SYSTEM AND METHOD FOR CARDIAC SEGMENTATION IN MR-CINE DATA USING INVERSE CONSISTENT NON-RIGID REGISTRATION

• US 20110081066A1
• Filed: 11/30/2010
• Published: 04/07/2011
• Est. Priority Date: 10/07/2009
• Status: Active Grant

• Current Assignee: Siemens Healthcare GMBH (Siemens AG)
• Original Assignee: Siemens AG
• Inventors: Guetter, Christoph, Jolly, Marie-Pierre, Guehring, Jens
• Granted Patent: US 8,494,236 B2
• Time in Patent Office: Days
• US Class Current: 382/131
• CPC Class Codes: G06T 2207/10076 4D tomography; Time-sequential 3D tomography
• CPC Class Codes: G06T 2207/30048 Heart; Cardiac
• CPC Class Codes: G06T 7/12 Edge-based segmentation
• CPC Class Codes: G06T 7/174 involving the use of two or more images
• CPC Class Codes: G06V 2201/031 of internal organs

Abstract

A method for cardiac segmentation in magnetic resonance (MR) cine data, includes providing a time series of 3D cardiac MR images acquired at a plurality of phases over at least one cardiac cycle, in which each 3D image includes a plurality of 2D slices, and a heart and blood pool has been detected in each image. Gray scales of each image are analyzed to compute histograms of the blood pool and myocardium. Non-rigid registration deformation fields are calculated to register a selected image slice with corresponding slices in each phase. Endocardium and epicardium gradients are calculated for one phase of the selected image slice. Contours for the endocardium and epicardium are computed from the gradients in the one phase, and the endocardium and epicardium contours are recovered in all phases of the selected image slice. The recovered endocardium and epicardium contours segment the heart in the selected image slice.

Claims

1. A method for cardiac segmentation in magnetic resonance (MR) cine data, the method comprising the steps of:

providing a time series of 3D cardiac MR images acquired at a plurality of phases over at least one cardiac cycle, wherein each 3D image comprises a plurality of 2D slices, wherein a heart and blood pool has been detected in each image;

analyzing gray scales of each image in the time series of images to compute histograms of the blood pool, myocardium, and the whole series of images;

calculating non-rigid registration deformation fields to register a selected image slice with corresponding slices in each phase of the time series of images;

computing endocardium and epicardium gradients for one phase of the selected image slice;

computing contours for the endocardium and epicardium from said gradients in said one phase of said selected image slice; and

recovering the endocardium and epicardium contours in all phases of the selected image slice, wherein said recovered endocardium and epicardium contours segment the heart in said selected image slice.

2. The method of claim 1, further comprising using image pixels in peaks of said histograms to seed a multiseeded fuzzy connected algorithm to group image pixels into homogenous regions for the heart, blood pool, and background to form a label image, wherein said heart, blood pool, and background are labeled.

3. The method of claim 1, wherein calculating non-rigid registration deformation fields to register a selected image slice with corresponding slices in each phase of the time series of images comprises finding an inverse consistent deformation field Φ

_pq;

R²R² between two phases f_p and f_q of a given slice that minimizes an inverse consistent similarity metric=J_C^ic=J_CC(f_p,f_q,Φ

_pq)+J_CC(f_p,f_q,Φ

_qp).

4. The method of claim 3, wherein said similarity metric is defined as

5. The method of claim 3, wherein the similarity metric is minimized using variational calculus and solving a partial differential equation that descends a gradient of J_CC.

6. The method of claim 2, wherein computing endocardium and epicardium gradients comprises:

computing probability images P_M and P_B as responses from the myocardium and blood pool histograms;

computing gradients of said probability images and of the selected image slice;

analyzing said label image L to find upwards transitions from the myocardium to non-myocardium, and smoothing said transitions; and

computing the gradients for the endocardium and epicardium according to

7. The method of claim 6, wherein computing contours for the endocardium and epicardium from said gradients comprises using a shortest path algorithm with a contour energy function of the form

8. The method of claim 1, wherein recovering the endocardium and epicardium contours in all phases of the selected image slice comprises:

propagating the endocardium and epicardium contours to all other phases of the selected image slice using the non-rigid registration deformation fields;

finding the endocardium and epicardium contours in each of the other phases of the selected image slice;

finding a best contour for the selected image slice by finding a phase having a contour with a lowest energy, wherein the energy K(p) of a contour in a given phase p is K(p)=Σ

_qE(C_q(C_p)), wherein C_p is the contour in phase p, C_q is the contour in phase q, and E is the energy of the contour in phase q; and

propagating said best contour to the other phases using the non-rigid registration deformation fields.

9. The method of claim 8, further comprising:

propagating said best contour to other slices using the non-rigid registration deformation fields; and

finding a best contour for each other slice by calculating

10. The method of claim 9, wherein for a first slice to be segmented, finding a best contour comprises minimizing

11. A method for cardiac segmentation in magnetic resonance (MR) cine data, the method comprising the steps of:

providing a time series of 3D cardiac MR images acquired at a plurality of phases over at least one cardiac cycle, wherein each 3D image comprises a plurality of 2D slices, wherein a heart and blood pool has been detected in each image;

calculating inverse consistent non-rigid registration deformation fields Φ

_pq;

R²R² to register a selected image slice f_p with corresponding slices f_q in each phase q of the time series of images;

computing contours for the endocardium and epicardium in said one phase of said selected image slice;

recovering the endocardium and epicardium contours in all phases of the selected image slice, wherein said recovered endocardium and epicardium contours segment the heart in said selected image slice;

propagating said endocardium and epicardium contours to other slices using the non-rigid registration deformation fields; and

finding best endocardium and epicardium contours for each other slice using the contour propagated from a previous slice as a prior, wherein said best endocardium and epicardium contours segment the heart in said time series of images.

12. The method of claim 11, further comprising analyzing gray scales of each image in the time series of images to compute histograms of the blood pool, myocardium, and the whole series of images;

and

using image pixels in peaks of said histograms to seed a multiseeded fuzzy connected algorithm to group image pixels into homogenous regions for the heart, blood pool, and background to form a label image, wherein said heart, blood pool, and background are labeled.

13. The method of claim 12, wherein computing contours for the endocardium and epicardium in said one phase of said selected image slice comprises:

computing probability images P_M and P_B as responses from the myocardium and blood pool histograms;

computing gradients of said probability images and of the selected image slice;

analyzing said label image L to find upwards transitions from the myocardium to non-myocardium, and smoothing said transitions;

computing the gradients for the endocardium and epicardium according to

14. The method of claim 11, wherein said inverse consistent non-rigid registration deformation fields minimize an inverse consistent similarity metric J_C^ic=J_CC(f_p,f_q,Φ

_pq)+J_CC(f_p,f_q,Φ

_qp), wherein

15. The method of claim 11, wherein finding best endocardium and epicardium contours for each other slice further comprises calculating

16. A program storage device readable by a computer, tangibly embodying a program of instructions executable by the computer to perform the method steps for cardiac segmentation in magnetic resonance (MR) cine data, the method comprising the steps of:

providing a time series of 3D cardiac MR images acquired at a plurality of phases over at least one cardiac cycle, wherein each 3D image comprises a plurality of 2D slices, wherein a heart and blood pool has been detected in each image;

analyzing gray scales of each image in the time series of images to compute histograms of the blood pool, myocardium, and the whole series of images;

calculating non-rigid registration defog nation fields to register a selected image slice with corresponding slices in each phase of the time series of images;

computing endocardium and epicardium gradients for one phase of the selected image slice;

computing contours for the endocardium and epicardium from said gradients in said one phase of said selected image slice; and

recovering the endocardium and epicardium contours in all phases of the selected image slice, wherein said recovered endocardium and epicardium contours segment the heart in said selected image slice.

17. The computer readable program storage device of claim 16, the method further comprising using image pixels in peaks of said histograms to seed a multiseeded fuzzy connected algorithm to group image pixels into homogenous regions for the heart, blood pool, and background to form a label image, wherein said heart, blood pool, and background are labeled.

18. The computer readable program storage device of claim 16, wherein calculating non-rigid registration deformation fields to register a selected image slice with corresponding slices in each phase of the time series of images comprises finding an inverse consistent deformation field Φ

_pq;

R²R² between two phases f_p and f_q of a given slice that minimizes an inverse consistent similarity metric J_C^ic=J_CC(f_p,f_q,Φ

_pq)+J_CC(f_q,f_q,Φ

_qp).

19. The computer readable program storage device of claim 18, wherein said similarity metric is defined as

20. The computer readable program storage device of claim 18, wherein the similarity metric is minimized using variational calculus and solving a partial differential equation that descends a gradient of J_CC.

21. The computer readable program storage device of claim 17, wherein computing endocardium and epicardium gradients comprises:

computing probability images P_M and P_B as responses from the myocardium and blood pool histograms;

computing gradients of said probability images and of the selected image slice;

analyzing said label image L to find upwards transitions from the myocardium to non-myocardium, and smoothing said transitions; and

computing the gradients for the endocardium and epicardium according to

22. The computer readable program storage device of claim 21, wherein computing contours for the endocardium and epicardium from said gradients comprises using a shortest path algorithm with a contour energy function of the form

23. The computer readable program storage device of claim 16, wherein recovering the endocardium and epicardium contours in all phases of the selected image slice comprises:

propagating the endocardium and epicardium contours to all other phases of the selected image slice using the non-rigid registration deformation fields;

finding the endocardium and epicardium contours in each of the other phases of the selected image slice;

finding a best contour for the selected image slice by finding a phase having a contour with a lowest energy, wherein the energy K(p) of a contour in a given phase p is K(p)=Σ

_qE(C_q(C_p)), wherein C_p is the contour in phase p, C_q is the contour in phase q, and E is the energy of the contour in phase q; and

propagating said best contour to the other phases using the non-rigid registration deformation fields.

24. The computer readable program storage device of claim 23, the method further comprising:

propagating said best contour to other slices using the non-rigid registration deformation fields; and

finding a best contour for each other slice by calculating

25. The computer readable program storage device of claim 24, wherein for a first slice to be segmented, finding a best contour comprises minimizing