Joint optimization of parameters for the detection of clustered microcalcifications in digital mammograms

• US 6,389,157 B2
• Filed: 01/11/2001
• Issued: 05/14/2002
• Est. Priority Date: 08/28/1997
• Status: Expired due to Fees

• Current Assignee: iCAD Incorporated
• Original Assignee: Qualia Computing Inc. (iCAD Incorporated)
• Inventors: Rathbun, Thomas F., Rosenstengel, John E., Rogers, Steven K., Broussard, Randy P., Ochoa, Edward M.
• Primary Examiner(s): Johns, Andrew W.
• Application Number:
  US09/758,889
• Publication Number:
  US 20010008562A1
• Time in Patent Office: 488 Days
• Field of Search: 382/128, 382/132, 382/155, 382/156, 382/206, 382/225, 382/227, 382/256, 382/257, 382/260, 382/270, 382/272, 128/922, 378/37, 706/13, 706/924
• US Class Current: 382/132
• CPC Class Codes: A61B 6/502 for diagnosis of breast, i.e. mammography
• CPC Class Codes: B25J 15/04 with provision for the remote detachment or exchange of the head or parts thereof
• CPC Class Codes: B29C 2791/006 Using vacuum
• CPC Class Codes: B29C 2791/007 Using fluid under pressure
• CPC Class Codes: B29C 49/0021 using moulds or mould parts movable in a closed path, e.g. mounted on movable endless supports
• CPC Class Codes: B29L 2023/18 Pleated or corrugated hoses
• CPC Class Codes: G06F 18/41 Interactive pattern learning with a human teacher
• CPC Class Codes: G06T 5/20 using local operators
• CPC Class Codes: G06T 7/0012 Biomedical image inspection
• CPC Class Codes: G06V 10/28 Quantising the image, e.g. histogram thresholding for discrimination between background and foreground patterns
• CPC Class Codes: G06V 10/443 by matching or filtering
• CPC Class Codes: G06V 2201/03 Recognition of patterns in medical or anatomical images
• CPC Class Codes: Y10S 128/922 including image analysis

Abstract

A method and system for detecting and displaying clustered microcalcifications in a digital mammogram, wherein a single digital mammogram is first automatically cropped to a breast area sub-image which is then processed by means of an optimized Difference of Gaussians filter to enhance the appearance of potential microcalcifications in the sub-image. The potential microcalcifications are thresholded, clusters are detected, features are computed for the detected clusters, and the clusters are classified as either suspicious or not suspicious by means of a neural network. Thresholding is preferably by sloping local thresholding, but may also be performed by global and dual-local thresholding. The locations in the original digital mammogram of the suspicious detected clustered microcalcifications are indicated. Parameters for use in the detection and thresholding portions of the system are computer-optimized by means of a genetic algorithm. The results of the system are optimally combined with a radiologist'"'"'s observation of the original mammogram by combining the observations with the results, after the radiologist has first accepted or rejected individual detections reported by the system.

Claims

1. A method for detecting clustered microcalcifications in a digital mammogram image by digital processing comprising, in sequence, the steps of:

a) optimizing a value for a predetermined filtering algorithm parameter;

b) filtering a digital mammogram image to produce a filtered image comprising essentially microcalcifications;

c) thresholding the filtered image from step b) to produce a thresholded image;

d) forming clusters of microcalcifications in the thresholded image from step c); and

e) inputting the identified clusters of microcalcifications to a classifier to distinguish suspicious clusters of microcalcifications from nonsuspicious clusters of microcalcifications;

wherein said step a) comprises;

i) providing an optimizing algorithm;

ii) providing a database of training images; and

iii) inputting the database of training images into said optimizing algorithm to determine an optimum value for said predetermined filtering algorithm parameter; and

wherein said step b) comprises;

i) providing a filtering algorithm incorporating the optimized predetermined filtering algorithm parameter; and

ii) applying the filtering algorithm to the digital mammogram image to produce a filtered image comprising essentially microcalcifications.

2. The method of claim 1 wherein said step a) comprises optimizing a parameter corresponding to the size of microcalcifications in the digital mammogram image.

3. A method for detecting clustered microcalcifications in a digital mammogram image by digital processing comprising, in sequence, the steps of:

a) optimizing at least one value for a predetermined threshold parameter;

b) filtering a digital mammogram image to produce a filtered image comprising essentially microcalcifications;

c) thresholding the filtered image from step b) to produce a threshold image;

d) forming clusters of microcalcifications in the filtered image from step c); and

e) inputting the identified clusters of microcalcifications to a classifier to distinguish suspicious clusters of microcalcifications from nonsuspicious clusters of microcalcifications;

wherein said step a) comprises;

i) providing an optimizing algorithm;

ii) providing a database of training images; and

iii) inputting the database of training images into said optimizing algorithm to determine optimum values for said predetermined threshold parameters; and

wherein said step c) comprises;

i) providing a thresholding algorithm incorporating the optimized predetermined threshold parameter; and

ii) applying the thresholding algorithm to the filtered image to identify potential microcalcifications.

4. The method of claim 3 wherein said step a) comprises computing the histogram of the filtered image and optimizing a parameter corresponding to an upper fraction of the histogram.

5. The method of claim 4 wherein said parameter corresponding to an upper fraction of the histogram comprises a parameter related to a maximum gray level in the filtered image.

6. The method of claim 3 wherein said step a) comprises optimizing a local threshold parameter set for determining a local threshold and optimizing a local neighborhood parameter for determining a local neighborhood size within which the local threshold is applied.

7. The method of claim 3 wherein:

said step a) further comprises;

A) computing the histogram of the filtered image and optimizing a parameter, f, corresponding to an upper fraction of the histogram;

B) optimizing a local threshold parameter set for determining a local threshold; and

C) optimizing a local neighborhood parameter for determining a local neighborhood size within which the local threshold is applied; and

said step c) comprises;

A) producing a globally thresholded image by retaining an upper fraction of the pixels in the filtered image; and

B) producing a locally thresholded image by locally thresholding pixels in a window of the filtered image, determined by the local neighborhood parameter, and retaining the pixels in the window whose gray levels satisfy a relation determined by the local threshold parameter set.

8. The method of claim 7 including the step of logically ANDing the globally thresholded image and the locally thresholded image.

9. A method for detecting clustered microcalcifications in a digital mammogram image by digital processing comprising, in sequence, the steps of:

a) optimizing a value for at least one predetermined clustering algorithm parameter;

b) filtering a digital mammogram image to produce a filtered image comprising essentially microcalcifications;

c) thresholding the filtered image from step b) to produce a thresholded image;

d) forming clusters of microcalcifications in the thresholded image from step c); and

e) inputting the identified clusters of microcalcifications to a classifier to distinguish suspicious clusters of microcalcifications from nonsuspicious clusters of microcalcifications;

wherein said step a) comprises;

i) providing an optimizing algorithm;

ii) providing a database of training images; and

iii) inputting the database of training images into said optimizing algorithm to determine optimum values for said predetermined clustering algorithm parameters; and

wherein said step d) comprises;

i) providing a clustering algorithm incorporating the optimized predetermined clustering algorithm parameter; and

ii) applying the clustering algorithm to the thresholded image to identify potential clusters of microcalcifications.

10. The method of claim 9 wherein said step a) comprises optimizing a parameter, d_nn, corresponding to a maximum distance between microcalcifications.

11. The method of claim 9 wherein said step a) comprises optimizing a parameter, μ

Cs_min, corresponding to a minimum number of microcalcifications for forming a cluster.

12. The method of claim 9 wherein said step a) comprises:

A) optimizing a parameter, d_nn, corresponding to a maximum distance between microcalcifications in a cluster; and

B) optimizing a parameter, μ

Cs_min, corresponding to a minimum number of microcalcifications for forming a cluster.

13. The method of claim 12 wherein said step d) comprises identifying groups of microcalcifications in which at least μ

Cs_min microcalcifications are within distance d_nn of each other to define the clusters of microcalcifications.

14. A method for detecting clustered microcalcifications in a digital mammogram image by digital processing comprising, in sequence, the steps of:

a) jointly optimizing a filtering algorithm parameter, a thresholding algorithm parameter and a clustering algorithm parameter;

b) providing a filtering algorithm incorporating the optimized filtering algorithm parameter and applying the filtering algorithm to the digital mammogram image to produce a filtered image comprising essentially microcalcifications;

c) providing a thresholding algorithm incorporating the optimized thresholding algorithm parameter and applying the thresholding algorithm to the filtered image to produce a thresholded image;

d) providing a clustering algorithm incorporating the optimized clustering algorithm parameter and applying the clustering algorithm to the thresholded image to identify potential clusters of microcalcifications; and

e) inputting the identified clusters of microcalcifications to a classifier to distinguish suspicious clusters of microcalcifications from non-suspicious clusters of microcalcifications.

15. The method of claim 14 wherein said step a) comprises:

i) providing an optimizing algorithm;

ii) providing a database of training images; and

iii) inputting the database of training images into said optimizing algorithm to determine an optimum value for each of said algorithm parameters.

16. The method of claim 14 wherein said step of optimizing the filtering algorithm parameter comprises optimizing a parameter corresponding to the size of microcalcifications in the digital mammogram image.

17. The method of claim 14 wherein said step of optimizing the thresholding algorithm parameter comprises computing the histogram of the filtered image and optimizing a parameter corresponding to an upper fraction of the histogram.

18. The method of claim 17 wherein said parameter corresponding to an upper fraction of the histogram comprises a parameter related to a maximum gray level in the filtered image.

19. The method of claim 14 wherein said step of optimizing a thresholding algorithm parameter comprises optimizing a local threshold parameter set for determining a local threshold and optimizing a local neighborhood parameter for determining a local neighborhood size within which the local threshold is applied.

20. The method of claim 14 wherein:

said step of optimizing a thresholding algorithm parameter further comprises;

i) computing the histogram of the filtered image and optimizing a parameter, f, corresponding to an upper fraction of the histogram;

ii) optimizing a local threshold parameter set for determining a local threshold; and

iii) optimizing a local neighborhood parameter for determining a local neighborhood size within which the local threshold is applied; and

said step c) comprises;

(i) producing a globally thresholded image by retaining an upper fraction of the pixels in the filtered image; and

(ii) producing a locally thresholded image by locally thresholding pixels in a window of the filtered image, determined by the local neighborhood parameter, and retaining the pixels in the window whose gray levels satisfy a relation determined by the local threshold parameter set.

21. The method of claim 20 including the step of logically ANDing the globally thresholded image and the locally thresholded image.

22. The method of claim 14 wherein said step of optimizing the clustering algorithm parameter comprises optimizing a parameter d_nn, corresponding to a maximum distance between microcalcifications.

23. The method of claim 14 wherein said step of optimizing the clustering algorithm parameter comprises optimizing a parameter, μ

Cs_min, corresponding to a minimum number of microcalcifications for forming a cluster.

24. The method of claim 14 wherein said step of optimizing the clustering algorithm parameter comprises:

i) optimizing a parameter, d_nn, corresponding to a maximum distance between microcalcifications in a cluster; and

ii) optimizing a parameter, μ

Cs_min, corresponding to a minimum number of microcalcifications for forming a cluster.

25. The method of claim 24 wherein said step d) comprises identifying groups of microcalcifications in which at least μ

Cs_min, microcalcifications are within distance d_nn, of each other to define the clusters of microcalcifications.