Computerized detection of masses and microcalcifications in digital mammograms

• US 5,627,907 A
• Filed: 12/01/1994
• Issued: 05/06/1997
• Est. Priority Date: 12/01/1994
• Status: Expired due to Term

• Current Assignee: University Of Pittsburgh (University of Pittsburgh of The Commonwealth System of Higher Education)
• Original Assignee: University Of Pittsburgh (University of Pittsburgh of The Commonwealth System of Higher Education)
• Inventors: Chang, Yuang-Hsiang, Zheng, Bin, Gur, David
• Primary Examiner(s): Couso, Jose L.
• Application Number:
  US08/352,169
• Time in Patent Office: 887 Days
• Field of Search: 382/128, 382/130, 382/132, 382/170, 382/203, 364/413.13, 364/413.17, 364/413.19, 364/413.22, 364/413.23, 378/99, 358/111, 128/634, 128/754
• US Class Current: 382/132
• CPC Class Codes: G06T 7/0012 Biomedical image inspection
• CPC Class Codes: G06V 20/69 Microscopic objects, e.g. biological cells or cellular parts

Abstract

Method and apparatus for detecting abnormal regions such as masses and microcalcifications in living tissue depicted in radiographs such as mammograms is presented by digitizing the radiograph to produce a digitized radiograph, and, by computer, identifying, according to predetermined criteria, any suspected abnormal regions; and for each identified suspected abnormal region, extracting multiple topographic layers of the region from the digitized radiograph; and evaluating the extracted multiple topographic layers'"'"' abnormal regions to determine whether the region is an abnormal region; and displaying the abnormal regions in a version of the radiograph. Layers are extracted by thresholding the digital radiograph at a particular threshold value for each topographic layer. The threshold value for each other layer is based on the threshold and contrast values in a previous layer.

Claims

1. A method of detecting abnormal regions in living tissue depicted in a radiograph, the method comprising the steps of:

(a) digitizing the radiograph to produce a digitized radiograph;

(b) identifying, according to predetermined criteria, suspected abnormal regions in the radiograph; and

(c) for each identified suspected abnormal region,

(i) extracting multiple topographic layers of the suspected abnormal region from the digitized radiography, the multiple topographic layers being extracted from the same radiograph; and

(ii) evaluating the extracted multiple topographic layers of the suspected abnormal region to determine whether the suspected region is an abnormal region.

2. A method as in claim 1, wherein the step of evaluating comprises the steps of:

for each layer of the multiple topographic layers, determining features of the suspected abnormal region in the layer, and

analyzing the determined features for different topographic layers of the multiple topographic layers.

3. A method as in claim 1, wherein the step of evaluating comprises the steps of:

for at least two topographic layers of the extracted multiple topographic layers, determining features of the suspected abnormal region in the at least two topographic layers, and

analyzing the determined features for different topographic layers of the at least two topographic layers.

4. A method of claim 2, wherein the different topographic layers are adjacent topographic layers.

5. A method of claim 3, wherein the different topographic layers are adjacent topographic layers.

6. A method as in claim 2, wherein the step of determining features comprises, determining at least one of a size, contrast, shape factor and digital value fluctuation of the suspected abnormal region, and wherein the step of analyzing the determined features comprises comparing ratios between the determined features of different ones of the topographic layers and evaluating changes between the determined features of different ones of the topographic layers.

7. A method as in claim 1, wherein the step of evaluating the extracted multiple topographic layers comprises the steps of:

determining, for each region, the rate of change in elevation of the region in a plurality of directions; and

selecting, based on said determining, whether the region includes an abnormal region.

8. A method as in claim 1, further comprising the step of generating a rule-based criteria database, and wherein the step of evaluating evaluates suspected regions based on rule-based selection criteria from the database.

9. A method as in claim 1, wherein the step of identifying comprises the steps of, in the digitized radiograph:

detecting and enhancing skin boundaries;

non-linear thresholding, based on digital values in the digitized radiograph;

blob labelling and size selecting of suspected regions; and

framing and locating suspected regions in the digitized radiograph.

10. A method as in claim 1, wherein the radiograph is an X-ray mammogram, and wherein the multiple topographic layers comprise top-view radiographic topographic layers.

11. A method as in claim 10, wherein the multiple topographic layers comprises three topographic layers.

12. A method as in claim 1, wherein the radiograph is an X-ray mammogram, and wherein the abnormal regions are masses.

13. A method as in claim 1, wherein the radiograph is an X-ray mammogram, and wherein the abnormal regions are microcalcifications.

33. The method of claim 1, wherein the topographic layers are extracted by thresholding the digital radiograph at a particular threshold value for each topographic layer.

34. The method of claim 33, wherein the threshold value for each other layer is based on the threshold and contrast values in a previous layer.

35. The method of claim 33, wherein the threshold value T_j for layer j is a function of threshold value T_j-1 and contrast value C_j-1 in layer j-1 as given by:

space="preserve" listing-type="equation">T.sub.j =T.sub.j-1 +256×

C.sub.j-1.

42. The method of claim 33, wherein the threshold value of the first layer is a function of a digital value of a minimum point in the suspected region.

43. The method of claim 42, wherein the threshold value of the first layer is given by:

space="preserve" listing-type="equation">T.sub.1 =I.sub.min +I.sub.min /10, when I.sub.min ≦

500;

space="preserve" listing-type="equation">T.sub.1 =I.sub.min ×

1.0246+37.68, when (500<

I.sub.min <

2000); and

space="preserve" listing-type="equation">T.sub.1 =I.sub.min +I.sub.min /23, when I.sub.min ≧

2000,

where I_min is the digital value of the minimum point in the suspected region.

14. A method of detecting masses in living tissue depicted in a mammogram, the method comprising the steps of:

(a) digitizing the mammogram to produce a digitized mammogram;

(b) identifying, as suspected regions, according to predetermined criteria, suspected masses in the radiograph; and

(c) for each identified suspected region,

(i) extracting multiple topographic layers of the suspected region from the digitized mammogram, the multiple topographic layers all being extracted from the same mammogram; and

(ii) evaluating the extracted multiple topographic layers to determine whether the suspected region is a mass.

15. A method as in claim 14, wherein the multiple topographic layers comprise top-view radiographic topographic layers.

16. A method as in claim 14, wherein the step of identifying comprises one of the steps of, in the digitized mammogram,

single-image segmentation and

bilateral-image subtraction.

17. A method as in claim 14 wherein the step of identifying comprises the steps of:

obtaining first and second low-pass filtered images of the digitized mammogram;

creating a binary image from subtraction and thresholding of the first and second low-pass filtered images; and

labelling blobs and determining suspicious areas in the binary image.

36. The method of claim 14, wherein the topographic layers are extracted by thresholding the digital radiograph at a particular threshold value for each topographic layer.

37. The method of claim 36, wherein the threshold value for each other layer is based on the threshold and contrast values in a previous layer.

38. The method of claim 36, wherein the threshold value T_j for layer j is a function of threshold value T_j-1 and contrast value C_j-1 in layer j-1 as given by:

space="preserve" listing-type="equation">T.sub.j =T.sub.j-1 +256×

C.sub.j-1.

18. An apparatus for detecting abnormal regions in living tissue depicted in a radiograph, comprising:

a digitizer for digitizing the radiograph to produce a digitized radiograph;

means for identifying, according to predetermined criteria, suspected abnormal regions in the radiograph;

extracting means for extracting multiple topographic layers of identified suspected abnormal region from the digitized radiograph; and

means for evaluating the extracted multiple topographic layers of a suspected abnormal region, the multiple topographic layers extracted from the same radiograph, to determine whether the identified suspected abnormal region is an abnormal region.

44. An apparatus as in claim 18, wherein the extracting means extracts layers based on a threshold value and a contrast value for a previous layer.

19. An apparatus for detecting masses in living tissue regions depicted in a mammogram, comprising:

a digitizer for digitizing the mammogram to produce a digitized mammogram;

means for identifying, as suspected abnormal regions, according to predetermined criteria, suspected masses in the radiograph;

extracting means for extracting multiple topographic layers of identified suspected abnormal regions from the digitized mammogram; and

means for evaluating the extracted multiple topographic layers, the multiple topographic layers being extracted from the same mammogram, to determine whether the suspected region depicts a mass.

20. An apparatus for locating abnormal regions in living tissue depicted in a radiograph comprising:

imaging means for obtaining a radiographic image of an anatomical area of a body;

digitizing means for digitizing the obtained radiographic image to produce a digitized radiograph;

segmenting means for segmenting into suspected regions the digitized radiograph obtained from the digitizing means; and

multi-layer topographic pruning means for pruning the suspected regions found by the segmenting means to locate the abnormal regions in the radiograph, said pruning being based on multiple topographic layers of the suspected regions, the layers all extracted from the radiograph.

21. An apparatus in claim 20, further comprising a database having rules and criteria relating to determinations made by the pruning means.

22. A method of detecting clusters of microcalcifications in living tissue depicted in a mammogram, the method comprising the steps of:

(a) digitizing the mammogram to produce a digitized mammogram;

(b) identifying, as suspected regions, according to predetermined criteria, suspected clusters of microcalcifications in the digitized mammogram; and

(c) for each identified suspected region,

(i) extracting multiple topographic layers of the suspected region from the digitized mammogram, the layers being extracted from the mammogram; and

(ii) evaluating the multiple topographic layers extracted from the same mammogram to determine whether the suspected region depicts microcalcifications.

23. A method as in claim 22, wherein the step of evaluating comprises the steps of:

for each layer of the multiple topographic layers, determining features of the region in the layer, and

analyzing determined features of the region for different topographic layers of the multiple topographic layers.

24. A method as in claim 22, wherein the step of evaluating comprises the steps of:

for at least two topographic layers of the multiple topographic layers, determining features of the region in the at least two topographic layers, and

analyzing the determined features of the region for different topographic layers of the at least two topographic layers.

25. A method of claim 24, wherein the different topographic layers are adjacent.

26. A method of claim 24, wherein the different topographic layers are adjacent.

27. A method as in claim 22, wherein the step of evaluating multiple topographic layers comprises the steps of:

determining, for each region, a rate of change in elevation of the region in a plurality of directions; and

selecting, based on said determining, whether the region depicts microcalcifications.

28. A method as in claim 22, further comprising the step of generating a rule-based criteria database, and wherein the step of evaluating evaluates suspected regions based on rule-based selection criteria from the database.

29. A method as in claim 22, wherein the step of identifying comprises the steps of:

lowpass filtering the digital mammogram using two different filters to produce two lowpass filtered images;

subtracting the two lowpass filtered images;

non-linear thresholding the subtracted image, based on digital values in the digital mammogram;

blob labelling and size selecting of suspected regions;

blob clustering and blob selecting of selected suspected regions.

30. A method as in claim 22, wherein the step of identifying comprises the steps of:

dual Gaussian filtering the digitized mammogram to produce two filtered mammograms;

subtracting the two filtered mammograms to produce a subtracted mammogram;

local minimum searching in the subtracted mammogram, and, based on the local minimum searching, blob labelling and first clustering to produce blobs and clusters;

performing cluster analysis of the clusters, and based on the analysis, deleting some clusters;

extracting topographic features;

evaluating the extracted features based on feature criteria, and based on the evaluating, deleting some blobs;

second clustering positive blobs;

determining the number of blobs in a cluster, and, based on the determining, deleting some clusters; and

highlighting positive detections on the digitized mammogram.

31. The method of claim 30, wherein the step of cluster analysis of a cluster comprises the steps of:

(b) determining a maximum length of the cluster;

(c) determining a standard deviation of every blob in the cluster to the line; and

(d) based on a relationship between the determined maximum length and the determined standard deviation, removing some clusters from consideration.

32. The method of claim 31, wherein the relationship is whether the length divided by the standard deviation is less than a predetermined value.

39. The method of claim 22, wherein the topographic layers are extracted by thresholding an image formed by subtracting two lowpass filtered versions of the digital radiograph at a particular threshold value for each topographic layer.

40. The method of claim 39, wherein the threshold value for each other layer is based on a digital value of a pixel in the digitized mammogram and on a predetermined contrast value.

41. The method of claim 39, wherein the predetermined contrast value is one of 1%, 1.5%, and 2%.

45. A method of detecting an abnormal region in living tissue depicted in a radiograph, the method comprising the steps of:

(a) digitizing the radiograph to produce a digitized radiograph;

(b) identifying, according to predetermined criteria, a suspected abnormal region in the radiograph;

(c) extracting multiple topographic layers of the suspected abnormal region from the digitized radiograph, all layers being extracted from the radiograph; and

(d) evaluating the extracted multiple topographic layers of the suspected abnormal region to determine whether the suspected region is an abnormal region.

46. A method as in claim 45, wherein the step of evaluating comprises the steps

for each layer of the multiple topographic layers, determining features of the suspected abnormal region in the layer, and

analyzing the determined features for different topographic layers of the multiple topographic layers.

47. A method as in claim 45, wherein the step of evaluating comprises the steps of:

for at least two topographic layers of the extracted multiple topographic layers, determining features of the suspected abnormal region in the at least two topographic layers, and

analyzing the determined features for different topographic layers of the at least two topographic layers.

48. A method of claim 46, wherein the different topographic layers are adjacent topographic layers.

49. A method of claim 47, wherein the different topographic layers are adjacent topographic layers.

50. A method as in claim 46, wherein the step of determining features comprises, determining at least one of a size, contrast, shape factor and digital value fluctuation of the suspected abnormal region, and wherein the step of analyzing the determined features comprises comparing ratios between the determined features of different ones of the topographic layers and evaluating changes between the determined features of different ones of the topographic layers.

51. A method as in claim 45, wherein the step of evaluating the extracted multiple topographic layers comprises the steps of:

determining the rate of change in elevation of the region in a plurality of directions; and

selecting, based on said determining, whether the region includes an abnormal region.

52. A method as in claim 45, further comprising the step of generating a rule-based criteria database, and wherein the step of evaluating evaluates a suspected region based on rule-based selection criteria from the database.

53. A method as in claim 45, wherein the step of identifying comprises the steps of, in the digitized radiograph:

detecting and enhancing skin boundaries;

non-linear thresholding, based on digital values in the digitized radiograph;

blob labelling and size selecting of suspected regions; and

framing and locating suspected regions in the digitized radiograph.

54. A method as in claim 45, wherein the radiograph is an X-ray mammogram, and wherein the multiple topographic layers comprise top-view radiographic topographic layers.

55. A method as in claim 54, wherein the multiple topographic layers comprises three topographic layers.

56. A method as in claim 45, wherein the radiograph is an X-ray mammogram, and wherein the abnormal regions are masses.

57. A method as in claim 45, wherein the radiograph is an X-ray mammogram, and wherein the abnormal regions are microcalcifications.

58. The method of claim 45, wherein the topographic layers are extracted by thresholding the digital radiograph at a particular threshold value for each topographic layer.

59. The method of claim 58, wherein the threshold value for each other layer is based on the threshold and contrast values in a previous layer.

60. The method of claim 58, wherein the threshold value T_j for layer j is a function of threshold value T_j-1 and contrast value C_j-1 in layer j-1 as given by:

space="preserve" listing-type="equation">T.sub.j =T.sub.j-1 +256×

C.sub.j-1.

61. The method of claim 58, wherein the threshold value of the first layer is a function of a digital value of a minimum point in the suspected region.

62. The method of claim 61, wherein the threshold value of the first layer is given by:

space="preserve" listing-type="equation">T.sub.1 =I.sub.min +I.sub.min /10, when I.sub.min ≦

500;

space="preserve" listing-type="equation">T.sub.1 =I.sub.min ×

1.0246+37.68, when (500<

I.sub.min <

2000); and

space="preserve" listing-type="equation">T.sub.1 =I.sub.min +I.sub.min /23, when I.sub.min ≧

2000,

where I_min is the digital value of the minimum point in the suspected region.

63. A method of detecting a mass in living tissue depicted in a mammogram, the method comprising the steps of:

(a) digitizing the mammogram to produce a digitized mammogram;

(b) identifying, as a suspected region, according to predetermined criteria, a suspected mass in the mammogram;

(c) extracting multiple topographic layers of the suspected region from the digitized mammogram, wherein all layers are extracted from the same mammogram; and

(d) evaluating the extracted multiple topographic layers to determine whether the suspected region is a mass.

64. A method as in claim 63, wherein the multiple topographic layers comprise top-view radiographic topographic layers.

65. A method as in claim 63, wherein the step of identifying comprises one of the steps of, in the digitized mammogram,

single-image segmentation and

bilateral-image subtraction.

66. A method as in claim 63, wherein the step of identifying comprises the steps

obtaining first and second low-pass filtered images of the digitized mammogram;

creating a binary image from subtraction and thresholding of the first and second low-pass filtered images; and

labelling a blob and determining a suspicious area in the binary image.

67. The method of claim 63, wherein the topographic layers are extracted by thresholding the digital radiograph at a particular threshold value for each topographic layer.

68. The method of claim 67, wherein the threshold value for each other layer is based on the threshold and contrast values in a previous layer.

69. The method of claim 67, wherein the threshold value T_j for layer j is a function of threshold value T_j-1 and contrast value C_j-1 in layer j-1 as given by:

space="preserve" listing-type="equation">T.sub.j =T.sub.j-1 +256×

C.sub.j-1.

70.

70. An apparatus for detecting an abnormal region in living tissue depicted in a radiograph, the apparatus comprising:

a digitizer for digitizing the radiograph to produce a digitized radiograph;

means for identifying, according to predetermined criteria, a suspected abnormal region in the radiograph;

means for extracting multiple topographic layers of the identified suspected abnormal region from the digitized radiograph, all layers being extracted from the same radiograph; and

means for evaluating the multiple topographic layers extracted from the same radiograph to determine whether the identified suspected abnormal region is an abnormal region.

71. An apparatus as in claim 70, wherein the extracting means extracts layers based on a threshold value and a contrast value for a previous layer.

72. An apparatus for detecting a mass in living tissue regions depicted in a mammogram, the apparatus comprising:

a digitizer for digitizing the mammogram to produce a digitized mammogram;

means for identifying, as a suspected abnormal region, according to predetermined criteria, a suspected mass in the mammogram;

means for extracting multiple topographic layers of the identified suspected abnormal region from the digitized mammogram, all layers being extracted from the same mammogram; and

means for evaluating the multiple topographic layers extracted from the same mammogram to determine whether the suspected region depicts a mass.

73. A method of detecting a cluster of microcalcifications in living tissue depicted in a mammogram, the method comprising the steps of:

(a) digitizing the mammogram to produce a digitized mammogram;

(b) identifying, as a suspected region, according to predetermined criteria, a suspected cluster of microcalcifications in the digitized mammogram;

(c) extracting multiple topographic layers of the suspected region from the digitized mammogram, wherein all layers are extracted from the same mammogram; and

(d) evaluating the extracted multiple topographic layers to determine whether the suspected region depicts a cluster of microcalcifications.

74. A method as in claim 73, wherein the step of evaluating comprises the steps of:

for each layer of the multiple topographic layers, determining features of the region in the layer, and

analyzing determined features of the region for different topographic layers of the multiple topographic layers.

75. A method as in claim 73, wherein the step of evaluating comprises the steps of:

for at least two topographic layers of the multiple topographic layers, determining features of the region in the at least two topographic layers, and

analyzing the determined features of the region for different topographic layers of the at least two topographic layers.

76. A method of claim 75, wherein the different topographic layers are adjacent.

77. A method of claim 75, wherein the different topographic layers are adjacent.

78. A method as in claim 73, wherein the step of evaluating multiple topographic layers comprises the steps of:

determining, for each region, a rate of change in elevation of the region in a plurality of directions; and

selecting, based on said determining, whether the region depicts a cluster of microcalcifications.

79. A method as in claim 73, further comprising the step of generating a rule-based criteria database, and wherein the step of evaluating evaluates a suspected region based on rule-based selection criteria from the database.

80. A method as in claim 73, wherein the step of identifying comprises the steps of:

lowpass filtering the digital mammogram using two different filters to produce two, lowpass filtered images;

subtracting the two lowpass filtered images;

non-linear thresholding the subtracted image, based on digital values in the digital mammogram;

blob labelling and size selecting of a suspected regions;

blob clustering and blob selecting of selected suspected regions.

81. The method of claim 73, wherein the topographic layers are extracted by thresholding an image formed by subtracting two lowpass filtered versions of the digital radiograph at a particular threshold value for each topographic layer.

82. The method of claim 81, wherein the threshold value for each other layer is based on a digital value of a pixel in the digitized mammogram and on a predetermined contrast value.

83. The method of claim 81, wherein the predetermined contrast value is one of 1%, 1.5%, and 2%.