Method and apparatus for processing magnetic resonance image

1. A method for reconstructing a plurality of magnetic resonance (MR) images, the method comprising:

generating a plurality of MR images that have different contrasts with respect to a first part of an object;

using the plurality of MR images as an input to an artificial neural network (ANN) model in order to reconstruct an original MR image, and determining a correlation between each of the plurality of MR images and the reconstructed original MR image;

reconstructing an MR image based on the generated plurality of MR images and the determined correlation via the ANN model; and

displaying the reconstructed MR image,

wherein the original MR image is obtained by fully sampling multi-contrast k-space data with respect to the first part of the object, and

wherein the determining of the correlation comprises:

extracting a plurality of first patches that includes a set of a plurality of pixels from each of the generated plurality of MR images used as the input to the ANN model;

extracting a plurality of second patches from the original MR image used as an output to the ANN model;

determining a correspondence relation between the plurality of first patches and the plurality of second patches; and

determining the correlation based on the determined correspondence relation.

2. The method of claim 1 , wherein the generating the plurality of MR images comprises generating the plurality of MR images by subsampling an MR signal acquired from the first part of the object and using a parallel imaging method with respect to a result of the subsampling.

3. A non-transitory computer-readable recording medium having recorded thereon a program for executing the method of claim 1 on a computer.

4. The method of claim 1 , wherein the determining the correlation comprises applying a Multi-layer Perceptron (MLP) to each of the generated plurality of MR images and the original MR image.

5. The method of claim 1 , wherein the determining the correlation comprises using a backpropagation method.

6. The method of claim 1 , further comprising acquiring noise pattern information that relates to a geometric factor of a radio frequency (RF) coil by using a parallel imaging method,

wherein the determining the correlation comprises using the noise pattern information as an input to the ANN model.

7. The method of claim 1 , wherein the determining the correlation comprises classifying each of the generated plurality of MR images and the original MR image into a magnitude image and a phase image and determining a correlation with respect to the magnitude image and the phase image.

8. The method of claim 1 , further comprising generating an MR image that has a different contrast than the contrasts of the generated plurality of MR images based on the determined correlation between the generated plurality of MR images and the original MR image.

9. A magnetic resonance imaging (MRI) apparatus, the apparatus comprising:

at least one radio frequency (RF) channel coil configured to receive an MR signal emitted from a first part of an object;

an RF receiver configured to acquire the MR signal;

an image processor configured to:

generate a plurality of MR images that have different respective contrasts with respect to the first part of the object;

use the generated plurality of MR images as an input to an artificial neural network (ANN) model in order to reconstruct an original MR image;

determine a correlation between each of the generated plurality of MR images and the original MR image; and

reconstruct an MR image based on the determined correlation; and

a display configured to display the reconstructed MR image,

wherein the original MR image is obtained by fully sampling multi-contrast k-space data with respect to the first part of the object, and

wherein the image processor is further configured to:

extract a plurality of first patches that includes a set of a plurality of pixels from each of the generated plurality of MR images used as the input to the ANN model;

extract a plurality of second patches from the original MR image used as an output to the ANN model;

determine a correspondence relation among the plurality of first patches and the plurality of second patches; and

determine the correlation based on the determined correspondence relation.

10. The apparatus of claim 9 , wherein the image processor is further configured to generate an MR image that has a different contrast than the contrasts of the plurality of MR images based on the determined correlation between the generated plurality of MR images and the original MR image.

11. The apparatus of claim 9 , wherein the image processor comprises:

an MR signal interpretation module configured to determine the correlation and to reconstruct the MR image; and

a memory configured to store the generated plurality of MR images, the original MR image, and the reconstructed MR image.

12. The apparatus of claim 9 , wherein the image processor is further configured to generate the plurality of MR images by subsampling the MR signal received by the RF receiver and by using a parallel imaging method with respect to a result of the subsampling.

13. The apparatus of claim 9 , wherein the image processor is further configured to determine the correlation between the generated plurality of MR images and the original MR image by using a backpropagation method.

14. The apparatus of claim 9 , wherein the image processor is further configured to determine the correlation between the generated plurality of MR images and the original MR image by applying a Multi-layer Perceptron (MLP) to each of the generated plurality of MR images and the original MR image.

15. The apparatus of claim 9 , wherein the image processor is further configured to acquire noise pattern information that relates to a geometric factor of the at least one RF channel coil by using a parallel imaging method.

16. The apparatus of claim 15 , wherein the image processor is further configured to determine the correlation between the generated plurality of MR images and the original MR image by using the noise pattern information as an input to the ANN model.