Automatic system calibration method of X-ray CT

• US 10,643,354 B2
• Filed: 03/21/2016
• Issued: 05/05/2020
• Est. Priority Date: 03/20/2015
• Status: Active Grant

• Current Assignee: Rensselaer Polytechnic Institute
• Original Assignee: Rensselaer Polytechnic Institute
• Inventors: Wang, Ge, Chen, Mianyi, Xi, Yan, Cong, Wenxiang
• Primary Examiner(s): Liew, Alex Kok S
• Application Number:
  US15/558,053
• Publication Number:
  US 20180068467A1
• Time in Patent Office: 1,506 Days
• Field of Search: 382128-134
• CPC Class Codes: A61B 6/032 Transmission computed tomography [CT]
• CPC Class Codes: A61B 6/5205 involving processing of raw data to produce diagnostic data
• CPC Class Codes: A61B 6/582 Calibration
• CPC Class Codes: G06T 11/005 Specific pre-processing for tomographic reconstruction, e.g. calibration, source positioning, rebinning, scatter correction, retrospective gating
• CPC Class Codes: G06T 11/006 Inverse problem, transformation from projection-space into object-space, e.g. transform methods, back-projection, algebraic methods
• CPC Class Codes: G06T 2207/10116 X-ray image
• CPC Class Codes: G06T 2207/20 Special algorithmic details
• CPC Class Codes: G06T 2211/412 Dynamic
• CPC Class Codes: G06T 2211/424 Iterative

Abstract

Systems and methods for geometric calibration and image reconstruction in computed tomography (CT) scanning using iterative reconstruction algorithms are provided. An iterative reconstruction algorithm can be used to reconstruct an improved image, and then the improved image can be used to adjust inaccurate parameters by using a Locally Linear Embedding (LLE) method. Adjusted parameters can then be used to reconstruct new images, which can then be used to further adjust the parameters. The steps of this iterative process can be repeated until a quality threshold is met.

Claims

1. A method of reconstructing a computed tomography (CT) image, the method comprising:

i) obtaining an initial CT image;

ii) performing a reconstruction algorithm on the initial CT image to obtain a reconstructed image of the initial CT image;

iii) using the reconstructed image to adjust one or more parameters associated with the image comprising using a Locally Linear Embedding (LLE) method on each of the one or more parameters;

iv) using the adjusted one or more parameters to perform the reconstruction algorithm on the reconstructed image to obtain an updated reconstructed image; and

v) repeating steps iii)-iv), using the most-recently updated reconstructed image in each repeated step iii), and updating the same one or more parameters in each repeated step iii) until a threshold value of a predetermined characteristic is met;

wherein the LLE method comprises;

with the K nearest vectors, representing the original data vector linearly with its neighboring vectors;

2. The method according to claim 1, wherein the reconstruction algorithm is based on Total Variation (TV).

3. The method according to claim 1, wherein the LLE method includes:

a) finding the K nearest neighbors for each high-dimensional point in terms of the Euclidean distance;

b) representing each point linearly with its neighbors and calculating weight coefficients for its neighbors; and

c) mapping high dimensional data with the weight coefficients to a low dimensional representation on an intrinsic manifold.

4. The method according to claim 1, wherein the LLE method includes:

finding the K nearest neighbors of a data vector b_i in its dataset vectors {tilde over (b)}_ij according to the Euclidean distance;

d_ij=∥

b_i−

{tilde over (b)}_ij∥

₂².

5. The method according to claim 1, wherein the LLE method (further) includes, calculating, with the weight coefficients W=(w_ik), the global internal coordinate Y=(y_i) by solving the equation:

6. The method according to claim 5, wherein the solution of the equation that can be solved to calculate Y=(y_i) is given by the bottom d+1 eigenvectors of the generalized eigenvalue problem:

MY=λ

Y,

where λ

is the eigenvalue, and M=(I−

W)^T(I−

W).

7. The method according to claim 1, wherein the number of nearest neighbors used in the LLE method is at least 2.

8. The method according to claim 1, wherein the LLE method includes solving the following linear system of equations:

Au=b_i,

where u=(u₁, u₂, . . . , u_J) is an image represented as a J dimensional vector, J is the number of pixels, b_i=(b₁, b₂, . . . , b_L) is data, L is the number of vector elements, and A=(a_jk) is a projection matrix related to the one or more (geometric) parameters,

wherein performing the reconstruction algorithm comprises calculating a projection matrix, which is affected by the one or more (geometric) parameters;

A=A(P),

where P is an estimated parameter vector, which includes the one or more parameters,

wherein the projections of the projection matrix are calculated using a distance-driven method,

wherein using the reconstructed image to adjust one or more parameters comprises minimizing the mean squared error between the projection data and re-projected projection data, which can be formulated as;

P=arg min∥

b_i−

{tilde over (b)}_ij∥

₂²s.t. A(P)u=b_i,

where b_i is the projection vector obtained from the measurement along different projection views, {tilde over (b)}_ij is the corresponding re-projected projection vector from a reconstructed image with sampled parameters, and P is the updated vector of parameters, and

wherein the re-projected projection vector can be calculated within a densely sampled parametric range;

{tilde over (P)}_j=(p_j1,p_j2, . . . ,p_jn).

9. The method according to claim 8, wherein a true parameter vector is close to neighboring sampled parameter vectors, and the measured projection vector can be linearly expressed by the K nearest re-projected projection vectors associated with the sampled parameter vectors, such that:

10. The method according to claim 1, wherein the predetermined characteristic, of which a threshold value must be met to stop repeating steps iii)-iv) is universal quality index (UQI).

11. The method according to claim 10, wherein the threshold value is at least 0.6.

12. The method according to claim 10, wherein the threshold value is at least 0.95.

13. The method according to claim 1, wherein the one or more parameters includes at least one of projection angle, source-to-object distance (SOD), object-to-detector distance (ODD), detector offset, detector tilt angle, object x-coordinate offset, object y-coordinate offset, and projection angular error.

14. The method according to claim 1, wherein obtaining the initial CT image comprises obtaining an initial parameter vector including initial value(s) for the one or more parameters associated with the image, and

wherein performing the reconstruction algorithm comprises updating the parameter vector with updated values of the one or more parameters based on the most recently reconstructed image.

15. The method according to claim 1, wherein the reconstruction algorithm comprises using an Ordered Subset Simultaneous Algebraic Reconstruction Technique (OS-SART).

16. The method according to claim 1, wherein each of step ii), step iii), step iv), and step v) is performed by a processor.

17. A computer readable medium having computer-executable instructions for performing the method according to claim 1.

18. A CT system, comprising:

an X-ray source;

a detector for detecting X-ray radiation from the X-ray source; and

a computer having the computer readable medium according to claim 17.

19. A method of reconstructing a computed tomography (CT) image, the method comprising:

i) obtaining an initial CT image;

ii) performing a reconstruction algorithm on the initial CT image to obtain a reconstructed image of the initial CT image;

iii) using the reconstructed image to adjust one or more parameters associated with the image comprising using a Locally Linear Embedding (LLE) method on each of the one or more parameters;

iv) using the adjusted one or more parameters to perform the reconstruction algorithm on the reconstructed image to obtain an updated reconstructed image; and

v) repeating steps iii)-iv), using the most-recently updated reconstructed image in each repeated step iii), and updating the same one or more parameters in each repeated step iii) until a threshold value of a predetermined characteristic is met;

wherein the LLE method comprises;

calculating, with the weight coefficients W=(w_ik), the global internal coordinate Y=(y_i) by solving the equation;

20. The method according to claim 19, wherein the solution of the equation that can be solved to calculate Y=(y_i) is given by the bottom d+1 eigenvectors of the generalized eigenvalue problem:

MY=λ

Y,

where λ

is the eigenvalue, and M=(I−

W)^T(I−

W).

21. The method according to claim 19, wherein the LLE method further comprises:

finding the K nearest neighbors of a data vector b_i in its dataset vectors {tilde over (b)}_ij according to the Euclidean distance;

d_ij=∥

b_i−

{tilde over (b)}_ij∥

₂².

22. The method according to claim 19, wherein the LLE method further comprises:

with the K nearest vectors, representing the original data vector linearly with its neighboring vectors;

23. The method of claim 19, wherein the LLE method further comprises solving the following linear system of equations:

Au=b_i,

where u=(u₁, u₂, . . . , u_J) is an image represented as a J dimensional vector, J is the number of pixels, b_i=(b₁, b₂, . . . , b_L) is data, L is the number of vector elements, and A=(a_jk) is a projection matrix related to the one or more (geometric) parameters,

wherein performing the reconstruction algorithm comprises calculating a projection matrix, which is affected by the one or more (geometric) parameters;

A=A(P),

where P is an estimated parameter vector, which includes the one or more parameters,

wherein the projections of the projection matrix are calculated using a distance-driven method,

wherein using the reconstructed image to adjust one or more parameters comprises minimizing the mean squared error between the projection data and re-projected projection data, which can be formulated as;

P=arg min∥

b_i−

{tilde over (b)}_ij∥

₂²s.t. A(P)u=b_i,

where b_i is the projection vector obtained from the measurement along different projection views, {tilde over (b)}_ij is the corresponding re-projected projection vector from a reconstructed image with sampled parameters, and P is the updated vector of parameters, and

wherein the re-projected projection vector can be calculated by the equations provide in claims 8 within a densely sampled parametric range;

{tilde over (P)}_j=(p_j1,p_j2, . . . ,p_jn).

24. The method according to claim 23, wherein a true parameter vector is close to neighboring sampled parameter vectors, and the measured projection vector can be linearly expressed by the K nearest re-projected projection vectors associated with the sampled parameter vectors, such that:

25. A method of reconstructing a computed tomography (CT) image, the method comprising:

i) obtaining an initial CT image;

ii) performing a reconstruction algorithm on the initial CT image to obtain a reconstructed image of the initial CT image;

iii) using the reconstructed image to adjust one or more parameters associated with the image comprising using a Locally Linear Embedding (LLE) method on each of the one or more parameters;

iv) using the adjusted one or more parameters to perform the reconstruction algorithm on the reconstructed image to obtain an updated reconstructed image; and

v) repeating steps iii)-iv), using the most-recently updated reconstructed image in each repeated step iii), and updating the same one or more parameters in each repeated step iii) until a threshold value of a predetermined characteristic is met;

wherein the LLE method comprises solving the following linear system of equations;

Au=b_i,

where u=(u₁, u₂, . . . , u_J) is an image represented as a J dimensional vector, J is the number of pixels, b_i=(b₁, b₂, . . . , b_L) is data, L is the number of vector elements, and A=(a_jk) is a projection matrix related to the one or more (geometric) parameters,

wherein performing the reconstruction algorithm comprises calculating a projection matrix, which is affected by the one or more (geometric) parameters;

A=A(P),

where P is an estimated parameter vector, which includes the one or more parameters,

wherein the projections of the projection matrix are calculated using a distance-driven method,

wherein using the reconstructed image to adjust one or more parameters comprises minimizing the mean squared error between the projection data and re-projected projection data, which can be formulated as;

P=arg min∥

b_i−

{tilde over (b)}_ij∥

₂²s.t. A(P)u=b_i,

where b_i is the projection vector obtained from the measurement along different projection views, {tilde over (b)}_ij is the corresponding re-projected projection vector from a reconstructed image with sampled parameters, and P is the updated vector of parameters, and

wherein the re-projected projection vector can be calculated within a densely sampled parametric range;

{tilde over (P)}_j=(p_j1,p_j2, . . . ,p_jn).

26. The method according to claim 25, wherein a true parameter vector is close to neighboring sampled parameter vectors, and the measured projection vector can be linearly expressed by the K nearest re-projected projection vectors associated with the sampled parameter vectors, such that:

27. The method according to claim 25, wherein the LLE method further comprises:

finding the K nearest neighbors of a data vector b_i in its dataset vectors {tilde over (b)}_ij according to the Euclidean distance;

d_ij=∥

b_i−

{tilde over (b)}_ij∥

₂².

28. The method according to claim 25, wherein the LLE method further comprises:

with the K nearest vectors, representing the original data vector linearly with its neighboring vectors;

29. The method according to claim 25, wherein the LLE method further comprises, calculating, with the weight coefficients W=(w_ik), the global internal coordinate Y=(y_i) by solving the equation:

30. The method according to claim 29, wherein the solution of the equation that can be solved to calculate Y=(y_i) is given by the bottom d+1 eigenvectors of the generalized eigenvalue problem:

MY=λ

Y,

where λ

is the eigenvalue, and M=(I−

W)^T(I−

W).