Method and apparatus for generating consistent image registration

• US 6,611,615 B1
• Filed: 06/25/1999
• Issued: 08/26/2003
• Est. Priority Date: 06/25/1999
• Status: Expired due to Fees

• Current Assignee: University of Iowa Research Foundation
• Original Assignee: University of Iowa Research Foundation
• Inventors: Christensen, Gary E.
• Primary Examiner(s): Patel, Jayanti K.
• Assistant Examiner(s): Choobin, Barry
• Application Number:
  US09/344,621
• Time in Patent Office: 1,523 Days
• Field of Search: 382/130, 382/128, 382/154, 382/294, 382/131, 382/276, 715/532, 600/407, 128/922, 424/9.4
• US Class Current: 382/130
• CPC Class Codes: G06T 5/20 using local operators
• CPC Class Codes: Y10S 128/922 including image analysis

Abstract

The present invention provides a method and apparatus for generating inversely related forward and reverse image transformations in any image registration technique with ambiguous correspondence based upon the joint estimation of the transformation. Accordingly, two image data sets are received, a source and a target. Current sets of forward and reverse displacement fields are initialized. From the current set of forward and reverse displacement fields, the consistent forward transformation and the consistent reverse transformation are jointly estimated. These transformations are outputted.

Claims

1. A method for generating consistent image registration between a first image and a second image, comprising the steps of:

(a) receiving a first image data set and a second image data set;

(b) initializing a current forward transformation and a current reverse transformation;

(c) computing an inverse of the current forward transformation and an inverse of the current reverse transformation;

(d) jointly estimating a consistent forward and a consistent reverse transformation between the first image data set and the second image data set by minimizing a difference between the current forward transformation and the inverse of the current reverse transformation and minimizing a difference between the current reverse transformation and the inverse of the current forward transformation; and

(e) outputting the jointly estimated forward and the reverse transformations.

2. The method of claim 1 wherein the step of jointly estimating a consistent forward and a consistent reverse transformation comprises the substeps of:

(i) computing a new consistent forward transformation; and

(ii) computing a new consistent reverse transformation.

3. The method of claim 2, wherein the step of jointly estimating a consistent forward and a consistent reverse transformation comprises the further substeps of:

(iii) setting the current forward transformation to the new consistent forward transformation;

(iv) setting the current reverse transformation to the new consistent reverse transformation; and

(v) repeating substeps (i)-(iv) until a predetermined condition is met.

4. The method of claim 3 wherein the predetermined condition is reaching a selected maximum number of repetitions.

5. The method of claim 3 wherein the predetermined condition is attaining a selected level of convergence.

6. The method of claim 3, wherein the step of initializing a current forward and a current reverse transformation further comprises initializing a current set of forward parameters and a current set of reverse parameters and wherein the substeps of computing the new consistent forward transformation and the new consistent reverse transformation are respectively performed based upon estimating a new set of forward parameters and a new set of reverse parameters for the transformation.

7. The method of claim 6, wherein the substep of setting the current forward transformation to the new consistent forward transformation further comprises setting the current set of forward parameters to the new set of forward parameters and wherein the substep of setting the current reverse transformation to the new consistent reverse transformation further comprises setting the current set of reverse parameters to the new set of reverse parameters.

8. The method of claim 7 wherein the step of jointly estimating a consistent forward and a consistent reverse transformation comprises the following substeps performed after substeps (i)-(iv) but prior to repetition in substep (v):

(vi) determining whether to increase a number of degrees of freedom;

(vii) increasing the number of degrees of freedom;

(viii) initializing a new set of forward transformation parameters and a new set of reverse transformation parameters based upon the increased number of degrees of freedom;

(ix) adding the new set of forward transformation parameters to the current set of forward transformation parameters; and

(x) adding the new set of reverse transformation parameters to the current set of reverse transformation parameters.

9. The method of claim 8 wherein the forward parameters are forward basis coefficients and the reverse parameters are reverse basis coefficients.

10. The method of claim 1 wherein the step of jointly estimating a consistent forward and a consistent reverse transformation estimates the forward and the reverse transformation utilizing a symmetric cost function and an inverse consistency constraint.

11. The method of claim 10 wherein the symmetric cost function is a mutual information cost function.

12. The method of claim 10 wherein the symmetric cost function is based upon a squared intensity difference cost function.

13. The method of claim 12 wherein the squared intensity difference cost function incorporates an intensity mapping function.

14. The method of claim 10 wherein the symmetric cost function is based upon an intensity variance cost function.

15. The method of claim 10 wherein the symmetric cost function is based upon matching features in the first image data set and the second image data set.

16. The method of claim 15 wherein the features conform to edges in the first image data set and the second image data set.

17. The method of claim 15 wherein the features conform to the gradient magnitude of the first image data set and the second image data set.

18. The method of claim 15, wherein the features conform to manifolds in the first image data set and the second image data set, wherein each image data set has a dimension, and wherein the manifold in the first image data set is of a dimension lesser than or equal to the dimension of the first image data set and the manifold in the second image data set is of a dimension lesser than or equal to the dimension of the second image data set.

19. The method of claim 18, wherein the manifolds are landmarks in the first image data set and the second image set.

20. The method of claim 18, wherein the manifolds are contours in the first image data set and the second image data set.

21. The method of claim 18, wherein the manifolds are subvolumes in the first image data set and the second image data set.

22. The method of claim 18, wherein the manifolds are surfaces in the first image data set and the second image data set.

23. The method of claim 10 wherein the step of jointly estimating a consistent forward and a consistent reverse transformation estimates the forward and the reverse transformation also utilizing a diffeomorphic constraint function.

24. The method of claim 23 wherein the diffeomorphic constraint function conforms to a regularization constraint.

25. The method of claim 23 wherein the diffeomorphic constraint function conforms to a linear differential operator constraint.

26. The method of claim 23 wherein the diffeomorphic constraint function conforms to a continuum mechanical model.

27. The method of claim 23 wherein the continuum mechanical model is a linear elasticity model.

28. The method of claim 10 wherein the step of jointly estimating a consistent forward and a consistent reverse transformation estimates the forward and the reverse transformation also utilizing a correspondence constraint function.

29. The method of claim 28, wherein the correspondence constraint function conforms to corresponding manifolds in the first image data set and the second image data set, wherein each image data set has a dimension, and wherein the manifold in the first image data set is of a dimension lesser than or equal to the dimension of the first image data set and the manifold in the second image data set is of a dimension lesser than or equal to the dimension of the second image data set.

30. The method of claim 29, wherein the manifolds are subvolumes in the first image data set and the second image data set.

31. The method of claim 29, wherein the manifolds are landmarks in the first image data set and the second image data set.

32. The method of claim 29, wherein the manifolds are contours in the first image data set and the second image data set.

33. The method of claim 29, wherein the manifolds are surfaces in the first image data set and the second image data set.

34. The method of claim 10 wherein the step of jointly estimating a consistent forward and a consistent reverse transformation estimates the forward and the reverse transformation also utilizing a diffeomorphic constraint function and a correspondence constraint function.

35. The method of claim 1, comprising the further step of:

(f) transforming the first image data set with the forward transformation and transforming the second image data set with the reverse transformation.

36. The method of claim 35, comprising the further step of:

(g) outputting the transformed first image data set and the transformed second image data set.

48. The storage device of claim 30 storing executable instructions which cause a processor to perform the further step of:

(f) transforming the first image data set with the forward transformation and transforming the second image data set with the reverse transformation.

49. The storage device of claim 48 storing executable instructions which cause a processor to perform the further step of:

(g) outputting the transformed first image data set and the transformed second image data set.

50. The storage device of claim 30, wherein the instructions causing a processor to jointly estimate a consistent forward and a consistent reverse transformation includes instructions to perform the substeps of:

(i) computing a new consistent forward transformation; and

(ii) computing a new consistent reverse transformation.

51. The storage device of claim 50, wherein the instructions causing a processor to jointly estimate a consistent forward and a consistent reverse transformation includes instructions to perform the further substeps of:

(iii) setting the current forward transformation to the new consistent forward transformation;

(iv) setting the current reverse transformation to the new consistent reverse transformation; and

(v) repeating substeps (i)-(iv) until a predetermined condition is met.

52. The storage device of claim 30, wherein the instructions causing a processor to jointly estimate a consistent forward and a consistent reverse transformation utilize a symmetric cost function and an inverse consistency constraint.

53. The storage device of claim 52, wherein the instructions causing a processor to jointly estimate a consistent forward and a consistent reverse transformation also utilize a correspondence constraint function.

54. The storage device of claim 53, wherein the correspondence constraint function conforms to corresponding manifolds in the first image data set and the second image data set, wherein each image data set has a dimension, and wherein the manifold in the first image data set is of a dimension lesser than or equal to the dimension of the first image data set and the manifold in the second image data set is of a dimension lesser than or equal to the dimension of the second image data set.

55. The storage device of claim 52, wherein the instructions causing a processor to jointly estimate a consistent forward and a consistent reverse transformation also utilize a diffeomorphic constraint function.

56. The storage device of claim 52, wherein the instructions causing a processor to jointly estimate a consistent forward and a consistent reverse transformation also utilize a diffeomorphic constraint function and a correspondence constraint function.

37. A system for generating consistent image registration between a first image data set and a second image data set comprising:

(a) a data storage system that stores one or more image data sets;

(b) one or more processor in communication with the data storage system for performing the steps comprising;

(i) receiving a first image data set and a second image data set;

(ii) initializing a current forward transformation and a current reverse transformation;

(iii) computing an inverse of the current forward transformation and an inverse of the current reverse transformation;

(iv) jointly estimating a consistent forward and a consistent reverse transformation between the first image data set and the second image data set by minimizing a difference between the current forward transformation and the inverse of the current reverse transformation and minimizing a difference between the current reverse transformation and the inverse of the current forward transformation; and

(v) outputting the jointly estimated forward and the reverse transformations.

38. The system of claim 37, wherein the processor performs the further step of:

(vi) transforming the first image data set with the forward transformation and transforming the second image data set with the reverse transformation.

39. The system of claim 38, wherein the processor performs the further step of:

(vii) outputting the transformed first image data set and the transformed second image data set.

40. The system of claim 37, wherein the processor jointly estimates a consistent forward and a consistent reverse transformation by performing the substeps of:

(A) computing a new consistent forward transformation; and

(B) computing a new consistent reverse transformation.

41. The system of claim 40, wherein the processor jointly estimates a consistent forward and a consistent reverse transformation by performing the further substeps of:

(C) setting the current forward transformation to the new consistent forward transformation;

(D) setting the current reverse transformation to the new consistent reverse transformation; and

(E) repeating substeps (A)-(D) until a predetermined condition is met.

42. The system of claim 37, wherein the processor jointly estimates a consistent forward and a consistent reverse transformation utilizing a symmetric cost function and an inverse consistency constraint.

43. The system of claim 42 wherein the processor jointly estimates a consistent forward and a consistent reverse transformation also utilizing a correspondence constraint function.

44. The system of claim 43, wherein the correspondence constraint function conforms to corresponding manifolds in the first image data set and the second image data set, wherein each image data set has a dimension, and wherein the manifold in the first image data set is of a dimension lesser than or equal to the dimension of the first image data set and the manifold in the second image data set is of a dimension lesser than or equal to the dimension of the second image data set.

45. The system of claim 42, wherein the processor jointly estimates a consistent forward and a consistent reverse transformation also utilizing a diffeomorphic constraint function and a correspondence constraint function.

46. The system of claim 42, wherein the processor jointly estimates a consistent forward and a consistent reverse transformation also utilizing a diffeomorphic constraint function.

47. A computer-readable, digital storage device storing executable instructions which cause a processor to generate a consistent image registration between a first image data set and a second image data set by performing steps comprising:

(a) receiving a first image data set and a second image data set;

(b) initializing a current forward transformation and a current reverse transformation;

(c) computing an inverse of the current forward transformation and an inverse of the current reverse transformation;

(d) jointly estimating a consistent forward and a consistent reverse transformation between the first image data set and the second image data set by minimizing a difference between the current forward transformation and the inverse of the current reverse transformation and minimizing a difference between the current reverse transformation and the inverse of the current forward transformation; and

(e) outputting the jointly estimated forward and the reverse transformations.