Line structure detection and analysis for mammography CAD

1. A method for image linear structure detection in medical imaging, executed using a processor, comprising:

locating microcalcification (mcc) candidate spots in a mammographic image;

forming candidate clusters;

assigning ranks to the candidate clusters;

identifying linear structures in the neighborhood where the candidate clusters reside; and

altering the ranks of the candidate clusters for which linear structures have been identified in the neighborhood.

2. The method of claim 1 wherein the step of locating microcalcification (mcc) candidate spots further comprises:

providing a plurality of image processing and computer vision procedures that find connected pixels that present characteristics which are similar to that of microcalcification in mammogram; and

grouping the connected pixels into microcalcification candidate spots.

3. The method of claim 1 wherein the step of forming candidate clusters further comprises:

grouping a plurality of microcalcification candidate spots that are close to each other within a certain distance into a cluster; and

attaching relevant parameters to the cluster.

4. The method of claim 1 wherein the step of assigning ranks to the candidate clusters further comprises:

generating classifiers that are trained with a plurality of true microcalcification clusters and non-cancer clusters; and

applying trained classifiers to the microcalcification candidate clusters and assigning ranks to classified clusters.

5. The method of claim 1 the step of identifying linear structures further comprises:

applying image processing and computer vision algorithms to image pixels including the cluster and its neighborhood;

extracting information related to linear structures from the processed image pixels; and

applying rules to extracted information to confirm the presence of linear structures.

6. The method of claim 1 wherein the step of altering the ranks of the candidate clusters for which linear structures have been identified in the neighborhood comprises:

evaluating the positions of the microcalcification candidate spots relative to the positions of the linear structures; and

altering the ranks of the candidate clusters if the number of microcalcification candidate spots falling on the linear structures exceeds a certain value for the candidate clusters.

7. A method for identifying linear structures that associated with mcc candidate clusters in mammographic images, executed using a processor, comprising:

locating microcalcification (mcc) candidate spots in a mammographic image;

forming microcalcification candidate clusters;

generating classifiers that are trained with a plurality of true microcalcification clusters and non-cancer clusters;

applying the trained classifiers to the microcalcification candidate clusters and assigning ranks to the classifiers clusters;

assigning ranks to the microcalcification candidate clusters;

determining a region of interest (ROI);

forming a line mask in the ROI; and

updating the ranks of the microcalcification candidate clusters.

8. The method of claim 7 comprising the steps of:

providing a plurality of image processing and computer vision procedures that find connected pixels that present characteristics which are similar to that of microcalcification in mammogram; and

grouping the connected pixels into microcalcification candidate spots.

9. The method of claim 7 comprises the steps of:

grouping a plurality of microcalcification candidate spots that are close to each other within a certain distance into a cluster; and

attaching relevant parameters to the cluster.

10. The A method of a microcalcification (mcc) candidate cluster driven linear structure detection in mammographic images, executed using a processor, comprising:

locating a plurality of mcc candidate clusters in digital mammograms;

extracting a region of interest (ROIs) in the mammographic image;

processing the ROIs to generate line masks;

analyzing the ROIs to verify the truthfulness of the generated line masks; and

tagging the mcc candidate clusters.

11. The method of claim 7 further comprising the steps of:

determining a candidate cluster whose rank is not within a specified range; and

using the positions of the candidate spots of the candidate cluster to define a region (ROI) that encloses the candidate spots in the mammographic image.

12. The method of claim 7 further comprising the steps of:

applying image processing and computer vision algorithms to image pixels within the ROI;

extracting information related to linear structures from the processed image pixels;

applying rules to extracted information to confirm the presence of linear structures; and

forming a line mask for the linear structures.

13. The method of claim 7 further comprising the steps of:

evaluating the positions of the microcalcification candidate spots relative to the positions of the line mask; and

updating the ranks of the candidate clusters if the number of microcalcification candidate spots falling in the line mask exceeds a certain value for the candidate clusters.

14. The method of claim 10 comprises the steps of:

generating classifiers that are trained with a plurality of true microcalcification clusters and non-cancer clusters; and

applying trained classifiers to the microcalcification candidate clusters and assigning ranks to classified clusters.

15. The method of claim 10 wherein the step of locating a plurality of mcc candidate clusters further comprises:

providing a plurality of image processing and computer vision procedures that find connected pixels that present characteristics which are similar to that of microcalcification in mammogram;

grouping the connected pixels into microcalcification candidate spots;

grouping a plurality of microcalcification candidate spots that are close to each other within a certain distance into a cluster; and attaching relevant parameters to the cluster.

16. The method of claim 10 wherein the step of extracting ROIs in the mammographic image further comprises:

determining a candidate cluster whose rank is not within a specified range; and

employing the positions of the candidate spots of the candidate cluster to define a region (ROI) that encloses the candidate spots in the mammographic image.

17. The method of claim 10 wherein the step of processing the ROIs to generate line masks further comprises:

applying image processing and computer vision algorithms to image pixels within the ROI;

extracting information related to linear structures from the processed image pixels;

applying rules to extracted information to confirm the presence of linear structures; and

forming a line mask for the linear structures.

18. The method of claim 10 wherein the step of analyzing the ROIs to verify the truthfulness of the generated line masks comprises:

applying image transformation/processing algorithms to pixels in the ROIs; and

analyzing the transformed/processed ROIs in the transformed/processed image domain to confirm the existence of linear structures that validate the generated line masks.

19. The method of claim 10 wherein the step of tagging the mcc candidate clusters comprises:

evaluating the positions of the microcalcification candidate spots relative to the position of the line mask; and

updating the status of the candidate clusters if the number of microcalcification candidate spots falling on the line masks exceeds a certain value for the candidate clusters.