System and method for image segmentation in generating computer models of a joint to undergo arthroplasty

• US 20090274350A1
• Filed: 04/14/2009
• Published: 11/05/2009
• Est. Priority Date: 04/30/2008
• Status: Active Grant

• Current Assignee: Howmedica Osteonics Corp. (Stryker Corporation)
• Original Assignee: Otismed Corporation (Stryker Corporation)
• Inventors: Pavlovskaia, Elena, Sarva, Venkata S.
• Granted Patent: US 8,311,306 B2
• Time in Patent Office: Days
• US Class Current: 382/128
• CPC Class Codes: A61B 17/155 Cutting femur
• CPC Class Codes: A61B 17/157 Cutting tibia
• CPC Class Codes: A61B 2034/102 Modelling of surgical devices, implants or prosthesis
• CPC Class Codes: G06F 18/00 Pattern recognition
• CPC Class Codes: G06V 10/24 Aligning, centring, orientation detection or correction of the image
• CPC Class Codes: G06V 10/46 Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features colour feature extraction G06V10/56
• CPC Class Codes: G06V 10/471 using approximation functions
• CPC Class Codes: G06V 2201/033 of skeletal patterns

Abstract

Systems and methods for image segmentation in generating computer models of a joint to undergo arthroplasty are disclosed. Some embodiments may include a method of partitioning an image of a bone into a plurality of regions, where the method may include obtaining a plurality of volumetric image slices of the bone, generating a plurality of spline curves associated with the bone, verifying that at least one of the plurality of spline curves follow a surface of the bone, and creating a 3D mesh representation based upon the at least one of the plurality of spline curve.

Claims

1. A method of partitioning an image of a bone into a plurality of regions, the method comprising the acts of:

obtaining a plurality of volumetric image slices of the bone;

generating a plurality of spline curves associated with the bone;

verifying that at least one of the plurality of spline curves follow a surface of the bone; and

creating a three dimensional (3D) mesh representation based upon the at least one of the plurality of spline curves.

2. The method of claim 1, further comprising the act of employing selective resolution over different areas of the bone.

3. The method of claim 1, wherein a uniform spacing between the plurality of slices is 2 millimeters.

4. The method of claim 1, wherein the plurality with slices are in the sagittal, transversal or coronal directions.

5. The method of claim 1, wherein the volumetric image slices represent a portion of the bone that makes contact with a second bone.

6. The method of claim 1, wherein at least one image slice within the plurality includes voxels that are distorted through voxel volume averaging.

7. The method of claim 1, wherein the plurality of spline curves define the bone'"'"'s shape and orientation.

8. The method of claim 1, wherein the method yields a planning model of the bone.

9. The method of claim 1, wherein the act of verifying comprises the act of examining neighboring slices in the plurality.

10. The method of claim 1, further comprising the act of integrating a representation of a model bone into the image of the bone.

11. The method of claim 10, where the representation of the model bone has no damaged cancellous or cortical matter.

12. A method of generating a representation of a model bone, the method comprising the acts of:

obtaining an image scan of the representation as a plurality of slices;

segmenting each slice in the plurality using one or more segmentation curves;

generating a mesh of the representation;

adjusting each slice in the plurality to include areas where the boundary area of the bone is stable between successive image scans; and

generating anchor segmentation such that the anchor segmentation follows a boundary of the representation of the model bone.

13. The method of claim 12, wherein the act of segmenting excludes segmenting cartilage associated with the representation of the model bone.

14. The method of claim 12, wherein the act of obtaining the image scan occurs through the use of magnetic resonance imaging (MRI).

15. The method of claim 12, wherein at least a portion of the representation of the model bone exists between two segmentation curves

16. The method of claim 12, wherein Gaussian smoothing is employed.

17. The method of claim 12, wherein the boundary of the representation of the model bone comprises a boundary between cortical and cancellous bone matter.

18. The method of claim 12, further comprising the act of connecting a plurality of cortical and cancellous boundary area together to obtain the anchor segmentation.

19. The method of claim 18, wherein the plurality of cortical and cancellous boundaries are disjoint.

20. The method of claim 12, wherein at least one cancellous voxel is within the anchor segmentation and at least one cortical voxel is external to the anchor segmentation.

21. The method of claim 12, wherein osteophytes are excluded from the anchor segmentation.

22. The method of claim 12, wherein at least one cortical voxels lies along the boundary of the representation of the model bone and at least one cancellous voxel lies adjacent the boundary of the representation of the model bone.

23. The method of claim 12, further comprising the act of generating a 3D mesh representing the representation of the model bone.

24. The method of claim 12, further comprising the act of repositioning a point on the one or more segmentation curves by comparing point placement of adjacent segmentation curves.

25. A method of segmenting a target bone using a representation of a model bone, the method comprising the acts of:

registering a segmented form of the representation to an image scan of the target bone;

refining the registration of the segmented form of the representation near a boundary of the target bone;

generating a mesh from the segmented form of the representation; and

generating a plurality of spline curves that approximate the intersection of the mesh and one or more slices from the image scan of the target bone.

26. The method of claim 25, wherein the method is performed on a joint.

27. The method of claim 25, further comprising applying a transform within one or more slices.

28. The method of claim 27, wherein the act of applying the transforms includes moving points within a plane of the slices but not transversal to the slices.

29. The method of claim 25, wherein the act of refining comprises averaging one or more voxels of adjacent slices.

30. The method of claim 25, further comprising the act of verifying that at least one of the plurality of spline curves follow a surface of the target bone.

31. The method of claim 30, wherein the act of verifying comprises the act of comparing at least two adjacent slices.

32. A method of mapping a representation of a model bone into an image scan of a target bone:

registering a generated portion of the representation into the image scan of the target bone using a translational transformation;

registering the generated portion of the representation into the image scan of the target bone using a similarity transformation;

registering a boundary portion of the representation into the image scan of the target bone using an affine transformation; and

registering the boundary portion of the representation into the image scan of the target bone using a spline transformation.

33. The method of claim 32, further comprising the act of adjusting one or more parameters of the translational transformation such that the translational transformation minimizes misalignment of the representation of the model bone and the image scan of the target bone.

34. The method of claim 32, wherein the similarity transform includes seven factors that are varied to optimize match between the representation and the target bone.

35. The method of claim 32, wherein the spline transformation is a 3D B-spline transformation.

36. The method of claim 32, wherein the translational transform occurs without rotating, scaling, or deforming the representation.

37. The method of claim 32, wherein the representation is rotated, scaled, or deformed during the similarity transform.

38. The method of claim 32, wherein in the event the bone is a tibia, the method further comprises the act of using at least one additional registration operation between the affine and spline transformations.

39. The method of claim 38, wherein the at least one additional registration operation uses transforms represented with one or more quadratic functions.

40. The method of claim 32, wherein the model bone is varus.

41. The method of claim 32, wherein the model bone is valgus.

42. The method of claim 32, wherein the target bone includes a knee joint and the acts of registering are performed such that a tibia does not penetrate a femur.

43. The method of claim 32, further comprising the act of introducing a penalty into the mapping in the event that the mapping causes the tibia to penetrate the femur.

44. The method of claim 32, wherein a step gradient descent optimizer is used during the similarity transform, affine transform, and spline transforms.

45. The method of claim 32, wherein the translational transform includes using an exhaustive optimizer.

46. The method of claim 32, further comprising the act of correcting for variations between the representation of the model bone and the image scan of the target bone by supplying each voxel of the image scan with the distance to the nearest boundary voxel.

47. The method of claim 32, wherein the similarity transform utilizes, for a center position, a center of a bounding box that encloses the target bone registered during translation registration.

48. A method for determining a degree of correspondence between an image of a target bone and a representation of a model bone, the method comprising the acts of:

selecting a plurality of sample points in the representation of the model bone to be registered;

partitioning the plurality of sample points into a plurality of groups;

sampling the image of the target bone;

determining a correlation of voxel intensities between the image of the target bone and the representation of the model bone for each group in the plurality; and

averaging the correlation determined for each group in the plurality.

49. The method of claim 48, wherein the act of partitioning comprises subdividing a three dimensional image space into one or more cells using a rectilinear grid and wherein points belonging to the same cell are placed into the same group in the plurality.

50. The method of claim 48, wherein the sample points are partitioned into the one or more groups to compensate for non-uniformities in the image of the target bone.

51. The method of claim 50, wherein the non-uniformities vary between different imaging technologies.

52. The method of claim 48, wherein the act of determining the correlation is performed according to the following formula:

$\mathrm{NC}<br><br>\left(A,B\right)=\frac{\underset{i=1}{\overset{N}{\sum <br><br>}}<br><br><br><br>A_i<br><br>B_i}{\sqrt{\left(\underset{i=1}{\overset{N}{\sum <br><br>}}<br><br><br><br>A_i^2\right)<br><br>\left(\underset{i=1}{\overset{N}{\sum <br><br>}}<br><br><br><br>B_i^2\right)}}<br><br>\frac{N<br><br>\sum <br><br>A_i<br><br>B_i-\sum <br><br>A_i<br><br>\sum <br><br>B_i}{\sqrt{N<br><br>\sum <br><br>A_i^2-{\left(\sum <br><br>A_i\right)}^2}<br><br>\sqrt{N<br><br>\sum <br><br>B_i^2-{\left(\sum <br><br>B_i\right)}^2}}$

wherein A_i is the intensity in the i^th voxel of the image of the target bone A, B_i is the intensity in the corresponding i^th voxel of the image of the representation of the model bone B and N is the number of voxels considered.

53. The method of claim 48, wherein the act of averaging includes weighting correlations in relation to the number of sample points in the group.

54. A method for refining registration of a representation of a model bone to a target bone, the method comprising the acts of:

transforming an anchor segmentation mesh;

generating a plurality of random points around the transformed anchor segmentation mesh;

determining if each point in the plurality lies inside one or more of the following meshes;

InDark-OutLight, InLight-OutDark, or Dark-In-Light determining whether one or more of the plurality of points lie within a threshold distance of the surface of the transformed anchor segmentation mesh; and

adding each point in the plurality as a dark point or light point depending upon whether the point lies within the InDark-OutLight, InLight-OutDark, or Dark-In-Light meshes and to determine of the point lies inside or outside of the transformed anchor segmentation mesh.

55. The method of claim 54, wherein the plurality of points are within image slices.

56. The method of claim 54, wherein the act of determining if each point lies within a threshold distance comprises the acts of:

intersecting the anchor segmentation mesh with each image slice;

noting one or more pixels of the image slice that intersect with the anchor segmentation mesh; and

filtering the one or more pixels outside the threshold distance.

57. The method of claim 56, wherein the act of filtering occurs using a Dilate filter.

58. The method of claim 54, wherein the threshold distance is 0.75 millimeters.

59. The method of claim 54, wherein the act of transforming comprises the act of generating one or more spline curves.

60. A method for generating spline curves outlining the surface of a feature of interest of a target bone, the method comprising the acts of:

intersecting a 3D mesh model of the feature surface with one or more slices of target data, the intersection defining a polyline curve;

paramaterizing the polyline curve as a function of length and tangent variation;

calculating a weighted sum of the length and tangent paramaterizations; and

sampling the polyline using the results of the act of calculating.

61. The method of claim 60, further comprising positioning a plurality of new sample points using the results of the act of calculating.

62. The method of claim 61, wherein the new sample points are used as one or more control points for the polyline curve.

63. The method of claim 62, wherein the control points are for a Kochanek spline.

64. A method of generating a spline curve of a current slice of a bone, the method comprising the acts of:

obtaining a spline curve from a previous slice; and

modifying the spline curve from the previous slice to emphasize one or more features present in both the current slice and the previous slice.

65. The method of 64, wherein the act of modifying the spline curve from the previous slice occurs automatically.

66. The method of claim 65, wherein the act of modifying the spline curve from the previous slice comprises registering a region around the spline curve from the previous slice, the region defined by one or more threshold values.

67. The method of claim 66, wherein the one or more threshold values include about 0.7 millimeters outside the spline curve from the previous slice to about 5.0 millimeters within the spline curve from the previous slice.

68. The method of claim 66, wherein the region is initially registered using an affine transformation.

69. The method of claim 68, wherein a result from the affine transformation is used as a starting value for a B-spline deformable transformation.

70. The method of claim 66, wherein a greater number of points are taken for computing a match between the current and previous slices as the region comes closer to the spline curve from the previous slice.

71. The method of claim 69, further comprising the act of modifying one or more points from the spline from the previous slice by applying a final transformation.

72. The method of claim 65, wherein a technician modifies one or more control points to the spline curve from the previous slice in the event that the bone boundary changes due to the current and previous slices being tangent to the bone boundary.

73. The method of claim 65, wherein a technician modifies one or more control points to the spline curve from the previous slice in the event that the bone boundary changes due to osteophyte growth.