SYSTEM AND METHOD FOR GENERATING MR MYOCARDIAL PERFUSION MAPS WITHOUT USER INTERACTION

• US 20090290778A1
• Filed: 04/29/2009
• Published: 11/26/2009
• Est. Priority Date: 05/22/2008
• Status: Active Grant

• Current Assignee: Siemens Healthineers AG (Siemens AG)
• Original Assignee: Siemens Corporate Research Incorporated (Siemens AG)
• Inventors: Sun, Ying, Gühring, Jens
• Granted Patent: US 8,139,838 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 7/149 involving deformable models, e.g. active contour models
• CPC Class Codes: G06V 10/25 Determination of region of interest [ROI] or a volume of interest [VOI]
• CPC Class Codes: G06V 10/267 by performing operations on regions, e.g. growing, shrinking or watersheds
• CPC Class Codes: G06V 10/755 Deformable models or variational models, e.g. snakes or active contours
• CPC Class Codes: G06V 2201/031 of internal organs

Abstract

A method for automatically generating a myocardial perfusion map from a sequence of magnetic resonance (MR) images includes determining a region of interest (ROI) in a reference frame selected from a time series of myocardial perfusion MR image slices, registering each image slice in the time series of slices to the reference frame to obtain a series of registered ROIs, and using the series of registered ROIs to segment endo- and epi-cardial boundaries of a myocardium in the ROI.

Claims

1. A method of automatically generating a myocardial perfusion map from a sequence of magnetic resonance (MR) images, said method comprising the steps of:

determining a region of interest (ROI) in a reference frame selected from a time series of myocardial perfusion MR image slices;

registering each image slice in said time series of slices to the reference frame to obtain a series of registered ROIs; and

using the series of registered ROIs to segment endo- and epi-cardial boundaries of a myocardium in the ROI.

2. The method of claim 1, wherein determining a region of interest (ROI) in a reference frame comprises:

cropping slices in said series of slices according to a temporal intensity variation at each pixel to obtain a global mask containing pixels exhibiting significant temporal intensity variation;

selecting a range of candidate slices with superior contrast;

detecting the LV blood pool and forming an LV mask in all candidate slices;

selecting one candidate slice with a detected LV blood pool as the reference frame;

detecting the right ventricle (RV) in the reference frame and forming an RV mask;

estimating a thickness of the myocardium based on the detected LV and RV masks; and

defining the ROI as a bounding box containing the LV, RV, and myocardium.

3. The method of claim 2, wherein detecting the LV blood pool comprises, for each candidate slice, selecting an intensity threshold for the blood, using said threshold to obtain a binary image and removing small connected components from said binary image.

4. The method of claim 3, wherein selecting one candidate slice as the reference frame comprises computing shape features for each remaining connected component to select a least eccentric, most circular, and most convex connected component, wherein the selected connected component among all among all candidate slices corresponds to the LV blood pool.

5. The method of claim 3, wherein said blood intensity threshold is selected through Gaussian clustering.

6. The method of claim 4, wherein said shape features include an eccentricity computed as a ratio between minimum and maximum radii on the connected components, a convexity computed as a ratio between an area of the connected component and an area of its convex hull, and a circularity of the convex hull defined as 4π

times the area of the convex hull over a square of a perimeter of said convex hull.

7. The method of claim 4, wherein when two connected components have similar shape features, a component with a larger area is selected.

8. The method of claim 2, wherein detecting the RV comprises finding a connected component in the binary image that is close to the LV and has a large area.

9. The method of claim 2, wherein segmenting endo- and epi-cardial boundaries of a myocardium in the ROI comprises:

initializing the endo-cardium and epi-cardium boundaries from the LV mask and myocardial thickness estimate;

detecting the endo-cardium by minimizing an energy functional;

masking out a left-ventricle (LV) blood pool inside of the endo-cardium from the image series, and detecting the epi-cardium by minimizing another energy functional,

wherein each said energy functional incorporates correlation information among pixels in a same slice as well as temporal correlations across the slices in the time series, prior shape information and anatomical constraints.

10. The method of claim 9, wherein the energy functional for the endo-cardium is E(C₁)=E_seq(C₁)+λ

E_img(C₁)+μ

E_int(C₁)+κ

E_shape(C₁), wherein C₁ is the endo-cardial boundary, λ

, μ

, and κ

are parameters that control weights of different terms, wherein

11. The method of claim 10, further comprising weighting each pixel on the endo-cardial boundary according to its distance to a convex hull of endo-cardial boundary, wherein pixels closer to the convex hull have a greater weight.

12. The method of claim 9, wherein the energy functional for the epi-cardium is E(C₂)=E_seq(C₂)+λ

E_img(C₂)+μ

E_int(C₂)+κ

E_shape(C₂)+E_con(C₂), wherein C₂ is the epi-cardial boundary, λ

, μ

, and κ

are parameters that control weights of different terms, wherein

13. A method of automatically generating a myocardial perfusion map from a sequence of magnetic resonance (MR) volumetric images, said method comprising the steps of:

providing a time series of registered myocardial perfusion MR image slices having reference image with a masked left ventricle (LV);

detecting the right ventricle (RV) in the reference frame and estimating a thickness of the myocardium based on the detected LV and RV;

initializing the endo-cardium and epi-cardium boundaries from the LV mask and myocardial thickness estimate;

segmenting the endo-cardium by minimizing an energy functional;

masking out a left-ventricle (LV) blood pool inside of the endo-cardium from the image series, and segmenting the epi-cardium by minimizing another energy functional,

wherein each said energy functional incorporates correlation information among pixels in a same slice as well as temporal correlations across the slices in the time series, prior shape information and anatomical constraints.

14. The method of claim 13, wherein providing a time series of registered myocardial perfusion MR image slices comprises providing a time series of myocardial perfusion MR image slices, determining a region of interest (ROI) in a reference frame selected from said time series, and registering each image slice in said time series of slices to the reference frame to obtain a series of registered ROIs, wherein the series of registered ROIs are used to segment endo- and epi-cardial boundaries of a myocardium in the ROI.

15. A program storage device readable by a computer, tangibly embodying a program of instructions executable by the computer to perform the method steps for automatically generating a myocardial perfusion map from a sequence of magnetic resonance (MR) images, said method comprising the steps of:

determining a region of interest (ROI) in a reference frame selected from a time series of myocardial perfusion MR image slices;

registering each image slice in said time series of slices to the reference frame to obtain a series of registered ROIs; and

using the series of registered ROIs to segment endo- and epi-cardial boundaries of a myocardium in the ROI.

16. The computer readable program storage device of claim 15, wherein determining a region of interest (ROI) in a reference frame comprises:

cropping slices in said series of slices according to a temporal intensity variation at each pixel to obtain a global mask containing pixels exhibiting significant temporal intensity variation;

selecting a range of candidate slices with superior contrast;

detecting the LV blood pool and forming an LV mask in all candidate slices;

selecting one candidate slice with a detected LV blood pool as the reference frame;

detecting the right ventricle (RV) in the reference frame and forming an RV mask;

estimating a thickness of the myocardium based on the detected LV and RV masks; and

defining the ROI as a bounding box containing the LV, RV, and myocardium.

17. The computer readable program storage device of claim 16, wherein detecting the LV blood pool comprises, for each candidate slice, selecting an intensity threshold for the blood, using said threshold to obtain a binary image and removing small connected components from said binary image.

18. The computer readable program storage device of claim 17, wherein selecting one candidate slice as the reference frame comprises computing shape features for each remaining connected component to select a least eccentric, most circular, and most convex connected component, wherein the selected connected component among all among all candidate slices corresponds to the LV blood pool.

19. The computer readable program storage device of claim 17, wherein said blood intensity threshold is selected through Gaussian clustering.

20. The computer readable program storage device of claim 18, wherein said shape features include an eccentricity computed as a ratio between minimum and maximum radii on the connected components, a convexity computed as a ratio between an area of the connected component and an area of its convex hull, and a circularity of the convex hull defined as 4π

times the area of the convex hull over a square of a perimeter of said convex hull.

21. The computer readable program storage device of claim 18, wherein when two connected components have similar shape features, a component with a larger area is selected.

22. The computer readable program storage device of claim 16, wherein detecting the RV comprises finding a connected component in the binary image that is close to the LV and has a large area.

23. The computer readable program storage device of claim 16, wherein segmenting endo- and epi-cardial boundaries of a myocardium in the ROI comprises:

initializing the endo-cardium and epi-cardium boundaries from the LV mask and myocardial thickness estimate;

detecting the endo-cardium by minimizing an energy functional;

masking out a left-ventricle (LV) blood pool inside of the endo-cardium from the image series, and detecting the epi-cardium by minimizing another energy functional,

wherein each said energy functional incorporates correlation information among pixels in a same slice as well as temporal correlations across the slices in the time series, prior shape information and anatomical constraints.

24. The computer readable program storage device of claim 23, wherein the energy functional for the endo-cardium is

E(C₁)=E_seq(C₁)+λ

E_int(C₁)+μ

E_int(C₁)+κ

E_shape(C₁),

wherein C₁ is the endo-cardial boundary, λ

, μ

, and κ and

K are parameters that control weights of different terms, wherein

25. The computer readable program storage device of claim 24, the method further comprising weighting each pixel on the endo-cardial boundary according to its distance to a convex hull of endo-cardial boundary, wherein pixels closer to the convex hull have a greater weight.

26. The computer readable program storage device of claim 23, wherein the energy functional for the epi-cardium is

E(C₂)=E_seq(C₂)+λ

E_img(C₂)+μ

E_int(C₂)+κ

E_shape(C₂)+E_con(C₂),

wherein C₂ is the epi-cardial boundary, λ

, μ

, and κ

are parameters that control weights of different terms, wherein