Method for reconstructing images of an imaged subject from a parallel MRI acquisition
View Patent ↗A parallel MR imaging method that uses a reconstruction algorithm that combines the GRAPPA image reconstruction method and the compressed sensing (CS) image reconstruction method in an iterative approach ( 200 ) or joint energy optimization approach ( 300 ).
1. A method for reconstructing images of an imaged subject from a parallel MRI acquisition, comprising:
a. generating, for each response signal received by a respective RF reception coil, a complete dataset in k-space from a reduced k-space dataset of the respective response signal and associated weighting factors for the reduced k-space dataset;
b. spatially transforming each complete k-space dataset generated to obtain an estimated reconstructed image for a respective RF reception coil, the estimated reconstructed image having a measure of sparsity;
c. maximizing the sparsity of the estimated reconstructed images;
d. repeating the generating and the spatially transforming steps using revised associated weighting factors obtained from each respective k-space dataset that results from the maximizing step; and
e. combining the estimated reconstructed images for the respective RF reception coils to obtain a reconstructed image of the imaged subject.
2. The method of claim 1 , wherein the generating step and the spatially transforming step are each performed using a generalized autocalibrating partially parallel acquisitions technique.
3. The method of claim 2 , wherein the maximizing step comprises maximizing using a compressed sensing technique.
4. The method of claim 1 , wherein the generating step comprises deriving the associated weighting factors from reference calibration data for a respective reduced k-space dataset.
5. The method of claim 1 , wherein the generating step comprises deriving the associated weighting factors from reference calibration data that is obtained from a respective reduced k-space dataset.
6. The method of claim 1 , wherein the generating step comprises deriving the associated weighting factors from reference calibration data that is obtained from a respective reduced k-space dataset and the repeating step comprises deriving the revised associated weighting factors from reference calibration data that is obtained from a respective complete k-space dataset that results from the maximizing step.
7. The method of claim 1 , wherein the maximizing step comprises maximizing using a compressed sensing technique.
8. The method of claim 7 , wherein the maximizing step comprises optimizing a sparsifying transform of an estimated reconstructed image over the set of values of the unacquired k-space data missing in the respective reduced k-space dataset.
9. The method of claim 1 , wherein the maximizing step comprises optimizing a sparsifying transform of an estimated reconstructed image over the set of values of the unacquired k-space data missing in the respective reduced k-space dataset.
10. The method of claim 9 , wherein the generating step comprises deriving the associated weighting factors from reference calibration data that is obtained from a respective reduced k-space dataset and the repeating step comprises deriving the revised associated weighting factors from reference calibration data that is obtained from a respective complete k-space dataset that results from the maximizing step.
11. The method of claim 1 , wherein the combining step comprises combining the estimated reconstructed images using a sum of squares reconstruction to obtain a reconstructed image of the imaged subject.
12. A method for obtaining an image of a subject by a parallel MR imaging device, comprising a) generating image reconstruction data from a generalized autocalibrating partially parallel acquisitions reconstruction of weighted under-sampled k-space data acquired by each RF reception coil of the device; b) calculating new weightings for respective under-sampled k-space data from k-space data obtained from a compressed sensing reconstruction of the image reconstruction data previously generated; and c) reconstructing the subject image according to the image reconstruction data generated from repeating the generating and calculating steps until convergence is achieved.
13. The method of claim 12 , wherein the generating step comprises deriving the weightings from calibration data that is obtained from respective under-sampled k-space data.
14. The method of claim 12 , wherein the compressed sensing reconstruction obtains k-space data by maximizing sparsity measures of a reconstructed image from the image reconstruction data generated.
15. The method of claim 12 , wherein the compressed sensing reconstruction obtains k-space data by optimizing a sparsifying transform of a reconstructed image over the set of values of the unacquired k-space data missing in the respective under-sampled k-space data.
16. The method of claim 12 , wherein the reconstructing step comprises reconstructing using a weighted combination of the individual image reconstructions from each RF reception coil of the device.
17. A method for obtaining an image of a subject using parallel MR imaging, comprising a) generating image reconstruction data from a generalized autocalibrating partially parallel acquisitions (GRAPPA) reconstruction of weighted under-sampled k-space data acquired by each RF reception coil; b) generating image reconstruction data from a compressed sensing (CS) reconstruction of under-sampled k-space data acquired by each RF reception coil; c) reconstructing the subject image according to the image reconstruction data generated and the formula
min
x
Ψ
x
0
+
λ
Ψ
x
-
C
D
W
y
_
2
,
where x represents the pixel intensities of the reconstructed images for the RF reception coils; Ψ represents a sparsifying transform; C represents a weighting matrix from the RF reception coils or the image estimates; D represents a block-diagonal matrix consisting of a discrete Fourier transform matrix; W represents a matrix operator that stores sets of weights which vary with the RF reception coils that may be applied to produce missing data in under-sampled k-space data; y represents the concatenated vector of the under-sampled k-space data from all the RF reception coils; and λ represents a parameter to control the degree to which the formula solution indicates a GRAPPA or a CS solution.