Method of reconstructing a dynamic series of motion-compensated magnetic resonance images
A Computer-implemented method of reconstructing a dynamic series of motion-compensated magnetic resonance images of a patient is provided. Images of a patient are acquired over time, at least partially in free-breathing, at a first image resolution and on a frame-by-frame basis. Each frame of the k-space data includes a first subset of data points having a first sample density and a second subset of data points having a second sample density. For each frame, a sub-group of the first subset and the second subset of the k-space data is selected, and an image is reconstructed at a second image resolution. The motion between the second image resolution images is estimated in the form of motion fields. The motion information is incorporated into a final reconstruction of a dynamic series of motion-compensated magnetic resonance images of the patient at a third image resolution.
1. A computer-implemented method of reconstructing a dynamic series of motion-compensated magnetic resonance images of a patient, the computer-implemented method comprising:
acquiring, over time, at a first image resolution and on a frame-by-frame basis, k-space data of a dynamic series of magnetic resonance images of the patient over successive respiratory cycles, cardiac cycles, or respiratory cycles and cardiac cycles, wherein each frame of the k-space data comprises a first subset of data points having a first sample density and a second subset of data points having a second sample density;
selecting, for each frame, a respective sub-group of the first subset of data points and the second subset of data points of the k-space data and reconstructing, for each frame, an image, at a second image resolution, from the selected sub-groups;
estimating a motion between the second image resolution images in a form of motion fields; and
incorporating the motion fields into a final reconstruction of the dynamic series of motion-compensated magnetic resonance images of the patient at a third image resolution,
wherein the first subset of data points is a region undersampled coherently across the dynamic series of magnetic resonance images in a dimension of k-space, and the second subset of data points is a region undersampled incoherently across the dynamic series of magnetic resonance images in the dimension of k-space.
2. The method of claim 1 , wherein the first subset of data points has a greater sample density than the second subset of data points.
3. The method of claim 1 , wherein the coherently undersampled region is a linearly undersampled region.
4. The method of claim 1 , wherein the coherently undersampled region is covering a center of k-space.
5. The method of claim 4 , wherein the coherently undersampled region covers the center of k-space across the dynamic series of magnetic resonance images.
6. The method of claim 4 , wherein the coherently undersampled region is obtained using a time-interleaved undersampling scheme, and the second subset of data points is obtained using a random or pseudo-random undersampling scheme.
7. The method of claim 4 , wherein a size of the coherently undersampled region covering the center of the k-space is determined by a minimum resolution at which physiological features or motion of interest are resolvable in a reconstructed image.
8. The method of claim 1 , wherein reconstructing the images at the second image resolution comprises spatial regularization of individual image frames.
9. The method of claim 8 , wherein reconstructing the images at the second image resolution is carried out without a temporal regularization constraint.
10. The method of claim 1 , wherein the final reconstruction of the dynamic series of motion-compensated magnetic resonance images of the patient at the third image resolution comprises a temporal regularization constraint across the dynamic series of magnetic resonance images.
11. The method of claim 1 , wherein the final reconstruction of the dynamic series of motion-compensated magnetic resonance images of the patient at the third image resolution comprises a spatial regularization of the individual image frames and a temporal regularization constraint across the dynamic series of magnetic resonance images.
12. The method of claim 10 , wherein the temporal regularization constraint in the reconstructing uses a total-variation regularization.
13. The method of claim 1 , further comprising:
registering the reconstructed images on a frame-by-frame basis to determine motion fields; and
interpolating the motion fields to a desired resolution for the final reconstruction of the dynamic series of motion-compensated magnetic resonance images.
14. The method of claim 1 , wherein the third image resolution is the same as the first image resolution.
15. The method of claim 1 , wherein the second image resolution is lower than the first image resolution.
16. The method of claim 15 , wherein the second image resolution is chosen as a minimum resolution at which physiological features or a motion of interest is resolvable in a reconstructed image.
17. The method of claim 1 , wherein the patient is free-breathing for at least a portion of the time over which the k-space data is obtained.
18. A data processing apparatus adapted to reconstruct a dynamic series of motion-compensated magnetic resonance images of a patient, the data processing apparatus comprising:
an imaging device configured to acquire, over time, at a first image resolution and on a frame-by-frame basis, k-space data of a dynamic series of magnetic resonance images of a patient over successive respiratory cycles, cardiac cycles, or respiratory cycles and cardiac cycles, wherein each frame of the k-space data comprises a first subset of data points having a first sample density and a second subset of data points having a second sample density; and
a processor configured to:
select, for each frame, a sub-group of the first subset of data points and the second subset of data points of the k-space data;
reconstruct, for each frame, an image, at a second image resolution, from the selected sub-groups; and
estimate a motion between the second image resolution images in a form of motion fields and incorporate the motion fields into a final reconstruction of the dynamic series of motion-compensated magnetic resonance images of the patient at a third image resolution,
wherein the first subset of data points is a region undersampled coherently across the dynamic series of magnetic resonance images in a dimension of k-space, and the second subset of data points is a region undersampled incoherently across the dynamic series of magnetic resonance images in the dimension of k-space.