Systems and methods for generating an imaging biomarker that indicates detectability of conspicuity of lesions in a mammographic image

• US 10,595,805 B2
• Filed: 06/26/2015
• Issued: 03/24/2020
• Est. Priority Date: 06/27/2014
• Status: Active Grant

• Current Assignee: Sunnybrook Research Institute
• Original Assignee: Sunnybrook Research Institute
• Inventors: Mainprize, James, Alonzo-Proulx, Olivier, Yaffe, Martin, Mawdsley, Gordon
• Primary Examiner(s): Gedeon, Brian T
• Application Number:
  US15/321,525
• Publication Number:
  US 20170202530A1
• Time in Patent Office: 1,733 Days
• Field of Search: None
• CPC Class Codes: A61B 6/032 Transmission computed tomography [CT]
• CPC Class Codes: A61B 6/12 Arrangements for detecting or locating foreign bodies
• CPC Class Codes: A61B 6/461 Displaying means of special interest
• CPC Class Codes: A61B 6/469 for selecting a region of interest [ROI]
• CPC Class Codes: A61B 6/481 involving the use of contrast agents
• CPC Class Codes: A61B 6/502 for diagnosis of breast, i.e. mammography
• CPC Class Codes: G16H 30/20 for handling medical images, e.g. DICOM, HL7 or PACS
• CPC Class Codes: G16H 30/40 for processing medical images, e.g. editing

Abstract

Systems and methods for generating imaging biomarkers that indicate detectability or conspicuity of lesions that may be present in mammographic images are provided. In general, a task-based measure of a signal-to-noise ratio (“SNR”) measurement of a detection task is computed over a number of small regions-of-interest (“ROIs”) in an image, and the computed parameter is used to predict what detection rates should be if a lesion was present in the image. As such, the computed parameter can be used to define an imaging biomarker that is a “masking measure” that indicates the degree of conspicuity of lesions that may be present in a given a mammographic image.

Claims

1. A method for generating an imaging biomarker that indicates a degree of lesion masking in a mammographic image obtained with an x-ray imaging system, the steps of the method comprising:

(a) providing a mammographic image acquired with an x-ray imaging system;

(b) processing the mammographic image to estimate a task-based measure that is based on a statistical measure of a detection task;

(c) generating an imaging biomarker that indicates a degree of lesion masking in mammographic image based on the estimated task-based measure; and

(d) generating a report that indicates the degree of lesion masking in the mammographic image based on the generated imaging biomarker.

2. The method as recited in claim 1, wherein step (b) includes selecting a region-of-interest (ROI) in the mammographic image and estimating the task-based measure for the selected ROI.

3. The method as recited in claim 2, wherein step (b) is repeated to select a plurality of ROIs in the mammographic image.

4. The method as recited in claim 3, wherein at least some of the plurality of ROIs are arranged to be overlapping with others of the plurality of ROIs.

5. The method as recited in claim 2, wherein the task-based measure is based on a signal-to-noise ratio (SNR) of the detection task.

6. The method as recited in claim 5, wherein the task-based measure is estimated using a model observer.

7. The method as recited in claim 6, wherein the model observer is at least one of a pre-whitening model observer, a non-pre-whitening model observer, a Hotelling model observer, or a channelized Hotelling model observer.

8. The method as recited in claim 6, wherein the model observer is a modified model observer that accounts for at least one of human visual response, human decision making, or human skill.

9. The method as recited in claim 6, wherein the task-based measure is based on a resolution-based measure of the x-ray imaging system, a task function that describes the detection task, and a noise-based measure that is estimated for the ROI.

10. The method as recited in claim 9, wherein the resolution-based measure is at least one of a point spread function and a modulation transfer function.

11. The method as recited in claim 9, wherein the noise-based measure is at least one of a normalized noise power spectrum or a noise covariance matrix.

12. The method as recited in claim 9, wherein the noise-based measure is estimated based on image intensity values in the ROI.

13. The method as recited claim 9, wherein the noise-based measure is estimated based on a system model of x-ray propagation.

14. The method as recited in claim 9, wherein the task function describes a detection of a lesion having at least one of a selected size, a selected shape, or a selected composition.

15. The method as recited in claim 1, wherein the generated imaging biomarker indicates a degree of difficulty for the detection task to detect whether lesion may be present in the mammographic image without identifying a presence of lesions in the mammographic image.

16. The method as recited in claim 1, wherein the generated imaging biomarker indicates a probability of missing a detection of a lesion.

17. The method as recited in claim 1, wherein generating the imaging biomarker includes computing a secondary measure of conspicuity of lesions that may be present in the mammographic image.

18. The method as recited in claim 17, wherein the secondary measure of conspicuity of lesions includes at least one of fractional area of increased lesion masking probability, overall lesion masking probability, or heterogeneity of lesion masking probability in the mammographic image.

19. The method as recited in claim 1, wherein generating the imaging biomarker includes computing a percentage risk of missing a detection of a lesion.

20. The method as recited in claim 1, wherein generating the imaging biomarker includes re-expressing the task-based measure as a breast imaging-reporting and data systems (BIRADS) equivalent scale measure.

21. The method as recited in claim 1, wherein the report generated in step (d) includes a map that relates values of the imaging biomarker to a colormap.

22. The method as recited in claim 1, further comprising providing the generated imaging biomarker to a computer-aided detection system.

23. The method as recited in claim 1, wherein the detection task indicates the detectability of lesions that may be present in the mammographic image by modeling the lesions that may be present in the mammographic image as at least one of a disk, a Gaussian profile blob, a synthetic spiculated breast cancer lesion, a single microcalcification, or a cluster of microcalcifications.

24. The method as recited in claim 1, wherein the x-ray imaging system is at least one of a digital mammography system, a digital tomosynthesis system, or a dual-energy mammography system.

25. The method as recited in claim 1, wherein step (c) includes calculating an average local detectability value across the mammographic image.

26. The method as recited in claim 1, wherein step (c) includes calculating a local detectability value for each quadrant of the mammographic image.

27. The method as recited in claim 1, wherein step (c) includes calculating a weighted average local detectability value for each quadrant.

28. The method as recited in claim 1, wherein step (c) includes calculating a local detectability value for each quadrant of the mammographic image and identifying the quadrant of the mammographic image having the lowest local detectability value.

29. The method as recited in claim 1, wherein step (c) includes calculating a local detectability value at each voxel location in the mammographic image and calculating a percent area of the mammographic image associated with local detectability values that are below a threshold value.