Method and System for Multi-Organ Segmentation Using Learning-Based Segmentation and Level Set Optimization

• US 20120230572A1
• Filed: 03/09/2012
• Published: 09/13/2012
• Est. Priority Date: 03/10/2011
• Status: Active Grant

• Current Assignee: Siemens Corp. (Siemens AG)
• Original Assignee: Siemens Corp. (Siemens AG), Siemens Molecular Imaging Incorporated (Siemens AG)
• Inventors: Sofka, Michal, Zhang, Jingdan, Zhou, Shaohua Kevin, Birkbeck, Neil, Declerck, Jerome, Kohlberger, Timo, Wetzl, Jens, Kaftan, Jens
• Granted Patent: US 9,042,620 B2
• Time in Patent Office: Days
• US Class Current: 382/131
• CPC Class Codes: G06T 2207/10081 Computed x-ray tomography [CT]
• CPC Class Codes: G06T 2207/20081 Training; Learning
• CPC Class Codes: G06T 2207/20161 Level set
• CPC Class Codes: G06T 2207/30048 Heart; Cardiac
• CPC Class Codes: G06T 2207/30084 Kidney; Renal
• CPC Class Codes: G06T 7/12 Edge-based segmentation
• CPC Class Codes: G06V 10/7553 based on shape, e.g. active shape models [ASM]

Abstract

A method and system for automatic multi-organ segmentation in a 3D image, such as a 3D computed tomography (CT) volume using learning-base segmentation and level set optimization is disclosed. A plurality of meshes are segmented in a 3D medical image, each mesh corresponding to one of a plurality of organs. A level set in initialized by converting each of the plurality of meshes to a respective signed distance map. The level set optimized by refining the signed distance map corresponding to each one of the plurality of organs to minimize an energy function.

Claims

1. A method for multi-organ segmentation in a 3D medical image comprising:

segmenting a plurality of meshes each corresponding to one of a plurality of organs in the 3D medical image;

initializing a level set by converting each of the plurality of meshes to a respective signed distance map; and

optimizing the level set by refining the respective signed distance map corresponding to each one of the plurality of organs to minimize a respective energy function.

2. The method of claim 1, further comprising:

generating a respective refined 3D mesh for each one of the plurality of organs from the respective refined distance map corresponding to each one of the plurality of organs.

3. The method of claim 1, wherein the respective energy function for each one of the plurality of organs includes a data energy term, a regularization energy term, and at least one energy term that imposes a geometric constraint on the respective signed distance function.

4. The method of claim 1, wherein the energy function includes a disjoint constraint energy term to remove overlaps between signed distance maps corresponding to adjacent organs.

5. The method of claim 4, wherein the disjoint energy term is expressed as:

E_O(φ

_A,φ

_B);

=∫

_Ω

H(φ

_A(x))H(φ

_B(x))φ

_B(x)dx,

where φ

_A and φ

_B are signed distance maps corresponding to adjacent organs, and H is a Heaviside step function.

6. The method of claim 1, wherein the energy function includes a local proximity constraint energy term to remove gaps between portions of signed distance maps corresponding to portions of adjacent organs that stay in fixed proximity.

7. The method of claim 6, wherein the local proximity constraint energy term is expressed as:

8. The method of claim 1, wherein the energy function includes a template constraint energy term enforce similarity of a refined distance map to the corresponding signed distance map initialized by converting one of the plurality of segmented meshes.

9. The method of claim 1, wherein the template constraint energy term is expressed as:

10. The method of claim 1, wherein the step of optimizing the level set by refining the respective signed distance map corresponding to each one of the plurality of organs to minimize a respective energy function comprises:

for each organ O_{i=1, . . . N}, calculating the energy minimization;

11. The method of claim 1, wherein the step of optimizing the level set by refining the respective signed distance map corresponding to each one of the plurality of organs to minimize a respective energy function comprises:

minimizing the respective energy function for each of the plurality of organs using a plurality of interleaved gradient descent iterations, wherein at each iteration, the respective signed distance map corresponding to each one of the plurality of organs is refined to reduce the respective energy function.

18. The apparatus of claim 1, wherein the means for optimizing the level set by refining the respective signed distance map corresponding to each one of the plurality of organs to minimize a respective energy function comprises:

means for minimizing the respective energy function for each of the plurality of organs using a plurality of interleaved gradient descent iterations.

12. An apparatus for multi-organ segmentation in a 3D medical image comprising:

means for segmenting a plurality of meshes each corresponding to one of a plurality of organs in the 3D medical image;

means for initializing a level set by converting each of the plurality of meshes to a respective signed distance map; and

means for optimizing the level set by refining the respective signed distance map corresponding to each one of the plurality of organs to minimize a respective energy function.

13. The apparatus of claim 12, further comprising:

means for generating a respective refined 3D mesh for each one of the plurality of organs from the respective refined distance map corresponding to each one of the plurality of organs.

14. The apparatus of claim 12, wherein the respective energy function for each one of the plurality of organs includes a data energy term, a regularization energy term, and at least one energy term that imposes a geometric constraint on the respective signed distance function.

15. The apparatus of claim 12, wherein the energy function includes a disjoint constraint energy term to remove overlaps between signed distance maps corresponding to adjacent organs.

16. The apparatus of claim 12, wherein the energy function includes a local proximity constraint energy term to remove gaps between portions of signed distance maps corresponding to portions of adjacent organs that stay in fixed proximity.

17. The apparatus of claim 12, wherein the energy function includes a template constraint energy term enforce similarity of a refined distance map to the corresponding signed distance map initialized by converting one of the plurality of segmented meshes.

19. A non-transitory computer readable medium encoded with computer executable instructions for multi-organ segmentation in a 3D medical image, the computer executable instructions defining a method comprising:

segmenting a plurality of meshes each corresponding to one of a plurality of organs in the 3D medical image;

initializing a level set by converting each of the plurality of meshes to a respective signed distance map; and

optimizing the level set by refining the respective signed distance map corresponding to each one of the plurality of organs to minimize a respective energy function.

20. The non-transitory computer readable medium of claim 19, further comprising:

generating a respective refined 3D mesh for each one of the plurality of organs from the respective refined distance map corresponding to each one of the plurality of organs.

21. The non-transitory computer readable medium of claim 19, wherein the respective energy function for each one of the plurality of organs includes a data energy term, a regularization energy term, and at least one energy term that imposes a geometric constraint on the respective signed distance function.

22. The non-transitory computer readable medium of claim 19, wherein the energy function includes a disjoint constraint energy term to remove overlaps between signed distance maps corresponding to adjacent organs.

23. The non-transitory computer readable medium of claim 22, wherein the disjoint energy term is expressed as:

E_O(φ

_A,φ

_B);

=∫

_Ω

H(φ

_A(x))H(φ

_B(x))φ

_B(x)dx,

where φ

_A and φ

_B are signed distance maps corresponding to adjacent organs, and H is a Heaviside step function.

24. The non-transitory computer readable medium of claim 19, wherein the energy function includes a local proximity constraint energy term to remove gaps between portions of signed distance maps corresponding to portions of adjacent organs that stay in fixed proximity.

25. The non-transitory computer readable medium of claim 24, wherein the local proximity constraint energy term is expressed as:

26. The non-transitory computer readable medium of claim 19, wherein the energy function includes a template constraint energy term enforce similarity of a refined distance map to the corresponding signed distance map initialized by converting one of the plurality of segmented meshes.

27. The non-transitory computer readable medium of claim 19, wherein the template constraint energy term is expressed as:

28. The non-transitory computer readable medium of claim 19, wherein the step of optimizing the level set by refining the respective signed distance map corresponding to each one of the plurality of organs to minimize a respective energy function comprises:

for each organ O_{i=1, . . . N}, calculating the energy minimization;

29. The non-transitory computer readable medium of claim 19, wherein the step of optimizing the level set by refining the respective signed distance map corresponding to each one of the plurality of organs to minimize a respective energy function comprises:

minimizing the respective energy function for each of the plurality of organs using a plurality of interleaved gradient descent iterations, wherein at each iteration, the respective signed distance map corresponding to each one of the plurality of organs is refined to reduce the respective energy function.