SYSTEM AND METHOD FOR CORONARY SEGMENTATION AND VISUALIZATION

• US 20080100621A1
• Filed: 10/11/2007
• Published: 05/01/2008
• Est. Priority Date: 10/25/2006
• Status: Active Grant

• Current Assignee: Siemens Healthcare GMBH (Siemens AG)
• Original Assignee: Siemens Corporate Research Incorporated (Siemens AG)
• Inventors: Aharon, Shmuel, Tek, Huseyin, Gulsun, Mehmet Akif
• Granted Patent: US 7,990,379 B2
• Time in Patent Office: Days
• US Class Current: 345/424
• CPC Class Codes: G06T 17/20 Finite element generation, e.g. wire-frame surface description, tesselation
• CPC Class Codes: G06T 2207/10072 Tomographic images
• CPC Class Codes: G06T 2207/20044 Skeletonization; Medial axis transform
• CPC Class Codes: G06T 2207/30101 Blood vessel; Artery; Vein; Vascular
• CPC Class Codes: G06T 7/12 Edge-based segmentation
• CPC Class Codes: G06T 7/162 involving graph-based methods
• CPC Class Codes: G06T 7/60 Analysis of geometric attributes

Abstract

A method of coronary vessel segmentation and visualization includes providing a digitized coronary image, placing a plurality of seed points along an estimated centerline of a coronary vessel, selecting a seed point and constructing a cyclic graph around the seed point in a plane perpendicular to the centerline at the seed point, performing a multi-scale-mean shift filtering in the perpendicular plane to estimate image gradient values, detecting a vessel boundary using a minimum-mean-cycle optimization that minimizes a ratio of a cost of a cycle to a length of a cycle, constructing a sub-voxel accurate vessel boundary about a point on the centerline, and refining the location of the centerline point from the sub-voxel accurate boundary, where the steps of constructing a sub-voxel accurate vessel boundary and refining the centerline point location are repeated until convergence.

Claims

1. A method of coronary vessel segmentation and visualization, comprising the steps of:

providing a digitized coronary image, said image comprising a plurality of intensities associated with a 3-dimensional grid of voxels;

placing a plurality of seed points along an estimated centerline of a coronary vessel;

selecting a seed point and constructing a cyclic graph around said seed point in a plane perpendicular to the centerline at the seed point;

performing a multi-scale-mean shift filtering in said perpendicular plane to estimate image gradient values;

detecting a boundary of said vessel using a minimum-mean-cycle optimization that minimizes a ratio of a cost of a cycle in said graph to a length of said cycle;

constructing a sub-voxel accurate vessel boundary about a point on said centerline; and

refining the location of said centerline point from said sub-voxel accurate boundary, wherein said steps of constructing a sub-voxel accurate vessel boundary and refining the centerline point location are repeated until convergence.

2. The method of claim 1, wherein said steps of selecting a seed point and constructing a cyclic graph, estimating image gradient values, and detecting a vessel boundary are repeated for each of the plurality of seed points, and further comprising constructing a 3-dimensional triangulated surface model from said sub-voxel accurate vessel boundaries to form a tubular mesh representing said coronary vessel, and rendering the tubular mesh from a viewer-specified direction.

3. The method of claim 1, wherein constructing a cyclic graph around said seed point comprises disposing an array of nodes are disposed at step at equally spaced angular intervals about each one of a circumference of plurality of concentric circles centered at the seed point;

and for each node in said graph, extending a plurality of edges to other nodes in said graph, wherein an angle of each edge termination node with respect to the seed point is greater than the angle of the edge origination node with respect to the seed point.

4. The method of claim 3, wherein the plurality of edges comprises 3 edges.

5. The method of claim 1, wherein performing a multi-scale-mean shift filtering comprises propagating one or more rays along the vessel, representing each ray as a collection of 2-dimensional points, performing a mean-shift filtering to each 2D points until convergence, selecting a plurality of edges based on a confidence function of said mean-shift filter size, and eliminating incorrect edges.

6. The method of claim 5, wherein said confidence function for a filter scale σ

_xk at a point i is given by;

7. The method of claim 5, wherein eliminating incorrect edges comprises eliminating those edges with a small confidence value.

8. The method of claim 2, wherein constructing a 3-dimensional triangulated surface model comprises, for each successive vessel boundary, sampling one boundary, locating corresponding points on a next boundary, and constructing a triangulated surface by connecting each point on the sampled boundary to two closest points on the next boundary.

9. The method of claim 2, wherein rendering the tubular mesh comprises:

specifying a less-than Z-buffer for said tubular mesh wherein every visible point in the resulting Z-buffer specifies the distance, along a viewing direction, from a viewer to a nearest visible part of the tubular mesh;

generating a set of two-dimensional points by extracting all Z-buffer points that are on a boundary between visible and non-visible parts of the tubular mesh;

generating a triangular mesh from said set of two-dimensional points;

projecting the triangular mesh into a 3-dimensional space by adding to each point of said mesh a Z coordinate calculated from a corresponding less-than Z-buffer function value and a known model-view transformation of the specific viewing direction;

creating a complete Z-buffer representation of the projected 3-dimensional triangular mesh representing said vessels by adding near-plane Z-buffer values of the projected 3-dimensional triangular mesh to corresponding non-visible points of the less-than Z-buffer function; and

volume rendering said 3-dimensional triangular mesh by calculating ray starting points and direction from the complete Z-buffer values and the specified viewing direction.

10. The method of claim 9, wherein the rays are propagated to a specified depth in the 3-dimensional triangular mesh.

11. The method of claim 9, further comprising generating a far Z-plane to provide an exit point for each ray.

12. The method of claim 9, wherein extracting all Z-buffer points that are on a boundary between visible and non-visible parts of the tubular mesh comprises checking, for every visible point in the Z-buffer, whether or not all neighboring points are visible, skipping to a next pixel if all the neighbors are visible, otherwise if there is at least one non-visible neighbor point, mark this point as a boundary between visible and non-visible points;

and adding additional points at corners of the Z-buffer according to a desired bounding box of an image area to be visualized.

13. The method of claim 9, wherein said volume rendering is performed using calculated mean-shift values of each point in said mesh, wherein an intensity value at each point is replaced with a local intensity of a mode obtained from the mean shift computation, wherein a mean-shift value of a triangle in said mesh is determined from an average value of the mean-shift values at the corners computed from a 1D mean-shift computation at each location.

14. A method of coronary vessel segmentation and visualization, comprising the steps of:

providing a digitized coronary image, said image comprising a plurality of intensities associated with a 3-dimensional grid of voxels;

constructing a plurality of vessel boundaries about seed points along an estimated centerline of a coronary vessel;

constructing a sub-voxel accurate vessel boundaries from said vessel boundaries;

refining the location of said centerline point from said sub-voxel accurate boundary, wherein said steps of constructing a sub-voxel accurate vessel boundary and refining the centerline point location are repeated until convergence;

constructing a 3-dimensional triangulated surface model from said sub-voxel accurate vessel boundaries to form a tubular mesh representing said coronary vessel;

specifying a less-than Z-buffer for said tubular mesh wherein every visible point in the resulting Z-buffer specifies the distance, along a viewing direction, from a viewer to a nearest visible part of the tubular mesh;

generating a set of two-dimensional points by extracting all Z-buffer points that are on a boundary between visible and non-visible parts of the tubular mesh;

generating a triangular mesh from said set of two-dimensional points;

projecting the triangular mesh into a 3-dimensional space by adding to each point of said mesh a Z coordinate calculated from a corresponding less-than Z-buffer function value and a known model-view transformation of the specific viewing direction; and

creating a complete Z-buffer representation of the projected 3-dimensional triangular mesh representing said vessels by adding near-plane Z-buffer values of the projected 3-dimensional triangular mesh to corresponding non-visible points of the less-than Z-buffer function.

15. The method of claim 14, further comprising volume rendering said 3-dimensional triangular mesh by calculating ray starting points and direction from the complete Z-buffer values and the specified viewing direction.

16. The method of claim 14, wherein constructing a plurality of vessel boundaries comprises:

placing a plurality of seed points along said estimated centerline of said coronary vessel;

selecting a seed point and constructing a cyclic graph around said seed point in a plane perpendicular to the centerline at the seed point;

performing a multi-scale-mean shift filtering in said perpendicular plane to estimate image gradient values; and

detecting a boundary of said vessel using a minimum-mean-cycle optimization that minimizes a ratio of a cost of a cycle in said graph to a length of said cycle, wherein said steps of selecting a seed point and constructing a cyclic graph, estimating image gradient values, and detecting a vessel boundary are repeated for each of the plurality of seed points.

17. The method of claim 14, further comprising performing a Ribbon Multi-Planar-Reconstruction (MPR) rendering, wherein for each point along the centerline, interpolating values of points along a line parallel to a viewing direction in a width of the vessel'"'"'s boundary, using said points as a boundary between visible and non-visible regions of said vessel, and filling a surrounding area with interpolated data from surrounding points.

18. The method of claim 17, wherein said ribbon MPR is calculated using image intensity values.

19. The method of claim 17, wherein said ribbon MPR is calculated using mean-shift values of each point in said mesh, wherein an intensity value at each point is replaced with a local intensity of a mode obtained from the mean shift computation, wherein a mean-shift value of a triangle in said mesh is determined from an average value of the mean-shift values at the corners computed from a 1D mean-shift computation at each location.

20. A program storage device readable by a computer, tangibly embodying a program of instructions executable by the computer to perform the method steps for coronary vessel segmentation and visualization, said method comprising the steps of:

providing a digitized coronary image, said image comprising a plurality of intensities associated with a 3-dimensional grid of voxels;

placing a plurality of seed points along an estimated centerline of a coronary vessel;

selecting a seed point and constructing a cyclic graph around said seed point in a plane perpendicular to the centerline at the seed point;

performing a multi-scale-mean shift filtering in said perpendicular plane to estimate image gradient values;

detecting a boundary of said vessel using a minimum-mean-cycle optimization that minimizes a ratio of a cost of a cycle in said graph to a length of said cycle;

constructing a sub-voxel accurate vessel boundary about a point on said centerline; and

refining the location of said centerline point from said sub-voxel accurate boundary, wherein said steps of constructing a sub-voxel accurate vessel boundary and refining the centerline point location are repeated until convergence.

21. The computer readable program storage device of claim 20, wherein said steps of selecting a seed point and constructing a cyclic graph, estimating image gradient values, and detecting a vessel boundary are repeated for each of the plurality of seed points, and further comprising constructing a 3-dimensional triangulated surface model from said sub-voxel accurate vessel boundaries to form a tubular mesh representing said coronary vessel, and rendering the tubular mesh from a viewer-specified direction.

22. The computer readable program storage device of claim 20, wherein constructing a cyclic graph around said seed point comprises disposing an array of nodes are disposed at step at equally spaced angular intervals about each one of a circumference of plurality of concentric circles centered at the seed point;

and for each node in said graph, extending a plurality of edges to other nodes in said graph, wherein an angle of each edge termination node with respect to the seed point is greater than the angle of the edge origination node with respect to the seed point.

23. The computer readable program storage device of claim 22, wherein the plurality of edges comprises 3 edges.

24. The computer readable program storage device of claim 20, wherein performing a multi-scale-mean shift filtering comprises propagating one or more rays along the vessel, representing each ray as a collection of 2-dimensional points, performing a mean-shift filtering to each 2D points until convergence, selecting a plurality of edges based on a confidence function of said mean-shift filter size, and eliminating incorrect edges.

25. The computer readable program storage device of claim 24, wherein said confidence function for a filter scale σ

_xk at a point i is given by;

26. The computer readable program storage device of claim 24, wherein eliminating incorrect edges comprises eliminating those edges with a small confidence value.

27. The computer readable program storage device of claim 21, wherein constructing a 3-dimensional triangulated surface model comprises, for each successive vessel boundary, sampling one boundary, locating corresponding points on a next boundary, and constructing a triangulated surface by connecting each point on the sampled boundary to two closest points on the next boundary.

28. The computer readable program storage device of claim 21, wherein rendering the tubular mesh comprises:

specifying a less-than Z-buffer for said tubular mesh wherein every visible point in the resulting Z-buffer specifies the distance, along a viewing direction, from a viewer to a nearest visible part of the tubular mesh;

generating a set of two-dimensional points by extracting all Z-buffer points that are on a boundary between visible and non-visible parts of the tubular mesh;

generating a triangular mesh from said set of two-dimensional points;

projecting the triangular mesh into a 3-dimensional space by adding to each point of said mesh a Z coordinate calculated from a corresponding less-than Z-buffer function value and a known model-view transformation of the specific viewing direction;

creating a complete Z-buffer representation of the projected 3-dimensional triangular mesh representing said vessels by adding near-plane Z-buffer values of the projected 3-dimensional triangular mesh to corresponding non-visible points of the less-than Z-buffer function; and

volume rendering said 3-dimensional triangular mesh by calculating ray starting points and direction from the complete Z-buffer values and the specified viewing direction.

29. The computer readable program storage device of claim 28, wherein the rays are propagated to a specified depth in the 3-dimensional triangular mesh.

30. The computer readable program storage device of claim 28, the method further comprising generating a far Z-plane to provide an exit point for each ray.

31. The computer readable program storage device of claim 28, wherein extracting all Z-buffer points that are on a boundary between visible and non-visible parts of the tubular mesh comprises checking, for every visible point in the Z-buffer, whether or not all neighboring points are visible, skipping to a next pixel if all the neighbors are visible, otherwise if there is at least one non-visible neighbor point, mark this point as a boundary between visible and non-visible points;

and adding additional points at corners of the Z-buffer according to a desired bounding box of an image area to be visualized.

32. The computer readable program storage device of claim 28, wherein said volume rendering is performed using calculated mean-shift values of each point in said mesh, wherein an intensity value at each point is replaced with a local intensity of a mode obtained from the mean shift computation, wherein a mean-shift value of a triangle in said mesh is determined from an average value of the mean-shift values at the corners computed from a 1D mean-shift computation at each location.