Method for calibration-free locally low-rank encouraging reconstruction of magnetic resonance images

Patent metadata

Field	Value
Publication number	US-9709650-B2
Application number	US-201214356287-A
Country	US
Kind code	B2
Filing date	Nov 6, 2012
Priority date	Nov 6, 2011
Publication date	Jul 18, 2017
Grant date	Jul 18, 2017

Abstract

A method for reconstructing an image of a subject from multi-coil data acquired with an array of radio frequency (“RF”) receiver coils that form a part of a magnetic resonance imaging (“MRI”) system without the need for coil sensitivity calibration is provided. A coil data set is acquired for each of the RF receiver coils, for example, by undersampling k-space. A set of image blocks that defines the location and size of each of a plurality of image blocks in the image domain is then selected. The acquired coil data sets and selected image block set are then used to jointly estimate a coil image for each of the plurality of RF receiver coils while promoting locally-low rank structure in the coil images. These coil images are then combined to produce the target image of the subject.

First claim

The invention claimed is:

1. A method for reconstructing an image from data acquired with a plurality of radio frequency (“RF”) receiver coils that form a part of a magnetic resonance imaging (“MRI”) system, the steps of the method comprising:
a) acquiring a coil data set from a subject for each of a plurality of RF receiver coils that form a part of an MRI system;
b) selecting a set of image blocks that defines a location and size of each of a plurality of image blocks in an image domain;
c) jointly estimating a coil image for each of the plurality of RF receiver coils using the coil data sets acquired in step a) and the set of image blocks selected in step b); and
d) combining the coil images jointly estimated in step c) to produce the image of the subject.

2. The method as recited in claim 1 in which the joint estimation performed in step c) includes minimizing an objective function while promoting locally-low rank structure in the jointly estimated coil images using the set of image blocks selected in step b).

3. The method as recited in claim 2 in which step c) includes minimizing the objective function using a greedy optimization algorithm that is based on singular value hard thresholding.

4. The method as recited in claim 2 in which the set of image blocks is spatially translated by at least one of a random and a deterministic amount at each iteration of the minimization of the objective function.

5. The method as recited in claim 2 in which the objective function includes a non-convex penalty term that promotes locally low-rank structure in the jointly estimated coil images.

6. The method as recited in claim 5 in which the non-convex penalty term includes at least one of a log-determinant and a Schatten p-norm functional.

7. The method as recited in claim 2 in which the objective function includes a regularization parameter that is automatically selected using a statistical method.

8. The method as recited in claim 2 in which the objective function includes a term that promotes locally low-rank structure in the jointly estimated coil images using the set of image blocks selected in step b).

9. The method as recited in claim 8 in which the term of the objective function includes calculating a nuclear norm of each image block defined by the set of images blocks selected in step b) and summing the calculated nuclear norms.

10. The method as recited in claim 2 in which locally-low rank structure in the jointly estimated coil images is promoted by imposing a penalty factor that penalizes non-low-rank image blocks in the set of image blocks selected in step b).

11. The method as recited in claim 10 in which the penalty factor includes calculating a singular value soft threshold for each image block in the set of image blocks selected in step b).

12. The method as recited in claim 2 in which the objective function has at least one additional term that promotes desired spatial characteristics of the jointly estimated coil images.

13. The method as recited in claim 2 in which the objective function has at least one additional term that promotes sparsity in the jointly estimated coil images.

14. The method as recited in claim 13 in which the at least one additional term includes a sparsifying transform.

15. The method as recited in claim 14 in which the at least one additional term of the objective function includes producing sparsified image blocks by applying the sparsifying transform to each image block defined by the set of images blocks selected in step b), calculating an λ p -λ q norm of each sparsified image block, and summing the calculated λ p -λ q norms.

16. The method as recited in claim 2 in which the joint estimation performed in step c) includes minimizing an objective function while also promoting sparsity of the jointly estimated coil images using the set of image blocks selected in step b).

17. The method as recited in claim 16 in which sparsity of the jointly estimated coil images is promoted in step c) by imposing a penalty factor that penalizes non-sparse image blocks in the set of image blocks selected in step b).

18. The method as recited in claim 17 in which the penalty factor includes producing sparsified image blocks by applying the sparsifying transform to each image block defined by the set of images blocks selected in step b), and calculating a joint-soft threshold for each sparsified image block.

19. The method as recited in claim 18 in which the sparsifying transform is at least one of a discrete cosine transform and a wavelet transform.

20. The method as recited in claim 1 in which each coil data set acquired in step a) is acquired by undersampling k-space.

21. The method as recited in claim 20 in which step a) includes undersampling k-space using an incoherent sampling operation.

22. The method as recited in claim 20 in which step a) includes undersampling k-space using non-Cartesian trajectories.

23. The method as recited in claim 1 in which the image produced in step d) is a three-dimensional image, and the set of image blocks selected in step b) is a set of three-dimensional image blocks.

24. The method as recited in claim 1 in which the joint estimation performed in step c) includes using a forward signal model that accounts for at least one of off-resonance, concomitant gradients, chemical shifts, and presence of multiple chemical species.

25. The method as recited in claim 1 in which the joint estimation performed in step c) includes modeling complex Gaussian noise.

26. The method as recited in claim 25 in which the modeled complex Gaussian noise is modeled as at least one of correlated and uncorrelated across the plurality of RF receiver coils.

27. The method as recited in claim 1 in which:
the coil data set acquired in step a) is indicative of at least one of a time series of images and a parametric series of images;
the set of image blocks selected in step b) are obtained from different frames of the at least one of a time series of images and a parametric series of images; and
the joint estimation performed in step c) includes promoting low-rankedness across the set of image blocks selected in step b).

28. The method as recited in claim 1 in which the joint estimation performed in step c) includes minimizing an objective function that does not require coil sensitivity calibration data as an input.

Assignees

• Mayo Foundation

Inventors

• Trzasko Joshua D
• Manduca Armando

Classifications

• G01R33/5611 Primary

  Parallel magnetic resonance imaging, e.g. sensitivity encoding [SENSE], simultaneous acquisition of spatial harmonics [SMASH], unaliasing by Fourier encoding of the overlaps using the temporal dimension [UNFOLD], k-t-broad-use linear acquisition speed-up technique [k-t-BLAST], k-t-SENSE (structural details of arrays of sub-coils G01R33/3415) · CPC title