Method and apparatus for fast detection of spiculated lesions in digital mammograms

• US 6,263,092 B1
• Filed: 06/23/1998
• Issued: 07/17/2001
• Est. Priority Date: 07/10/1996
• Status: Expired due to Term

• Current Assignee: Hologic Incorporated
• Original Assignee: R2 Technology, Inc. (Hologic Incorporated)
• Inventors: Roehrig, Jimmy R., Zhang, Wei, Romsdahl, Harlan M.
• Primary Examiner(s): Patel, Jayanti K.
• Application Number:
  US09/103,290
• Time in Patent Office: 1,120 Days
• Field of Search: 382/128-134, 382/100, 382/155, 382/156, 382/172, 382/181, 382/227, 382/260, 382/296, 382/194, 378/21, 378/37, 600/300, 600/410, 600/425, 128/915
• US Class Current: 382/128
• CPC Class Codes: A61B 6/502 for diagnosis of breast, i.e. mammography
• CPC Class Codes: G06F 18/00 Pattern recognition
• CPC Class Codes: G06T 2207/30068 Mammography; Breast
• CPC Class Codes: G06T 7/0012 Biomedical image inspection
• CPC Class Codes: G06V 10/25 Determination of region of interest [ROI] or a volume of interest [VOI]
• CPC Class Codes: G06V 10/36 Applying a local operator, i.e. means to operate on image points situated in the vicinity of a given point; Non-linear local filtering operations, e.g. median filtering
• CPC Class Codes: G06V 10/457 by analysing connectivity, e.g. edge linking, connected component analysis or slices
• CPC Class Codes: G06V 10/48 by mapping characteristic values of the pattern into a parameter space, e.g. Hough transformation
• CPC Class Codes: G06V 2201/03 Recognition of patterns in medical or anatomical images

Abstract

A method and apparatus for the fast detection of spiculated lesions in a digital mammogram, the method for use in a computer aided diagnosis system for assisting a radiologist in identifying and recognizing the spiculations among a multiplicity of lines corresponding to standard fibrous breast tissue. A line and direction image is created from a digital mammogram, and a region of potential intersection for substantially every pixel in the digital mammogram image is determined. The region of potential intersection for each pixel is a predetermined pattern, such as a high aspect ratio rectangle or trapezoid, positioned around the pixel and rotated in a direction corresponding to direction information for that pixel. The regions of potential intersection are accumulated among the pixels to produce a cumulative array, and information in the cumulative array is processed for identifying spiculations in the digital mammogram.

Claims

1. A method of detecting spiculations in an image, the image having pixels, comprising the steps of:

determining a region of potential intersection for each of a plurality of image pixels using line information and direction information related to that image pixel;

accumulating said regions of potential intersection to produce a cumulative array;

processing information contained in said cumulative array for identifying the spiculations in the image.

2. The method of claim 1, wherein said step of determining a region of potential intersection is performed for substantially every image pixel.

3. The method of claim 2, wherein said step of determining a region of potential intersection comprises the steps of:

determining, according to the line information related to the image pixel, whether the image pixel is located along a line; and

if the image pixel is located along a line, selecting a region centered on the image pixel corresponding to a predetermined pattern, said predetermined pattern being rotated by an amount related to the direction information related to the image pixel.

4. The method of claim 3, wherein said step of determining a region of potential intersection comprises the step of selecting a null region if the image pixel is not located along a line.

5. The method of claim 4, said predetermined pattern having a center, said predetermined pattern comprising two generally rectangular portions symmetrically positioned around the center and extending from an inner radius to an outer radius.

6. The method of claim 5, wherein the inner and outer radii of the predetermined pattern are chosen according to a radius of a desired spiculation size.

7. The method of claim 6, wherein the inner radius of the predetermined pattern is less than the radius of the desired spiculation size, and wherein the outer radius of the predetermined pattern is greater than the radius of the desired spiculation size.

8. The method of claim 4, the image being M by N pixels in size, said cumulative array being M by N pixels.

9. The method of claim 4, said step of accumulating said regions of potential intersection comprising the step of, for each of said image pixels, incrementing all pixels in said cumulative array located within said region of potential intersection for the image pixel by a first amount.

10. The method of claim 9, wherein said first amount is a fixed amount independent of the direction information corresponding to the pixel.

11. The method of claim 9, further comprising the steps of:

computing line information and direction information related to substantially every image pixel; and

computing a weighting function based on statistical information related to the direction information for substantially every image pixel and being a function of direction information;

wherein

said first amount is equal to the weighting function corresponding to the direction information related to the image pixel.

12. The method of claim 11, the image pixel having coordinates (i,j), said weighting function being a function WT(theta), said direction information related to the pixel being an angle THETA(i,j) and corresponding to a tangent of a line passing through the image pixel, wherein said first amount is equal to WT(THETA(i,j)).

13. The method of claim 12, said step of computing the weighting function WT(theta) comprising the steps of:

computing a histogram H(theta) of the direction information THETA(i,j) for all image pixels;

computing the weighting function WT(theta) having an inverse relationship to the histogram H(theta), the weighting function WT(theta) having a minimum value for a value of theta corresponding to a maximum value of the histogram H(theta).

14. The method of claim 2, the image being M by N pixels in size, said cumulative array having a size of AM by BN pixels, where A<

1 and B<

1, said step of determining a region of potential intersection comprising the steps of;

determining, according to the line information related to the image pixel, whether the image pixel is located along a line, the image pixel having coordinates; and

if the image pixel is located along a line, selecting a region centered on a proportional center pixel corresponding to a predetermined pattern, said predetermined pattern being rotated by an amount related to the direction information related to the image pixel, wherein said proportional center pixel has coordinates equal to said coordinates of the image pixel scaled by A and B.

15. The method of claim 14, wherein said coordinates of the image pixel comprise a first coordinate and a second coordinate, and wherein said coordinates of said proportional center pixel comprise a first coordinate equal to said first coordinate of said image pixel multiplied by A and integerized, and wherein said coordinates of said proportional center pixel further comprise a second coordinate equal to said second coordinate of said image pixel multiplied by A and integerized.

16. The method of claim 15, said step of accumulating said regions of potential intersection comprising the step of, for each of said image pixels, incrementing all pixels in said cumulative array located within said region of potential intersection for the image pixel by a first amount.

17. The method of claim 1, said step of processing information contained in said cumulative array for identifying the spiculations in the image comprising the step of locating local maxima in said cumulative array for locating spiculations in the image.

18. The method of claim 1, said step of processing information contained in said cumulative array for identifying the spiculations in the image comprising the step of thresholding said cumulative array for locating spiculations in the image.

19. A method of detecting regions of interest in an image, the image having pixels, comprising the steps of:

computing mass information corresponding to said image, including mass location information;

determining a region of potential intersection for each of a plurality of image pixels using line information and direction information related to that image pixel;

accumulating said regions of potential intersection to produce a cumulative array; and

using information contained in said cumulative array in conjunction with said mass information for identifying regions of interest in said image.

20. The method of claim 19, said step of using information contained in said cumulative array comprising the steps of:

locating a spiculation by using information in said cumulative array; and

comparing said spiculation location with said mass location information for determining the presence of a region of interest.

21. The method of claim 20, said step of locating a spiculation comprising the step of identifying a location of a local maximum in said cumulative array.

22. The method of claim 20, said step of locating a spiculation comprising the step of thresholding said cumulative array for identifying spiculations near locations where said cumulative array exceeds a threshold value.

23. The method of claim 22, said threshold value being a constant value.

24. The method of claim 19, said step of using information contained in said cumulative array comprising the steps of:

identifying a location of a strong circumscribed mass candidate using said mass information;

thresholding a first region of said cumulative array with a first threshold, said first region not including said strong circumscribed mass candidate location;

thresholding a second region of said cumulative array with a second threshold, said second region including said strong circumscribed mass candidate location, said second threshold being lower than said first threshold; and

if said second threshold is exceeded at said strong circumscribed mass candidate location, indicating a region of interest at said strong circumscribed mass candidate location.

25. The method of claim 19, said step of computing mass information comprising the steps of:

computing a gradient plane from said image, said gradient plane having pixels, each gradient plane pixel having a gradient intensity value and a gradient direction value;

determining a region of potential centroid for each gradient plane pixel using the gradient intensity value and gradient direction value for that pixel;

accumulating said regions of potential centroid to produce a sphericity array; and

using information contained in said sphericity array for determining said mass information.

26. The method of claim 25, wherein said step of determining a region of potential centroid comprises the step of selecting a region centered on the gradient plane pixel corresponding to a predetermined pattern, said predetermined pattern being rotated by an amount related to the gradient direction value for the gradient plane pixel.

27. The method of claim 26, said predetermined pattern having a center, said predetermined pattern comprising two generally rectangular portions symmetrically positioned around the center and extending from an inner radius to an outer radius.

28. The method of claim 27, wherein the inner and outer radii of the predetermined pattern are chosen according to a radius of a desired mass size.

29. The method of claim 28, said step of accumulating said regions of potential centroid comprising the step of, for each of said gradient plane pixels, incrementing all pixels in said sphericity array located within said region of potential centroid by an amount corresponding to the gradient intensity value for that gradient plane pixel.

30. The method of claim 25, said step of using information contained in said cumulative array comprising the steps of:

identifying a location of a strong circumscribed mass candidate using said mass information;

thresholding a first region of said cumulative array with a first threshold, said first region not including said strong circumscribed mass candidate location;

thresholding a second region of said cumulative array with a second threshold, said second region including said strong circumscribed mass candidate location, said second threshold being lower than said first threshold; and

if said second threshold is exceeded near said strong circumscribed mass candidate location, indicating a region of interest near said strong circumscribed mass candidate location.

31. The method of claim 19, said mass information including mass events, each event comprising mass centroid location, mass area, mass elongation, and mass contrast, said step of using information contained in said cumulative array comprising the step of using linear classifiers for prioritizing said mass events.

32. A method of detecting spiculations in an image, said method being capable of locating noneccentric spiculations, the image having pixels, said method comprising the steps of:

determining a region of potential intersection for each of a plurality of image pixels using line information and direction information related to that image pixel;

computing a plurality of weights corresponding to each of said plurality of image pixels;

accumulating, for each of said plurality of image pixels, said plurality of weights into a plurality of accumulation planes for those pixels located within said region of potential intersection for that image pixel; and

processing information contained in said plurality of accumulation planes for identifying the noneccentric spiculations in the image.

33. The method of claim 32, wherein said plurality of image pixels comprises substantially every pixel in said image.

34. The method of claim 33, wherein said step of determining a region of potential intersection comprises the steps of:

determining, according to the line information related to the image pixel, whether the image pixel is located along a line; and

if the image pixel is located along a line, selecting a region centered on the image pixel corresponding to a predetermined pattern, said predetermined pattern being rotated by an amount related to the direction information related to the image pixel.

35. The method of claim 34, wherein said step of determining a region of potential intersection further comprises the step of selecting a null region if the image pixel is not located along a line.

36. The method of claim 35, said plurality of accumulation planes comprising a first accumulation plane ACC1, a second accumulation plane ACC2, and a third accumulation plane ACC3, wherein said first, second, and third accumulation planes ACC1, ACC2, and ACC3 are capable of being processed for producing a spiculation activity plane ACT and a spiculation eccentricity plane ECC for use in locating noneccentric spiculations.

37. The method of claim 36, said plurality of accumulation planes, said spiculation activity plane ACT and said spiculation eccentricity plane ECC each comprising pixels with coordinates (i,j), said step of processing information in said plurality of accumulation planes further comprising the steps of:

computing said spiculation activity plane ACT and said spiculation eccentricity plane ECC using information in said plurality of accumulation planes such that said spiculation activity plane ACT comprises pixel values related to the presence of spiculations, and such that said spiculation eccentricity plane ECC comprises pixel values related to the presence of eccentric spiculations;

forming a spiculation output plane SO, said spiculation output plane comprising pixels with coordinates (i,j), said spiculation output plane SO being a function of said spiculation activity plane ACT and said spiculation eccentricity plane ECC at each pixel; and

using information in said spiculation output plane SO for identifying the noneccentric spiculations in the image.

38. The method of claim 37, wherein SO(i,j) is set equal to. a first constant multiplied by ACT(i,j) added to a second constant multiplied by ECC(i,j).

39. The method of claim 38, wherein said first constant is equal to 2 and said second constant is equal to −

1.

40. The method of claim 39, the image pixels having coordinates (i,j), said direction information being an angle THETA(i,j), said plurality of weights corresponding to each image pixel comprising a first weight W1(i,j) for accumulating into said first accumulation plane ACC1, a second weight W2(i,j) for accumulating into said second accumulation plane ACC2, and a third weight W3(i,j) for accumulating into said third accumulation plane ACC3, said first weight W1(i,j) being proportional to (cos(THETA(i,j))**2.

41. The method of claim 40, said second weight W2(i,j) being proportional to (sin(THETA(i,j))**2, and said third weight W3(i,j) being proportional to 2cos(THETA(i,j))sin(THETA(i,j)).

42. The method of claim 41, said step of computing said spiculation activity plane ACT and said spiculation eccentricity plane ECC comprising the steps of:

for each pixel at location (i,j), setting ACT(i,j) equal to W1(i,j)+W2(i,j); and

for each pixel at location (i,j), setting ECC(i,j) equal to SQRT((W1(i,j)−

W2 (i,j))**2+(W3(i,j))**2).

43. The method of claim 42, said method further comprising the steps of:

computing mass information corresponding to said image, said mass information including mass events, each event comprising mass centroid location, mass area, mass elongation, and mass contrast; and

using information contained in said SO, ACT and ECC arrays in conjunction with said mass information for identifying regions of interest in said image, including the step of using linear classifiers for prioritizing said mass events.

44. An apparatus for detecting spiculated lesions in a digital mammogram, said apparatus comprising:

a first memory for storing a line and direction image comprising pixels, each pixel containing line information and direction information derived from said digital mammogram;

a processor for determining a region of potential intersection corresponding to each of said line and direction image pixels using said line information and direction information; and

an accumulator for accumulating said regions of potential intersection to produce a cumulative array;

wherein the spiculated lesions in the digital mammogram may be detected by processing information contained in said cumulative array.

45. A computer-readable medium which can be used for directing an apparatus to detect spiculations in an image, comprising:

means for directing a first memory of said apparatus to store line and direction image comprising pixels, each pixel containing line information and direction information derived from said digital mammogram;

means for directing a processor of said apparatus to determine a region of potential intersection corresponding to each line and direction image pixel using said line information and direction information; and

means for directing an accumulator of said apparatus for accumulating said regions of potential intersection to produce a cumulative array;

wherein the spiculated lesions in the digital mammogram may be detected by processing information contained in said cumulative array.

46. A method of detecting spiculations in an image, the image having pixels, comprising the steps of:

computing line information and direction information for each image pixel, said line information being equal to a nonzero constant if said pixel lies along a line, said line information being equal to zero otherwise, said direction information being equal to the angle THETA of a tangent to said line if said pixel lies along said line;

determining a region of potential intersection for each image pixel using line information and direction information related to that image pixel, comprising the steps of;

determining, according to the line information related to the image pixel, whether the image pixel is located along a line; and

if the image pixel is located along a line, selecting a region centered on the image pixel corresponding to a predetermined pattern, said predetermined pattern being rotated by the angle THETA for that image pixel; and

selecting a null region if the image pixel is not located along a line;

accumulating said regions of potential intersection to produce a cumulative array;

processing information contained in said cumulative array for identifying the spiculations in the image, comprising the step of identifying local maxima in said cumulative array;

wherein said predetermined pattern has a center, and wherein said predetermined pattern comprises two generally rectangular portions symmetrically positioned around the center and extending from an inner radius to an outer radius, said inner radius and said outer radius being determined according to a desired spiculation size.