Breast segmentation in radiographic images

• US 8,675,933 B2
• Filed: 06/24/2011
• Issued: 03/18/2014
• Est. Priority Date: 04/30/2010
• Status: Active Grant

• Current Assignee: iCAD Incorporated
• Original Assignee: VuCOMP, Inc.
• Inventors: Wehnes, Jeffrey C., Pike, James H., Monaco, James P.
• Primary Examiner(s): TUCKER, WESLEY J
• Application Number:
  US13/168,588
• Publication Number:
  US 20110280465A1
• Time in Patent Office: 998 Days
• Field of Search: 382/128, 382130-133, 382/266
• US Class Current: 382/128
• CPC Class Codes: G06F 18/24 Classification techniques
• CPC Class Codes: G06T 2207/10116 X-ray image
• CPC Class Codes: G06T 2207/20016 Hierarchical, coarse-to-fine, multiscale or multiresolution image processing; Pyramid transform
• CPC Class Codes: G06T 2207/30068 Mammography; Breast
• CPC Class Codes: G06T 2207/30096 Tumor; Lesion
• CPC Class Codes: G06T 7/0012 Biomedical image inspection
• CPC Class Codes: G06T 7/11 Region-based segmentation
• CPC Class Codes: G06T 7/143 involving probabilistic approaches, e.g. Markov random field [MRF] modelling

Abstract

An image segmentation embodiment comprises generating a start model comprising a set of model points approximating an outline of an object in an initial image, smoothing the image at a first smoothing level, generating a curvature image by applying a second derivative operator, locating second derivative local maxima in the curvature image that are orthogonal to a respective model point and within a search region having a first boundary on one side of the start model and a second boundary on an opposite side of the start model, generating a set of contours, shifting the start model to an outer boundary of the contours, and generating a segmentation mask of the object based on the shifted start model.

Claims

1. A method for segmenting an image comprising pixels, the method comprising:

generating a start model comprising a set of model points approximating an outline of an object in an initial image;

subsampling the initial image at a first scale to generate a subsampled image;

smoothing the subsampled image at a first smoothing level to generate a smoothed image;

generating a curvature image by applying a second derivative operator to the smoothed image;

locating, for each of the model points in the start model, second derivative local maxima in the curvature image that are orthogonal to a respective model point and within a search region having a first boundary on one side of the start model and a second boundary on an opposite side of the start model;

generating a set of contours by growing a contour for each of the second derivative local maxima;

shifting the start model to an outer boundary of the contours; and

generating a segmentation mask of the object based on the shifted start model.

2. The method of claim 1, wherein the generating the curvature image further comprises applying a first derivative operator to the smoothed image.

3. The method of claim 1, wherein the generating the curvature image further comprises determining first and second eigenvalues and an angle of orientation at each point in the curvature image.

4. The method of claim 3, wherein the growing the contour comprises using the eigenvalues of curvature image points in a plurality of directions around the respective model point to determine a growth direction of the contour.

5. The method of claim 3, wherein the growing the contour comprises using a maximum angle change per growth step and/or an angular filter to determine a growth direction of the contour.

6. The method of claim 1, wherein the generating the set of contours further comprises pruning contours having at least a portion outside of the search region and/or having a length less than a length threshold.

7. The method of claim 1, wherein the generating the set of contours further comprises grouping together overlapping ones of the contours.

8. The method of claim 7, further comprising mapping the set of contours to a linear model space before the grouping.

9. The method of claim 1, wherein the generating the set of contours comprises utilizing each of the second derivative local maxima as a starting point to grow each respective contour.

10. The method of claim 1, further comprising repeating the subsampling, the smoothing, the generating a curvature image, the locating the local maxima, and the generating the set of contours steps using the shifted start model, a second scale and a second smoothing level.

11. The method of claim 1, wherein the image is a mammogram image, and wherein the object is a breast.

12. A system for segmenting an image comprising pixels, the system comprising:

a object model generator generating a start model comprising a set of model points approximating an outline of an object in an initial image;

an image subsampler subsampling the initial image at a first scale to generate a subsampled image;

an image smoother smoothing the subsampled image at a first smoothing level to generate a smoothed image;

a curvature image generator generating a curvature image by applying a second derivative operator to the smoothed image;

a local maxima detector locating, for each of the model points in the start model, second derivative local maxima in the curvature image that are orthogonal to a respective model point and within a search region having a first boundary on one side of the start model and a second boundary on an opposite side of the start model;

a contour generator generating a set of contours by growing a contour for each of the second derivative local maxima;

an object model shifter shifting the start model to an outer boundary of the contours; and

a segmentation mask generator generating a segmentation mask of the object based on the shifted start model.

13. A computer program product for segmenting an image, the computer program product having a non-transitory computer-readable medium with a computer program embodied thereon, the computer program comprising:

computer program code for generating a start model comprising a set of model points approximating an outline of an object in an initial image;

computer program code for subsampling the initial image at a first scale to generate a subsampled image;

computer program code for smoothing the subsampled image at a first smoothing level to generate a smoothed image;

computer program code for generating a curvature image by applying a second derivative operator to the smoothed image;

computer program code for locating, for each of the model points in the start model, second derivative local maxima in the curvature image that are orthogonal to a respective model point and within a search region having a first boundary on one side of the start model and a second boundary on an opposite side of the start model;

computer program code for generating a set of contours by growing a contour for each of the second derivative local maxima;

computer program code for shifting the start model to an outer boundary of the contours; and

computer program code for generating a segmentation mask of the object based on the shifted start model.

14. The computer program product of claim 13, wherein the computer program code for generating the curvature image further comprises computer program code for applying a first derivative operator to the smoothed image.

15. The computer program product of claim 13, wherein the computer program code for generating the curvature image further comprises computer program code for determining first and second eigenvalues and an angle of orientation at each point in the curvature image.

16. The computer program product of claim 15, wherein the computer program code for growing the contour comprises computer program code for using the eigenvalues of curvature image points in a plurality of directions around the respective model point to determine a growth direction of the contour.

17. The computer program product of claim 15, wherein the computer program code for growing the contour comprises computer program code for using a maximum angle change per growth step and/or an angular filter to determine a growth direction of the contour.

18. The computer program product of claim 13, wherein the computer program code for generating the set of contours further comprises computer program code for pruning contours having at least a portion outside of the search region and/or having a length less than a length threshold.

19. The computer program product of claim 13, wherein the computer program code for generating the set of contours further comprises computer program code for grouping together overlapping ones of the contours.

20. The computer program product of claim 19, further comprising computer program code for mapping the set of contours to a linear model space before the grouping.

21. The computer program product of claim 13, wherein the computer program code for generating the set of contours comprises computer program code for utilizing each of the second derivative local maxima as a starting point to grow each respective contour.

22. The computer program product of claim 13, further comprising computer program code for repeating the subsampling, the smoothing, the generating a curvature image, the locating the local maxima, and the generating the set of contours steps using the shifted start model, a second scale and a second smoothing level.

23. The computer program product of claim 13, wherein the image is a mammogram image, and wherein the object is a breast.