Apparatus and method for reconstructing magnetic resonance image using learning, and under-sampling apparatus method and recording medium thereof

1. A magnetic resonance image reconstruction apparatus using learning, comprising:

an under-sampling portion that performs full-sampling with respect to a first region that corresponds to a center of a first k-space image and performs under-sampling with respect to a second region in the first k-space image, excluding the first k-space image, to under-sample the first k-space image; and

a magnetic resonance image reconstruction portion that outputs a magnetic resonance image based on the first k-space image through a pre-learned neural network, generates a third k-space image by correcting a second k-space image that is based on the output magnetic resonance image, based on the first k-space image with respect to a region that correspondings to the first k-space image in the second k-space, and acquires a magnetic resonance image based on the third k-space image.

2. The MR image reconstruction apparatus using machine learning of claim 1 , wherein the MR image reconstruction portion comprises:

a first image converter that acquires a first MR image by converting the first k-space image;

a first image reconstruction portion that acquires a second MR image through the learned neural network by receiving the first MR image;

a second image converter that acquires the second k-space image by converting the second MR image;

a second image reconstruction portion that generates the third k-space image through correction that replaces a portion overlapping the first k-space image with the first k-space image in the second k-space image; and

a third image converter that acquires a reconstructed MR image by converting the third k-space image.

3. The MR image reconstruction apparatus using machine learning of claim 2 , wherein the first image converter and the third image converter use inverse Fourier transform, and the second image converter uses Fourier transform.

4. The MR image reconstruction apparatus of claim 1 , wherein the first image reconstruction portion is pre-learned by using a reference MR image converted from a k-space image that corresponds to an MR image and in which full-sampling is performed on a center region of the k-space image and under-sampling is formed on a remaining region of the k-space image as an input value, and an MR image converted from a reference full-sample k-space image as a label.

5. The MR image reconstruction apparatus using machine learning of claim 1 , wherein the first region is set to a region corresponding to 3% to 8% of a center of the first k-space image.

6. The MR image reconstruction apparatus using machine learning of claim 1 , wherein the under-sampling performed on the second region performs encoding by selecting one per at least four lines.

7. The MR image reconstruction apparatus using machine learning of claim 1 , further comprising a region setting portion that sets a region that corresponds to a center of the first k-space image as the first region, and a remaining region as the second region.

8. An MR image reconstruction apparatus using learning, comprising:

a decomposition portion that generates n×k1 low-resolution input images by applying k1 first filters n times to a high-resolution input image;

a learning portion that generates n×k2 low-resolution output images by applying a corresponding neural network according to the n×k1 low-resolution input image; and

a reconstruction portion that generates a high-resolution output image by applying k2 second dual filters n times to the n×k2 low-resolution output image,

wherein the high-resolution input image is generated by converting a k-space image where full-sampling is performed on a first region that corresponds to a center of the k-space image and under-sampling is performed on a second region in the k-space image, excluding the first region, and

the k2 second dual filters are dual filters with respect to k2 second filters that generate a reference high-resolution label image for learning of the neural network as n×k2 low-resolution label images.

9. The MR image reconstruction apparatus using machine learning of claim 8 , wherein the learning portion comprises a neural network is learned so that a result acquired by performing deep learning on n×kl reference low-resolution input images that are generated by iteratively performing the k1 filters n times on a reference high-resolution input image, becomes n×k2 reference low-resolution label images, and

the n×k2 reference low-resolution label images are generated by repeating the k2 filters n times on a high-resolution label image.

10. The MR image reconstruction apparatus using machine learning of claim 8 , wherein the learning portion uses different neural networks for a first low-resolution input image among the k1 low-resolution input images and for a second low-resolution input image among the k1 low-resolution input images.

11. A MR image reconstruction method using learning, comprising:

under-sampling a first k-space image by full-sampling a first region that corresponds to a center of the first k-space image and under-sampling a second region, excluding the first region, in the first k-space image;

outputting an MR image based on the first k-space image through a pre-learned neural network;

generating a third k-space image by correcting a second k-space image that is based on the output MR image, based on the first k-space image with respect to a region corresponding to the first k-space image in the second k-space image; and

acquiring an MR image based on the third k-space image.

12. The MR image reconstruction method using machine learning of claim 11 , wherein the outputting the MR image based on the first k-space image through the pre-learned neural network comprises:

acquiring a first MR image by converting the first k-space image; and

acquiring a second MR image through the learned neural network by receiving the first MR image.

13. The MR image reconstruction method using machine learning of claim 12 , wherein the acquiring the first MR image comprises acquiring the first MR image by inverse Fourier transforming the first k-space image.

14. The MR image reconstruction method using machine learning of claim 11 , wherein the generating the third k-space image comprises:

acquiring the second k-space image by converting the output MR image; and

generating the third k-space image through correction that replaces a portion overlapping the first k-space image with the first k-space image in the second k-space image.

15. The MR image reconstruction method using machine learning of claim 14 , wherein the acquiring the second k-space image comprises acquiring the second k-space image by Fourier transforming the output MR image.

16. The MR image reconstruction method using machine learning of claim 11 , wherein the pre-learned neural network is pre-learned by using a reference MR image converted from a k-space image that corresponds to an MR image and in which full-sampling is performed on a center region of the k-space image and under-sampling is formed on a remaining region of the k-space image as an input value, and an MR image converted from a reference full-sample k-space image as a label.

17. The MR image reconstruction method using machine learning of claim 11 , further comprising setting a region corresponding to a center of the first k-space image as the first region and a remaining region as the second region.

18. An MR image reconstruction method using learning, comprising:

generating a high-resolution input image by converting a k-space image where full-sampling is performed on a first region that corresponds to a center of the k-space image and under-sampling is performed on a second region in the k-space image, excluding the first region;

decomposing for generating n×k1 low-resolution input images by applying k1 first filters on the high-resolution input image;

generating n×k2 low-resolution output images by applying a corresponding neural network according to the n×k1 low-resolution input images; and

generating a high-resolution output image by applying nk2 second dual filters to the n×k2 low-resolution output images,

wherein the k2 second dual filters are dual filters with respect to k2 second filters that generate a reference high-resolution label image for learning of the neural network as n×k2 low-resolution label images.

19. The MR image reconstruction method using mahine learning of claim 18 , wherein the neural network is learned so that a result acquired by performing deep learning on n×k1 reference low-resolution input images that are generated by iteratively performing the k1 filters n times on a reference high-resolution input image, becomes n×k2 reference low-resolution label images, and

the n×k2 reference low-resolution label images are generated by repeating the k2 filters n times on a high-resolution label image.

20. The MR image reconstruction method using machine learning of claim 18 , wherein the generating the n×k2 low-resolution output images comprises:

applying a first neural network that corresponds to a first low-resolution input image among the n×k1 low-resolution input images; and

applying a second neural network that corresponds to a second low-resolution input image among the n×k1 low-resolution input images.

21. A computer readable recording medium on which a program that executes an MR image reconstruction method of claim 11 is recorded.