Systems and methods for identifying bone marrow in medical images

• US 9,058,665 B2
• Filed: 12/30/2009
• Issued: 06/16/2015
• Est. Priority Date: 12/30/2009
• Status: Active Grant

• Current Assignee: General Electric Company
• Original Assignee: General Electric Company
• Inventors: Bar-Shalev, Avi, Manjeshwar, Ravindra Mohan
• Primary Examiner(s): Holcomb, Mark
• Assistant Examiner(s): Ng, Jonathan K
• Application Number:
  US12/650,036
• Publication Number:
  US 20110158494A1
• Time in Patent Office: 1,994 Days
• Field of Search: 382128-132
• US Class Current: 1/1
• CPC Class Codes: A61B 6/032 Transmission computed tomography [CT]
• CPC Class Codes: A61B 6/505 for diagnosis of bone
• CPC Class Codes: A61B 6/508 for non-human patients
• CPC Class Codes: A61B 6/5241 combining overlapping images of the same imaging modality, e.g. by stitching
• CPC Class Codes: G06F 18/00 Pattern recognition
• CPC Class Codes: G06T 2207/10081 Computed x-ray tomography [CT]
• CPC Class Codes: G06T 2207/20164 Salient point detection; Corner detection
• CPC Class Codes: G06T 2207/30008 Bone
• CPC Class Codes: G06T 7/0012 Biomedical image inspection
• CPC Class Codes: G06T 7/11 Region-based segmentation
• CPC Class Codes: G06T 7/13 Edge detection
• CPC Class Codes: G06T 7/187 involving region growing; involving region merging; involving connected component labelling

Abstract

Systems and method for identifying bone marrow in medical images are provided. A method includes obtaining a three-dimensional (3D) computed tomography (CT) volume data set corresponding to an imaged volume and identifying voxels in the 3D CT volume data set having a Hounsfield Unit (HU) value below a bone threshold. The voxels are identified without using image continuity. The method further includes marking the identified voxels as non-bone voxels, determining definite tissue voxels based on the identified non-bone voxels and expanding a region defined by the definite tissue voxels. The method also includes segmenting the expanded region to identify bone voxels and bone marrow voxels and identifying bone marrow as voxels that are not the bone voxels.

Claims

1. A method for identifying bone marrow in medical image data, the method comprising:

obtaining a three-dimensional (3D) computed tomography (CT) volume data set corresponding to an imaged volume, the 3D CT volume data set acquired by a CT imaging system;

identifying voxels in the 3D CT volume data set having a Hounsfield Unit (HU) value below a bone threshold, the voxels identified by one or more processors without using image continuity;

marking the identified voxels as non-hone voxels and the non-identified voxels as bone voxels;

determining, using the one or more processors, from the marked non-bone voxels, definite tissue voxels and non-definite tissue voxels, wherein determining includes identifying voxels surrounded by only marked non-bone voxels and connected, either directly or indirectly, to an initial voxel that is a marked non-bone voxel by other voxels that have been identified as definite tissue voxels, the identified voxels are marked as definite tissue voxels;

expanding a region defined by the definite tissue voxels;

masking the definite tissue voxels to remove the definite tissue voxels from the 3D CT volume data set; and

identifying bone marrow from the non-definite tissue as un-masked voxels that are not the marked bone voxels.

2. A method in accordance with claim 1 wherein the expanding comprises adding at least one layer of voxels to the determined definite tissue voxels.

3. A method in accordance with claim 1 further comprising displaying an image of a 3D CT volume generated from the 3D CT volume data set and highlighting the voxels corresponding to the identified bone marrow.

4. A method in accordance with claim 1 further comprising dividing the voxels identified as bone marrow into red bone marrow and yellow bone marrow using corresponding red bone marrow HU values and yellow bone marrow HU values.

5. A method in accordance with claim 1 further comprising selecting as an initial voxel for the determining definite tissue voxels a non-bone voxel.

6. A method in accordance with claim 5 wherein the initial voxel is a corner voxel selected based on being not surrounded by the marked bone voxels in the 3D CT volume data set.

7. A method in accordance with claim 1 further comprising acquiring a full body CT scan of a patient to obtain the 3D CT volume data set.

8. A method in accordance with claim 1 further comprising determining a total mass of bone marrow based on the voxels corresponding to the identified bone marrow voxels.

9. A method in accordance with claim 1 wherein the bone threshold is 200 Hounsfield Units (HU).

10. A method in accordance with claim 1 wherein the bone threshold is determined based on an HU value for hard bone.

11. A method in accordance with claim 1 further comprising counting a number of neighbor marked non-bone voxels for each marked non-hone voxel, wherein the number represents a count value for each marked non-bone voxel;

wherein the determining includes identifying definite voxels based on a count number of the non-bone voxel.

12. A method in accordance with claim 1 wherein the expanding comprises adding one or more layers of voxels adjacent to voxels that are marked non-bone voxels and are not determined definite tissue voxels.

13. A method in accordance with claim 1 wherein the definite tissue voxels comprise voxels identified as definitely not bone.