METHOD FOR SEGMENTATION OF LESIONS

• US 20090129673A1
• Filed: 11/15/2007
• Published: 05/21/2009
• Est. Priority Date: 11/15/2007
• Status: Active Grant

• Current Assignee: Philips Healthcare Informatics Incorporated (Koninklijke Philips N.V.)
• Original Assignee: CareStream Health Incorporated
• Inventors: Ray, Lawrence A., Simon, Richard A., Gindele, Edward B.
• Granted Patent: US 8,144,949 B2
• Time in Patent Office: Days
• US Class Current: 382/173
• CPC Class Codes: G06T 2207/10072 Tomographic images
• CPC Class Codes: G06T 2207/30096 Tumor; Lesion
• CPC Class Codes: G06T 2207/30101 Blood vessel; Artery; Vein; Vascular
• CPC Class Codes: G06T 7/0012 Biomedical image inspection
• CPC Class Codes: G06T 7/11 Region-based segmentation

Abstract

A method of image segmentation includes receiving a set of voxels, segmenting the set of voxels into a foreground group and a background group, and classifying voxels of the foreground group as either lesion voxels or normal anatomy voxels. The method also includes blocking the normal anatomy voxels and performing a second segmentation on voxels of the background group and the lesion voxels, the second segmentation forming a stage two foreground group comprising the lesion voxels and a portion of the voxels of the background group. The method further includes classifying voxels of the stage two foreground group as either stage two lesion voxels or stage two normal anatomy voxels.

Claims

1. A method of image segmentation, comprising:

a) receiving a set of voxels;

b) segmenting the set of voxels into a foreground group and a background group;

c) classifying voxels of the foreground group as either lesion voxels or normal anatomy voxels;

d) blocking the normal anatomy voxels;

e) performing a second segmentation on voxels of the background group and the lesion voxels, the second segmentation forming a stage two foreground group comprising the lesion voxels and a portion of the voxels of the background group; and

f) classifying voxels of the stage two foreground group as either stage two lesion voxels or stage two normal anatomy voxels.

2. The method of claim 1, wherein the second segmentation is more aggressive than the segmentation of step b.

3. The method of claim 1, further including blocking the stage two normal anatomy voxels, performing a third segmentation on voxels of the background group and the stage two lesion voxels, the third segmentation forming a stage three foreground group comprising the stage two lesion voxels and a different portion of the voxels of the background group, and classifying voxels of the stage three foreground group as either stage three lesion voxels or stage three normal anatomy voxels.

4. The method of claim 1, further comprising calculating a volume of a lesion associated with the lesion voxels or the stage two lesion voxels.

5. The method of claim 1, further comprising limiting the segmentation of step b with a first set of stopping conditions.

6. The method of claim 5, wherein the first set of stopping conditions is a Hounsfield unit threshold.

7. The method of claim 5, further comprising limiting the second segmentation with a second set of stopping conditions different than the first set.

8. The method of claim 1, further comprising creating a composite class map based on the segmentation of step b and the second segmentation.

9. The method of claim 1, wherein blocking the normal anatomy voxels comprises excluding a plurality of voxels adjacent to the normal anatomy voxels from the second segmentation.

10. A method of image segmentation, comprising:

a) receiving a set of voxels that were segmented into at least three classes by previous voxel segmentation;

b) defining a region for a further voxel segmentation, the region excluding voxels of one of the at least three classes;

c) performing the further voxel segmentation within the region, the further voxel segmentation being more aggressive than the previous voxel segmentation, the further voxel segmentation separating voxels in the region into at least two of the at least three classes; and

d) creating a composite class map based on the previous voxel segmentation and the further voxel segmentation.

11. The method of claim 10, further comprising identifying a seed point for the further voxel segmentation based on the previous voxel segmentation.

12. The method of claim 10, wherein creating the composite class map comprises rectifying conflicting voxel classifications.

13. The method of claim 10, wherein the first and second sets of stopping conditions are Hounsfield unit thresholds.

14. The method of claim 10, further comprising calculating a volume of a lesion based on a final composite class map.

15. The method of claim 10, wherein the region excludes a plurality of voxels adjacent to a normal anatomy.

16. The method of claim 10, wherein the region excludes voxels associated with more than one anatomical structure.

17. The method of claim 10, further comprising repeating steps b through e at least once.

18. The method of claim 10, wherein the at least three classes comprises a lesion class, a normal anatomy class, and a background class.

19. A method of image segmentation, comprising:

a) performing a first segmentation capable of distinguishing between a first structure and a second structure but incapable of determining a full extent of the first structure;

b) defining a region for a second segmentation, the region excluding the second structure; and

c) performing the second segmentation within the region, the second segmentation capable of determining a greater extent of the first structure than the first segmentation.

20. The method of claim 19, further including

defining a second region for a further segmentation, the second region excluding the first structure, and

performing the further segmentation within the second region.

21. The method of claim 19, wherein the first segmentation uses a first algorithm limited by a first set of stopping conditions and the second segmentation uses a second algorithm limited by a second set of stopping conditions more aggressive than the first.

22. The method of claim 19, further comprising creating a composite class map based on the first and second segmentations.