Method and system for processing an image of body tissues
View Patent ↗A method and system for detecting a boundary between a first tissue and a second tissue in a two-dimensional image. A cost is assigned to each pixel in a detection region indicative of the likelihood that the pixel is not located on the boundary between the first tissue and the second tissue. The boundary between the first and second tissues or organs is a path of voxels of minimal cost in the detection region from among two or more paths, the cost of a path being obtained in a calculation based upon the costs assigned to the voxels of the path. The method may be used to detect a boundary between a first tissue or organ and a second tissue or organ in a three-dimensional scan by applying the method to each of one or more two-dimensional sections and joining the two dimensional boundaries into a three dimensional boundary in the three dimensional scan. The method may also be used in segmenting a three-dimensional body scan.
1. A method of segmenting tissue of interest in a CT scan of the area of the tissue of interest, comprising:
processing a first section of the CT scan data using a computer to define a contour representing a 2D boundary between the tissue of interest and adjacent tissue;
processing a succession of further sections using the computer to define a contour representing a 2D boundary of the tissue of interest in the section being processed, each 2D boundary being within a selected distance from the 2D boundary in a previously processed section; and
generating a 3D boundary enclosing the tissue of interest in relation to the adjacent tissue from the succession of 2D boundaries defined in the succession of sections,
wherein the processing to define the 2D boundaries in the first section and the boundaries in the successive sections is performed by the computer using a cost function that defines the likelihood that a voxel in a section is not located on the boundary between the heart and the first adjacent tissue, the boundary in the section being processed being determined as a path between a pair of start and end points having the lowest cost,
wherein a cost assigned to a given voxel is calculated by the computer as a difference of a bad-grade and a good-grade, the good-grade being indicative of the extent that the given voxel's arrangement of neighbors has attributes of a voxel at the boundary between the tissue of interest and adjacent tissues and the bad-grade being indicative of the extent that the given voxel's arrangement of neighbors has attributes different from attributes of a voxel at the boundary between adjacent tissues, and
wherein the good-grade of given voxel is based upon any one or more of the following, for a predetermined n:
(a) the number of adjacent tissue voxels neighboring the given voxel within an n-voxels neighborhood, divided by n; and
(b) the number of fat voxels within an n-voxels neighborhood, from above times the total number of tissue of interest voxels or adjacent tissue voxels and contrast voxels within an n-voxels neighborhood, from below, divided by n (n+1).
2. The method according to claim 1 , wherein a cost is assigned to a given voxel in a section by:
(a) classifying one or more voxels in a neighborhood of the given voxel into a tissue class from among one or more tissue classes; and
(b) assigning the cost to the given voxel in a computer calculation involving the tissue classification of voxels in the neighborhood.
3. The method according to claim 2 wherein the given voxel has a gray level and a voxel is classified into a tissue class by:
(a) providing a gray level range for each of the one or more tissue classes; and
(b) classifying the voxel into a tissue class whose gray level range includes the gray level of the voxel.
4. The method according to claim 1 wherein the CT scan is a cardiac CT scan.
5. The method according to claim 4 wherein the tissue of interest is cardiac tissue.
6. The method according to claim 5 wherein the adjacent tissue is one or more of lung tissue, fat tissue, or liver tissue, sternum tissue, and contrast-material.
7. The method according to claim 6 wherein a cost assigned to a given voxel is calculated by the computer based on a difference between a bad-grade and a good-grade, the good-grade being indicative of the extent that the given voxel's arrangement of neighbors has attributes of a voxel at the boundary between the tissue of interest and the adjacent tissue and the bad-grade being indicative of the extent that the given voxel's arrangement of neighbors has attributes different from attributes of a voxel at the boundary between the tissue of interest and the adjacent tissue.
8. The method according to claim 5 wherein the adjacent tissue is bone tissue.
9. The method according to claim 8 wherein the bone tissue is sternum.
10. The method according to claim 1 , wherein the processing using a cost function is performed by a fast marching algorithm.
11. A non-transitory computer program storage device readable by a computer, tangibly embodying a computer program of instructions executable by the computer to perform the method of claim 10 .
12. The method according to claim 1 , wherein the processing of successive adjacent sections continues until a defined criterion is satisfied.
13. The method according to claim 1 wherein the sets of start and end points in the first section are defined arbitrarily, and the sets of start and end points in the successive sections are determined by:
generating a domain using the computer consisting of all points in a previously processed adjacent section that are a given distance from the boundary determined in that section;
selecting start and end points in a succeeding section that are at the given distance from the boundary of the previous section; and
using the cost function to determine the boundary in the succeeding section, with the determination being restricted to points in the generated domain.
14. The method according to claim 13 , wherein determination of the boundaries using the cost function in the succession of sections is made using a fast marching algorithm.
15. A non-transitory computer program storage device readable by a computer, tangibly embodying a computer program of instructions executable by the computer to perform the method of claim 13 .
16. A method of segmenting tissue of interest in a CT scan comprising:
repeating the processing as defined in claim 1 to develop a set of 2D boundaries between the tissue of interest and a second adjacent tissue; and
generating a composite 3D boundary enclosing the tissue of interest in relation to the first and second adjacent tissues from the succession of 2D boundaries between the tissue of interest and the adjacent tissue with the succession of 2D boundaries between the tissue of interest and the second adjacent tissue using the computer to form a composite 3D boundary.
17. The method according to claim 16 wherein the composite 3D boundary is incorporated into the CT scan.
18. A non-transitory computer program storage device readable by a computer, tangibly embodying a computer program of instructions executable by the computer to perform the method of claim 17 .
19. A non-transitory computer program storage device readable by a computer, tangibly embodying a computer program of instructions executable by the computer to perform the method of claim 16 .
20. The method according to claim 1 wherein the cost assigned to a path is a sum of the costs of the voxels of the path.
21. A non-transitory computer program storage device readable by a computer, tangibly embodying a computer program of instructions executable by the computer to perform the method of claim 20 .
22. A non-transitory computer program storage device readable by a computer, tangibly embodying a computer program of instructions executable by the computer to perform the method of claim 1 .
23. The method according to claim 1 wherein the succession of selected distances are between 2D boundaries in a previously processed adjacent section.
24. The method according to claim 1 wherein each of the succession of further sections are parallel to each other.
25. A method of segmenting tissue of interest in a CT scan of the area of the tissue of interest, comprising:
processing a first section of the CT scan data using a computer to define a contour representing a 2D boundary between the tissue of interest and adjacent tissue;
processing a succession of further sections using the computer to define a contour representing a 2D boundary of the tissue of interest in the section being processed, each 2D boundary being within a selected distance from the 2D boundary in a previously processed section; and
generating a 3D boundary enclosing the tissue of interest in relation to the adjacent tissue from the succession of 2D boundaries defined in the succession of sections;
wherein the processing to define the 2D boundaries in the first section and the boundaries in the successive sections is performed by the computer using a cost function that defines the likelihood that a voxel in a section is not located on the boundary between the heart and the first adjacent tissue, the boundary in the section being processed being determined as a path between a pair of start and end points having the lowest cost;
wherein a cost assigned to a given voxel is calculated by the computer as a difference of a bad-grade and a good-grade, the good-grade being indicative of the extent that the given voxel's arrangement of neighbors has attributes of a voxel at the boundary between the tissue of interest and adjacent tissues and the bad-grade being indicative of the extent that the given voxel's arrangement of neighbors has attributes different from attributes of a voxel at the boundary between adjacent tissues; and
wherein the bad-grade is based upon any one or more of the following, for a predetermined n:
(a) the total number of adjacent tissue voxels and contrast voxels within an n-voxels neighborhood, from above, times the number of voxels of a further adjacent tissue within an n-voxels neighborhood, from below, divided by n(n+1);
(b) the number of contrast voxels within an n-voxels neighborhood, from above times the number of contrast voxels within an n-voxels neighborhood from below, divided by n (n+1);
(c) the number of tissue of interest voxels or adjacent tissue voxels within an n-voxels neighborhood from above times the total number of heart or liver voxels, within an n-voxels neighborhood and missing voxels, from below, divided by n(n+1); and
(d) a quantity having the value 1 if there are contrast values having a gray value above 200, and 0 otherwise;
wherein a cost is assigned to a given voxel in a section by:
(a) classifying one or more voxels in a neighborhood of the given voxel into a tissue class from among one or more tissue classes; and
(b) assigning the cost to the given voxel in a computer calculation involving the tissue classification of voxels in the neighborhood.
26. A non-transitory computer program storage device readable by a computer, tangibly embodying a computer program of instructions executable by the computer to perform the method of claim 25 .
27. A method of segmenting tissue of interest in a CT scan of the area of the tissue of interest comprising:
processing a first section of the CT scan data using a computer to define a contour representing a 2D boundary between the tissue of interest and adjacent tissue;
processing a succession of further sections using the computer to define a contour representing a 2D boundary of the tissue of interest in the section being processed, each 2D boundary being within a selected distance from the 2D boundary in a previously processed section; and
generating a 3D boundary enclosing the tissue of interest in relation to the adjacent tissue from the succession of 2D boundaries defined in the succession of sections;
wherein the processing to define the 2D boundaries in the first section and the boundaries in the successive sections is performed by the computer using a cost function that defines the likelihood that a voxel in a section is not located on the boundary between the heart and the first adjacent tissue, the boundary in the section being processed being determined as a path between a pair of start and end points having the lowest cost;
wherein a cost is assigned to a given voxel in a section by:
(a) classifying one or more voxels in a neighborhood of the given voxel into a tissue class from among one or more tissue classes; and
(b) assigning the cost to the given voxel in a computer calculation involving the tissue classification of voxels in the neighborhood;
wherein a cost assigned to a given voxel is calculated as a difference of a bad-grade and a good-grade, the good-grade being indicative of the extent that the given voxel's arrangement of neighbors has attributes of a voxel at the boundary between cardiac and a selected adjacent tissue and the bad-grade being indicative of the extent that the given voxel's arrangement of neighbors has attributes different from attributes of a voxel at the boundary between cardiac and the selected adjacent tissue;
wherein the tissue of interest is cardiac tissue, and the one or more tissue classes are selected from lung tissue, fat tissue, heart tissue and contrast or bone tissue;
wherein a cost assigned to a given voxel is calculated as a difference of a bad-grade and a good-grade, the good-grade being indicative of the extent that the given voxel's arrangement of neighbors has attributes of a voxel at the boundary between cardiac and a selected adjacent tissue, and the bad-grade being indicative of the extent that the given voxel's arrangement of neighbors has attributes different from attributes of a voxel at the boundary between cardiac and the selected adjacent tissue; and
wherein the good-grade is based upon any one or more of the following, for a predetermined n:
(a) the number of inside fat-value neighbors within an n-voxels neighborhood, times the sum of the number of outside heart-muscle-value and the number of contrast-bone value voxels within an n-voxels neighborhood, divided by n(n+1); and
(b) the sum of inside fat voxels and heart-muscle voxels within an n-voxels neighborhood, times the number of outside lung voxels within an n-voxels neighborhood, divided by n(n+1).