Method and system for motion compensation in a temporal sequence of images

• US 20050265611A1
• Filed: 05/20/2005
• Published: 12/01/2005
• Est. Priority Date: 05/25/2004
• Status: Active Grant

• Current Assignee: Siemens Healthcare GMBH (Siemens AG)
• Original Assignee: Siemens Medical Solutions USA Incorporated (Siemens AG)
• Inventors: Valadez, Gerardo Hermosillo
• Granted Patent: US 7,616,836 B2
• Time in Patent Office: Days
• US Class Current: 382/236
• CPC Class Codes: G06T 2207/10076 4D tomography; Time-sequential 3D tomography
• CPC Class Codes: G06T 2207/10088 Magnetic resonance imaging [MRI]
• CPC Class Codes: G06T 2207/20016 Hierarchical, coarse-to-fine, multiscale or multiresolution image processing; Pyramid transform
• CPC Class Codes: G06T 2207/30068 Mammography; Breast
• CPC Class Codes: G06T 2207/30096 Tumor; Lesion
• CPC Class Codes: G06T 7/0016 involving temporal comparison
• CPC Class Codes: G06T 7/269 using gradient-based methods
• CPC Class Codes: G06T 7/32 using correlation-based methods
• CPC Class Codes: G06T 7/37 using transform domain methods

Abstract

A method and system for performing motion compensation in a temporal sequence of images include performing a conjugate gradient maximization of a similarity measure between two images, based on the local cross-correlation of corresponding regions to obtain a displacement field for warping one of the images. The non-singularity of the deformation is ensured by utilizing a composition of regularized gradients of the similarity measure when building the solution.

Claims

1. A method for motion compensation in first and second images in a temporal sequence by computing a deformation function, said method comprising:

performing a conjugate gradient maximization of a similarity measure between said first and second images; and

deriving said deformation function by utilizing said gradient maximization in conjunction with composition criteria so as to ensure non-singularity of said deformation function.

2. A method in accordance with claim 1, wherein said conjugate gradient maximization is based on local cross-correlation of corresponding regions around each point in said images.

3. A method in accordance with claim 1, wherein said composition criteria comprise deriving said deformation function by composition of a plurality of regularized gradients of said similarity measure such that said deformation exhibits essentially no singularities.

4. A method in accordance with claim 1, wherein said step of performing a conjugate gradient maximization comprises:

deriving respective deformations by composition of a plurality of regularized gradients of said similarity measure for obtaining said deformation function.

5. A method in accordance with claim 1, wherein said step of deriving said deformation function comprises applying the algorithm:

$\{\begin{array}{c}\mathrm{Let}<br><br>\text{\hspace{1em}<br><br>}<br><br>U_0=\mathrm{id}\\ \begin{array}{c}\mathrm{For}<br><br>\mathrm{ <br><br> <br><br> <br><br>}<br><br>k=1<br><br>\text{\hspace{1em}<br><br>}<br><br>\mathrm{\cdots <br><br>}<br><br>\text{\hspace{1em}<br><br>}<br><br>n<br><br>\text{;<br><br>}\\ <br><br>v_k=G_{\mathrm{\sigma <br><br>}}*\nabla <br><br>S<br><br>\left(I_1,I_2<br><br>\text{\hspace{1em}<br><br>}\circ <br><br>\text{\hspace{1em}<br><br>}<br><br>{\mathrm{\varphi <br><br>}}_k\right)\\ <br><br>{\mathrm{\varphi <br><br>}}_{k+1}={\mathrm{\varphi <br><br>}}_k<br><br>\text{\hspace{1em}<br><br>}\circ <br><br>\text{\hspace{1em}<br><br>}<br><br>\left(\mathrm{id}+{\mathrm{\varepsilon <br><br>}}_k<br><br>v_k\right)\end{array}\\ \mathrm{End}\end{array}\hspace{1em}<br><br>$

wherein operator G_σ

* denotes convolution by a Gaussian kernel,

S denotes the similarity measure,

∇

S denotes the gradient of S with respect to φ

,

I₁;

→

R denotes said reference image,

I₂;

→

R denotes said floating image,

φ

;

Ω

→

R³ denotes said deformation such that said similarity measure is maximized,

∘

denotes composition of functions, and

ε

_k is the step size, which is made sufficiently small to ensure invertibility.

6. A method in accordance with claim 5, wherein said step of deriving a deformation comprises deriving a deformation by composition of displacements which are small as compared with the size of said images.

7. A method in accordance with claim 5, wherein said step of deriving a deformation comprises deriving a deformation by composition of respective regularized gradients of said similarity measure.

8. A method in accordance with claim 5, wherein said step of warping comprises computing said deformation composed with said floating image.

9. A method in accordance with claim 5, wherein said step of warping comprises computing (I₂∘

φ

(x), wherein I₂;

→

R denotes said floating image, φ

;

Ω

→

R³ denotes said deformation such that said similarity measure is maximized, and ∘

denotes composition of functions.

10. A method in accordance with claim 9, wherein said step of computing comprises utilizing tri-linear interpolation at each voxel of said floating image being warped.

11. A method in accordance with claim 1, wherein said step of acquiring a reference image and a floating image comprises:

utilizing a medical imaging technique, including any of magnetic resonance imaging (MRI), X-ray imaging, and CT scan imaging.

12. A method in accordance with claim 1, wherein said step of obtaining a reference image and a floating image comprises:

utilizing a medical imaging technique for detection of at least one of tumors and potential tumors in breast cancer.

13. A method in accordance with claim 1, wherein said step of obtaining a reference image and a floating image comprises:

utilizing a medical imaging technique for obtaining images indicating suspect tissue exhibiting at least one of (a) a more rapid intake or wash-in of contrast agent and (b) a more rapid washout than adjacent, non-tumor tissue.

14. A method in accordance with claim 1, comprising a step of utilizing said motion corrected image to detect suspect tissue through comparison of images of a patient made before and after at least one of such wash-in and washout.

15. A method in accordance with claim 1, wherein said conjugate gradient maximization is based on local cross-correlation of corresponding regions around each point in said images.

16. A method in accordance with claim 1, wherein said composition criteria comprise deriving said deformation function by composition of a plurality of regularized gradients of said similarity measure such that said deformation exhibits essentially no singularities.

17. A method in accordance with claim 1, wherein said step of performing a conjugate gradient maximization comprises:

deriving respective deformations by composition of a plurality of regularized gradients of said similarity measure for obtaining said deformation function.

18. A method for performing image motion compensation between a reference image and an initial floating image by computing a deformation function from an initial deformation function, said method comprising:

(a) setting said initial deformation function as a current deformation function;

(b) warping said initial floating image with said current deformation function so as to obtain a current warped floating image;

(c) computing a current similarity measure between said reference image and said current warped floating image;

(d) computing the current gradient of said current similarity measure with respect to said current deformation function;

(e) regularizing said current gradient so as to ensure its invertibility to obtain a current regularized gradient;

(f) (A) in the first performance of this step, setting said current regularized gradient as current conjugate gradient, and

(B) in subsequent iterations of this step, conjugating said current regularized gradient with said current conjugate gradient so as to obtain a subsequent conjugate gradient;

(g) composing said subsequent conjugate gradient with said current deformation function so as to obtain a subsequent deformation function;

(h) setting said current deformation function as said subsequent deformation function;

(i) setting said current conjugate gradient as said subsequent conjugate gradient;

(j) going to step (b) and repeating until reaching a predetermined stop criterion; and

(k) defining the current warped floating image as a final warped floating image.

19. A method in accordance with claim 18, wherein said iterations begin at a low resolution level and proceed through increasingly higher resolutions.

20. A method in accordance with claim 18, wherein said stop criterion comprises reaching at least one of:

a predetermined maximum number of repetitions or iterations of the recited sequence of steps;

a predetermined lower limit for said gradient;

zero for said gradient; and

a predetermined processing time limit.

21. A method for performing image motion compensation between a reference image and an initial floating image by computing a deformation function from an initial deformation function, said method comprising:

(a) acquiring reference and initial floating images in a time sequence;

(b) setting said initial deformation function as a current deformation function;

(c) warping said initial floating image with said current deformation function so as to obtain a current warped floating image;

(d) computing a current similarity measure between said reference image and said current warped floating image;

(e) computing the current gradient of said current similarity measure with respect to said current deformation function;

(f) regularizing said current gradient so as to ensure its invertibility to obtain a current regularized gradient;

(g) setting said current regularized gradient as current conjugate gradient;

(h) composing said subsequent conjugate gradient with said current deformation function so as to obtain a subsequent deformation function;

(i) setting said current deformation function as said subsequent deformation function;

(j) setting said current conjugate gradient as said subsequent conjugate gradient;

(k) warping said initial floating image with said current deformation function so as to obtain a current warped floating image;

(l) computing a current similarity measure between said reference image and said current warped floating image;

(m) computing the current gradient of said current similarity measure with respect to said current deformation function;

(n) regularizing said current gradient so as to ensure its invertibility to obtain a current regularized gradient;

(o) conjugating said current regularized gradient with said current conjugate gradient so as to obtain a subsequent conjugate gradient;

(p) composing said subsequent conjugate gradient with said current deformation function so as to obtain a subsequent deformation function;

(q) setting said current deformation function as said subsequent deformation function;

(r) setting said current conjugate gradient as said subsequent conjugate gradient;

(s) going to step (k) and repeating until a predetermined stop criterion is reached; and

(t) defining the current warped floating image as a final warped floating image.

22. A method in accordance with claim 21, wherein said iterations begin at a low resolution level and proceed through increasingly higher resolutions.

23. A method in accordance with claim 21, wherein said predetermined stop criterion comprises reaching at least one of:

a predetermined maximum number of repetitions or iterations of the recited sequence of steps;

a predetermined lower limit for said gradient;

zero for said gradient; and

a predetermined processing time limit.

24. A method in accordance with claim 21, wherein said step of acquiring reference and initial floating images comprises:

utilizing a medical imaging technique, including any of magnetic resonance imaging (MRI), X-ray imaging, and CT scan imaging.

25. A method in accordance with claim 21, wherein said step of acquiring reference and initial floating images comprises:

utilizing a medical imaging technique for detection of at least one of tumors and potential tumors in breast cancer.

26. A method in accordance with claim 21, wherein said step of acquiring reference and initial floating images comprises:

utilizing a medical imaging technique for obtaining images indicating suspect tissue exhibiting a more rapid intake or wash-in of contrast agent, as well as a more rapid washout than adjacent, non-tumor tissue.

27. A method in accordance with claim 26, comprising a step of utilizing said motion-corrected final warped floating image to detect suspect tissue through comparison of images of a patient made before and after such wash-in and/or washout.

28. A method in accordance with claim 26, comprising a step of utilizing said retrieving at least one of said images from any of:

a storage medium, a computer, a radio link, the Internet, an infrared link, an acoustic link, a scanning device, and a live imaging device.

46. A method in accordance with claim 27, comprising a step of utilizing said motion-corrected image to detect suspect tissue through comparison of images of a patient made before and after at least one of such wash-in and washout.

29. A method for performing image motion compensation, comprising:

acquiring a reference image and an initial floating image;

warping said initial floating image with a given initial deformation function so as to obtain a current warped floating image;

computing local cross-correlation between respective corresponding regions in said reference image and said warped floating image;

performing a conjugate gradient maximization of a similarity measure based on said local cross-correlation so as to derive a deformation function for which said similarity measure is maximal, by utilizing a combination of a conjugate gradient optimization with a composition of displacements which are small as compared with the size of said images whereby fast convergence is achieved while ensuring invertibility so that said deformation stays non-singular; and

warping said floating image in accordance with said subsequent deformation function to obtain a motion-corrected image.

30. A method in accordance with claim 29, comprising:

initializing said initial deformation function at the top level of a multi-resolution pyramid having a given number of progressively decreasing image resolution levels, with the lowest resolution level being at the top;

extrapolating said initial deformation function to the next lower level;

determining a next lower deformation function at said next lower level;

if said next lower level is the lowest level, then warping said floating image using a deformation function determined at said lowest level, and if not,

then performing a cycle of steps comprising;

extrapolating said next lower deformation function to a next yet lower level, and

determining a yet lower deformation function at said yet lower level,

repeatedly as needed until the lowest level is reached; and

warping said floating image using a deformation function determined at said lowest level.

31. A method in accordance with claim 29, comprising:

forming said progressively decreasing resolution levels by progressively reducing the size of image planes used in forming said levels.

32. A method in accordance with claim 31, comprising:

progressively reducing the size of said image planes by using a low-pass filter for reducing in half the sampling frequency for said image planes.

33. A method in accordance with claim 31, wherein said step of performing a conjugate gradient maximization of said similarity measure comprises utilizing a combination of a conjugate gradient optimization with a composition of displacements which are small as compared with the size of said images whereby fast convergence is achieved while ensuring that said deformation stays nonsingular.

34. A method in accordance with claim 31, wherein said step of performing a conjugate gradient maximization of said similarity measure comprises utilizing a combination of a conjugate gradient optimization with a composition of displacements which are small as compared with the size of said images, whereby fast convergence is achieved while ensuring that said deformation stays nonsingular.

35. A method in accordance with claim 31, wherein said step of performing a conjugate gradient maximization of said similarity measure comprises utilizing a combination of a conjugate gradient optimization with composition of respective regularized gradients of said similarity measure, whereby fast convergence is achieved while ensuring that said deformation stays nonsingular.

36. A method in accordance with claim 29, wherein said step of deriving a deformation comprises deriving a deformation exhibiting no singularities.

37. A method in accordance with claim 29, wherein said step of deriving a deformation function comprises applying the algorithm:

$\{\begin{array}{c}\mathrm{Let}<br><br>\text{\hspace{1em}<br><br>}<br><br>U_0=\mathrm{id}\\ \begin{array}{c}\mathrm{For}<br><br>\mathrm{ <br><br> <br><br> <br><br>}<br><br>k=1<br><br>\text{\hspace{1em}<br><br>}<br><br>\mathrm{\cdots <br><br>}<br><br>\text{\hspace{1em}<br><br>}<br><br>n<br><br>\text{;<br><br>}\\ <br><br>v_k=G_{\mathrm{\sigma <br><br>}}*\nabla <br><br>S<br><br>\left(I_1,I_2<br><br>\text{\hspace{1em}<br><br>}\circ <br><br>\text{\hspace{1em}<br><br>}<br><br>{\mathrm{\varphi <br><br>}}_k\right)\\ <br><br>{\mathrm{\varphi <br><br>}}_{k+1}={\mathrm{\varphi <br><br>}}_k<br><br>\text{\hspace{1em}<br><br>}\circ <br><br>\text{\hspace{1em}<br><br>}<br><br>\left(\mathrm{id}+{\mathrm{\varepsilon <br><br>}}_k<br><br>v_k\right)\end{array}\\ \mathrm{End}\end{array}\hspace{1em}<br><br>$

wherein operator G_σ

* denotes convolution by a Gaussian kernel,

S denotes the similarity measure,

∇

S denotes the gradient of S with respect to φ

,

I_i;

→

R denotes said reference image,

I₂;

→

R denotes said floating image,

φ

;

Ω

→

R³ denotes said deformation such that said similarity measure is maximized, and

∘

denotes composition of functions, and

ε

_k is the step size, which is made sufficiently small to ensure invertibility.

38. A method in accordance with claim 37, wherein said step of deriving a deformation comprises deriving a deformation by composition of displacements which are small as compared with the size of said images.

39. A method in accordance with claim 37, wherein said step of deriving a deformation comprises deriving a deformation by composition of respective regularized gradients of said similarity measure.

40. A method in accordance with claim 37, wherein said step of warping comprises computing said deformation composed with said floating image.

41. A method in accordance with claim 37, wherein said step of warping comprises computing (I₂∘

φ

) (x), wherein I₂;

→

R denotes said floating image, φ

;

Ω

→

R³ denotes said deformation such that said similarity measure is maximized, and ∘

denotes composition of functions.

42. A method in accordance with claim 41, wherein said step of computing comprises utilizing tri-linear interpolation at each voxel of said floating image being warped.

43. A method in accordance with claim 29, wherein said step of acquiring a reference image and a floating image comprises:

utilizing a medical imaging technique, including any of magnetic resonance imaging (MRI), X-ray imaging, and CT scan imaging.

44. A method in accordance with claim 29, wherein said step of obtaining a reference image and a floating image comprises:

utilizing a medical imaging technique for detection of at least one of tumors and potential tumors in breast cancer.

45. A method in accordance with claim 29, wherein said step of obtaining a reference image and a floating image comprises:

utilizing a medical imaging technique for obtaining images indicating suspect tissue exhibiting at least one of (a) a more rapid intake (wash-in) of contrast agent and (b) a more rapid washout than adjacent, non-tumor tissue.

47. A method for performing image motion compensation, comprising:

inputting a given temporal sequence of images, I₁, . . . I_i, . . . I_n;

selecting an image of said temporal sequence with index k as a reference image;

starting with i=1, then if i is not equal to k,

performing a multi-resolution motion correction between I_k and I by

performing a conjugate gradient maximization of a similarity measure between first and second images in said temporal sequence;

deriving a deformation function by utilizing said gradient maximization in conjunction with composition criteria so as to ensure non-singularity of said deformation function;

warping one of said first and second images by said deformation function;

following such correction, defining J₁ as an output of the foregoing step of performing a multi-resolution motion correction until;

in the event that i is equal to k,

defining J_i as I_k and incrementing i by 1,

comparing the incremented result with n;

in the event the incremented result is less than or equal to n,

continuing to repeat the foregoing steps until the incremented result is greater than n, and thereupon

terminating said steps, thereby resulting in a series of respective motion-corrected images, J₁, J₂. J_n.

48. A method in accordance with claim 47, wherein said multi-resolution motion correction comprises:

determining a number of levels required for a multi-resolution pyramid;

constructing said multi-resolution pyramid with a lowest resolution top level, proceeding down in levels of increasing resolution;

initializing said deformation function to zero at the top level of said pyramid;

determining said deformation function, level by level down said pyramid, a current deformation function being determined within a current level based on initialization using the foregoing deformation function, and in accordance with the following steps;

if said current level is not the bottom level, then extrapolating said current deformation function to the next lower level,

determining a deformation function at said next lower level; and

when said bottom level of the pyramid is reached, then warping said floating image with said deformation function at said next lower level.

49. A method in accordance with claim 48, wherein said step of determining said deformation function comprises:

computing local cross-correlation between regions in said reference image and corresponding respective regions in said floating image;

determining a similarity measure between said reference image and said floating image; and

performing a conjugate gradient maximization of said similarity measure based on said local cross-correlation so as to derive a deformation function for which said similarity measure is maximal.

50. A method in accordance with claim 49, comprising:

performing said local cross-correlation between images I₁ and I₂ in accordance with the following;

$J_{\mathrm{CC}}={\int <br><br>}_{\mathrm{\Omega <br><br>}}^{\text{\hspace{1em}<br><br>}}<br><br>J_{\mathrm{CC}}<br><br>\left(x\right)<br><br>\text{\hspace{1em}<br><br>}<br><br>d<br><br>x={\int <br><br>}_{\mathrm{\Omega <br><br>}}^{\text{\hspace{1em}<br><br>}}<br><br>\frac{{v_{1,2}<br><br>\left(x\right)}^2}{v_1<br><br>\left(x\right)<br><br>\text{\hspace{1em}<br><br>}<br><br>v_2<br><br>\left(x\right)}<br><br>\text{\hspace{1em}<br><br>}<br><br>d<br><br>x,$

where ν

_1,2(x), ν

₁(x) and ν

₂(x) are respectively the covariance and variances of the intensities of I₁ and I₂ around x;

whereof the first order variation defines a gradient given by

$\nabla <br><br>J_{\mathrm{CC}}<br><br>\left(x\right)=f_{\mathrm{CC}}<br><br>\left(I<br><br>\left(x\right),x\right)<br><br>\nabla <br><br>I_2<br><br>\left(x\right),<br><br>\mathrm{where}$$f_{\mathrm{CC}}<br><br>\left(i,x\right)=G_{\mathrm{\gamma <br><br>}}*L_{\mathrm{CC}}<br><br>\left(i,x\right),\mathrm{and}$$L_{\mathrm{CC}}<br><br>\left(i,x\right)=\frac{v_{1,2}<br><br>\left(x\right)}{v_2<br><br>\left(x\right)}<br><br>\left(\frac{i_1-{\mathrm{\mu <br><br>}}_1<br><br>\left(x\right)}{v_1<br><br>\left(x\right)}\right)+J_{\mathrm{CC}}<br><br>\left(x\right)<br><br>\left(\frac{i_2-{\mathrm{\mu <br><br>}}_2<br><br>\left(x\right)}{v_2<br><br>\left(x\right)}\right),$

function L_CC is estimated as

$L_{\mathrm{CC}}<br><br>\left(i,x\right)=\left(G_{\mathrm{\gamma <br><br>}}*f_1\right)<br><br>\left(x\right)<br><br>i_1+\left(G*f_2\right)<br><br>\left(x\right)<br><br>i_2+\left(G_{\mathrm{\gamma <br><br>}}*f_3\right)<br><br>\left(x\right),<br><br>\mathrm{where}$$\begin{array}{c}{f}_1<br><br>\left(x\right)=\frac{v_{1,2}<br><br>\left(x\right)}{v_1<br><br>\left(x\right)<br><br>v_2<br><br>\left(x\right)},\\ f_2<br><br>\left(x\right)=\frac{J_{\mathrm{CC}}<br><br>\left(x\right)}{v_2<br><br>\left(x\right)},\mathrm{and}\\ f_3<br><br>\left(x\right)=f_1<br><br>\left(x\right)<br><br>{\mathrm{\mu <br><br>}}_1<br><br>\left(x\right)+f_2<br><br>\left(x\right)<br><br>{\mathrm{\mu <br><br>}}_2<br><br>\left(x\right);<br><br>\end{array}$

all required space dependent quantities including μ

₁(x) are computed by convolution with a Gaussian kernel.

51. A method in accordance with claim 48, wherein said step of determining said deformation function comprises:

assigning the initial value of k as zero;

warping F with U to obtain F;

computing the similarity measure between U and F;

defining G as the gradient of the similarity measure with respect to U_k;

if G_U is not zero and k is equal to or greater than 1, then conjugating G_U, with G_Uk-1;

determining, using the conjugation of G_Uk, with G_Uk-1, whether an optimal step has been found;

if an optimal step has been found, then updating U_k with G_Uk, determining whether k≧

k_max;

if k≧

k_max, the process is ended, and if not,

then k is incremented by 1;

using the incremented value of k, warping F with U_K to obtain F_U

following again with the foregoing steps for computing the similarity measure, and so forth; and

if G_U is equal to zero, then ending the process, whereby the output is an improved version U_k of the deformation function.

52. A method in accordance with claim 48, wherein said step of constructing said multi-resolution pyramid comprises:

reducing only the size of said images in constructing said pyramid.

53. A method in accordance with claim 52, wherein said step of reducing only the size of said images in constructing said pyramid comprises:

using a low-pass filter designed for reducing in half the sampling frequency with minimal loss of information.

54. A method for performing image motion compensation between images in a temporal sequence by computing a deformation function from an initial deformation function, said method comprising:

(a) acquiring a reference image and an initial floating image;

(b) setting said initial deformation function as a current deformation function;

(c) warping said initial floating image with said current deformation function so as to obtain a current warped floating image;

(d) computing a current similarity measure between said reference image and said current warped floating image;

(e) computing the current gradient of said current similarity measure with respect to said current deformation function;

(f) regularizing said current gradient so as to ensure its invertibility to obtain a current regularized gradient;

(g) (A) in the first performance of this step, setting said current regularized gradient as current conjugate gradient, and

(B) in subsequent iterations of this step, conjugating said current regularized gradient with said current conjugate gradient so as to obtain a subsequent conjugate gradient;

(h) composing said subsequent conjugate gradient with said current deformation function so as to obtain a subsequent deformation function;

(i) setting said current deformation function as said subsequent deformation function;

(j) setting said current conjugate gradient as said subsequent conjugate gradient;

(k) going to step (c) and repeating until a predetermined stop criterion is reached; and

(l) defining the current warped floating image as a final warped floating image.

55. A method in accordance with claim 54, wherein said predetermined stop criterion comprises reaching at least one of:

a predetermined maximum number of repetitions or iterations of the recited sequence of steps;

a predetermined lower limit for said gradient;

zero for said gradient; and

a predetermined processing time limit.

56. A method in accordance with claim 54 comprising beginning iterations at a low resolution level and proceeding through increasingly higher resolutions.

57. A method in accordance with claim 54, wherein said step of acquiring a reference image and an initial floating image comprises:

utilizing a medical imaging technique, including any of magnetic resonance imaging (MRI), X-ray imaging, and CT scan imaging.

58. A method in accordance with claim 57, wherein said step of acquiring a reference image and an initial floating comprises:

utilizing a medical imaging technique for detection of at least one of tumors and potential tumors in breast cancer.

59. A method in accordance with claim 57, wherein said step of acquiring a reference image and an initial floating image comprises:

utilizing a medical imaging technique for obtaining images indicating suspect tissue exhibiting a more rapid intake (wash-in) of contrast agent, as well as a more rapid washout than adjacent, non-tumor tissue.

60. A method in accordance with claim 57, comprising a step of utilizing said motion-corrected image to detect suspect tissue through comparison of images of a patient made before and after such wash-in and/or washout.

61. A method in accordance with claim 54, wherein said step of acquiring reference and initial floating images comprises:

retrieving at least one of said images from any of;

a storage medium, a computer, a radio link, the Internet, an infrared link, an acoustic link, a scanning device, and a live imaging device.

62. A system for performing image motion compensation, comprising:

a memory device for storing a program and other data; and

a processor in communication with said memory device, said processor operative with said program to perform;

(a) setting said initial deformation function as a current deformation function;

(b) warping said initial floating image with said current deformation function so as to obtain a current warped floating image;

(c) computing a current similarity measure between said reference image and said current warped floating image;

(d) computing the current gradient of said current similarity measure with respect to said current deformation function;

(e) regularizing said current gradient so as to ensure its invertibility to obtain a current regularized gradient;

(f) (A) in the first performance of this step, setting said current regularized gradient as current conjugate gradient, and

(B) in subsequent iterations of this step, conjugating said current regularized gradient with said current conjugate gradient so as to obtain a subsequent conjugate gradient;

(g) composing said subsequent conjugate gradient with said current deformation function so as to obtain a subsequent deformation function;

(h) setting said current deformation function as said subsequent deformation function;

(i) setting said current conjugate gradient as said subsequent conjugate gradient;

(j) going to step (b) and repeating until reaching a predetermined stop criterion; and

(k) defining the current warped floating image as a final warped floating image.

63. A system in accordance with claim 62, including

performing said iterations beginning at a low resolution level and proceeding through increasingly higher resolutions.

64. A system in accordance with claim 62, wherein reaching said stop criterion comprises at least one of reaching:

a predetermined maximum number of repetitions or iterations of the recited sequence of steps;

a predetermined lower limit for said gradient;

zero for said gradient; and

a predetermined processing time limit.

65. A computer program product comprising a computer useable medium having computer program logic recorded thereon for program code for performing image motion compensation by:

(a) setting said initial deformation function as a current deformation function;

(b) warping said initial floating image with said current deformation function so as to obtain a current warped floating image;

(c) computing a current similarity measure between said reference image and said current warped floating image;

(d) computing the current gradient of said current similarity measure with respect to said current deformation function;

(e) regularizing said current gradient so as to ensure its invertibility to obtain a current regularized gradient;

(f) (A) in the first performance of this step, setting said current regularized gradient as current conjugate gradient, and

(B) in subsequent iterations of this step, conjugating said current regularized gradient with said current conjugate gradient so as to obtain a subsequent conjugate gradient;

(g) composing said subsequent conjugate gradient with said current deformation function so as to obtain a subsequent deformation function;

(h) setting said current deformation function as said subsequent deformation function;

(i) setting said current conjugate gradient as said subsequent conjugate gradient;

(j) going to step (b) and repeating until reaching a predetermined stop criterion; and

(k) define the current warped floating image as a final warped floating image.

66. A computer program product in accordance with claim 65, including

performing said iterations beginning at a low resolution level and proceeding through increasingly higher resolutions.

67. A computer program product in accordance with claim 65, wherein said stop criterion comprises at least one of reaching:

a predetermined maximum number of repetitions or iterations of the recited sequence of steps;

a predetermined lower limit for said gradient;

zero for said gradient; and

a predetermined processing time limit.