thereby producing two characterizations for said same area of the subject as represented by the first and second digital images, wherein the numeric image analysis includes generating at least one summary value characterizing accuracy of focus for each of the overlapping margins and comparing relative accuracy of focus for the first and second digital images in the margins; and

,

at least one of normalizing, encoding, storing and indicating a measurement associated with the first and second digital images, based on comparison of the two characterizations for said same area.

2. The method of claim 1, wherein the subject being imaged comprises a specimen slide and further comprising obtaining a plurality of microscopic digital pictures, each of the digital pictures corresponding to a respective said limited area and having at least one said area adjacent thereto, and the plurality of microscopic digital pictures together encompassing at least a predetermined portion of the specimen slide.

3. The method of claim 2, wherein at least some of the plurality of microscopic digital pictures overlap other microscopic digital pictures on at least two edges.

4. The method of claim 1, wherein the numeric image analysis comprises producing a numerical value relating to differences in focus measurements in the margins from at least one of a derivative based algorithm, a statistical algorithm, a histogram based algorithm and an intuitive algorithm.

5. The method of claim 1, wherein the numeric analysis comprises applying at least two feature extraction algorithms that differ by at least one of algorithm type and scaling, each of the feature extraction algorithms producing algorithm output values, and discriminating among the digital images based on at least one of individual said algorithm output values and combinations of said algorithm output values.

6. The method of claim 5, further comprising normalizing the algorithm output values before said discriminating.

7. An image processing method, comprising:

acquiring a first digital image encompassing a limited area of a subject being imaged;

acquiring at least one second digital image encompassing an area of the subject adjacent to the limited area, wherein the first and second digital images overlap along margins thereof, the first and second digital images providing pixel data sets wherein pixel data subsets corresponding to the margins encompass a same area of the subject;

applying a numeric image analysis algorithm to the first and second digital images at the pixel data subsets corresponding to the margins, thereby producing two characterizations for said same area of the subject as represented by the first and second digital images;

further comprising acquiring a plurality of said first and second digital images forming image segments in an array encompassing the subject, wherein the numeric analysis includes generating at least one summary value characterizing accuracy of focus for plural overlapping margins of the image segments, and the comparison of the two characterizations includes determining a difference value having a sign and a value; and

at least one of normalizing, encoding, storing and indicating a measurement associated with the first and second digital images, based on comparison of the two characterizations for said same area.

8. The method of claim 7, wherein the margins occur on multiple sides of the digital images, and further comprising quantifying the difference values for at least two sides for each member of a population of the digital images and discriminating among the digital images based on the difference values.

9. The method of claim 8, further comprising determining a population of the difference values for at least one of all the sides of the digital images and selectively oriented sides of the digital images, and applying an acceptance criterion for encoding a relative rank in the population of a difference value for at least one oriented side of the digital images.

10. The method of claim 9, further comprising at least one of labeling the digital images in a composite image, and shading the digital images in the composite image, for providing a visual indication of said relative rank in the population.

11. The method of claim 10, further comprising providing the visual indication of said relative rank for a selected subset comprising one or more of said sides of the digital images.

12. The method of claim 10, wherein the visual indication is color coded to distinguish between at least one of the sign and the difference value for at least one of the margins of the digital images.

13. The method of claim 9, further comprising shading the digital images in the composite image, for providing a visual indication to distinguish between digital images that meet and do not meet a predetermined threshold.

14. The method of claim 7, wherein the margins occur on multiple sides of the digital images, and further comprising quantifying the difference values for at least one side for each member of a population of the digital images, wherein the at least one side for each member is oriented toward a same direction, and discriminating among the digital images based on the difference values and orientation of the direction.

15. The method of claim 14, further comprising associating pairs of sides that are one of opposite and adjacent sides of the digital images, and discriminating among the digital images based on the difference values and the orientation of said pairs of sides.

16. The method of claim 14, wherein said discriminating comprises discarding selected ones of the digital images based on the difference values and re-acquiring digital images encompassing a same area of the subject.

17. A method for digital pathology imaging, comprising:

mounting a pathology specimen on a slide defining an X-Y plane;

arranging the slide on a stage of a microscope having a digital camera for recording at a Z axis focal distance a magnified image of the specimen at an X-Y position, the microscope having a relative positioning control in the X-Y plane and a relative focal distance control along the Z axis;

obtaining and at least temporarily storing successive pixel data images at different focal distances, applying a focus accuracy algorithm to determine and select a preferred optimal focal distance, and collecting a pixel data image;

advancing the slide in at least one of an X direction and a Y direction by a distance sufficient to encompass a magnified image of the specimen adjacent to a previously collected pixel data image, with a marginal area of overlap along a margin with the previously collected pixel data image, and collecting a next pixel data image;

continuing to advance over the slide, and to collect pixel data images with marginal areas of overlap until the successive pixel data images encompass at least a predetermined portion of the specimen, each pixel data image corresponding to an image segment in a composite image that overlaps other said image segments at a respective said marginal area, such that at least two pixel data characterizations of the respective marginal area are provided in the first and second digital images;

analyzing pixel data in the marginal areas of overlap by applying at least one numeric image analysis algorithm correlated with image quality to at least two pixel data sets in corresponding areas, and obtaining a sign and difference value for an output of the at least one numeric image analysis algorithm;

discriminating among the pixel data images based on the difference values for the marginal areas of the pixel data sets in the composite image.

18. The method of claim 17, further comprising merging the pixel data images across the marginal areas of overlap to form the composite image of the predetermined portion of the specimen, and navigating the composite image using a digital display.

19. The method of claim 17, further comprising re-acquiring pixel data images where the difference values fail to meet an acceptance criterion.

20. The method of claim 17 further comprising representing at least one of said difference values, a ranking of said difference values and an acceptance status for the pixel data images, on a visual presentation having areas corresponding to the pixel data images in the composite image.

21. The method of claim 17, wherein discriminating among the pixel data images based on the difference values is at least partly specific to an orientation of the marginal area of overlap relative to an associated one of the pixel data images.