Granted Patent B2

US 7,593,561 · App. 11/284,570 · Granted Sep 22, 2009

Abstract

Claims (25)

1. A method for computer-aided detection of microcalcification clusters using a digital computer, comprising steps of:

a) obtaining digital mammography data for a single view image;

b) normalizing the single view image data;

c) filtering the single view image data to reduce the image noise;

d) generating and applying a first mask to the single view image data for defining the breast structure, forming a first cropped image thereby and generating and applying a second mask to the single view image data for defining the muscle structure, forming a second cropped image thereby;

e) generating, for each of the first and second cropped images, an artifact mask corresponding to vascular calcifications and known imaging artifacts;

f) applying, to each of the first and second cropped images, the corresponding artifact mask, defining corresponding first and second artifact-masked cropped images thereby;

g) processing portions of each artifact-masked cropped image by:

(g1) applying an enhancement algorithm for enhancing small particles in the artifact-masked cropped image, forming an enhanced artifact-masked cropped image thereby;

(g2) reducing edge effects introduced by artifacts in the enhanced artifact-masked cropped image; and

(g3) obtaining a set of microcalcification cluster candidates according to a segmentation threshold;

h) repeating step g) at least one time for each of the first and second artifact-masked cropped images, using an alternate enhancement algorithm in sub-step (g1);

i) combining results from steps g) and h) to form a listing of suspected microcalcification clusters;

j) applying image processing algorithms to remove false positives from the listing of suspected microcalcification clusters to obtain an initial listing of candidate microcalcification clusters;

k) extracting features of each microcalcification in the candidate microcalcification cluster and classifying each microcalcification to remove false positive microcalcifications;

l) regrouping the remaining microcalcifications to form a regrouped listing of candidate microcalcification clusters; and

m) extracting features of each microcalcification cluster in the regrouped listing of candidate microcalcification clusters and classifying each microcalcification cluster in the regrouped microcalcification cluster listing to identify true positives.

2. The method according to claim 1 , wherein the step of generating the artifact mask comprises the step of identifying structures exceeding a predetermined boundary size.

3. The method according to claim 1 , wherein the step of generating the artifact mask comprises one or more of the following steps:

(i) detecting artifacts having characteristics of vascular tissue structures;

(ii) detecting large benign calcifications; and

(iii) detecting foreign objects in the image.

4. The method according to claim 1 , wherein the step of applying an enhancement algorithm in step g(1) comprises the step of applying a linear transformation.

5. The method according to claim 1 , wherein the step of applying an enhancement algorithm in step g(1) comprises the step of applying a non-linear transformation.

6. The method according to claim 1 , wherein the step of generating a first mask comprises the step of using image contrast characteristics.

Assignments (9)