Method for varying undersampling dimension for accelerating multiple-acquisition magnetic resonance imaging and device for the same

Patent metadata

Field	Value
Publication number	US-11029381-B2
Application number	US-201916246598-A
Country	US
Kind code	B2
Filing date	Jan 14, 2019
Priority date	Jan 12, 2018
Publication date	Jun 8, 2021
Grant date	Jun 8, 2021

Abstract

Provided is an MRI image generation method including: acquiring first phase encoding lines obtained by undersampling along a first direction using an MRI device; acquiring second phase encoding lines obtained by undersampling in a second direction different from the first direction using the MRI device; generating a first MRI image based on the first phase encoding lines and the second phase encoding lines; and generating a second MRI image different from the first MRI image based on the first phase encoding lines and the second phase encoding lines.

First claim

What is claimed is:

1. An MRI image generation method comprising:
acquiring, with an MRI device comprising a first coil and a computing device, first phase encoding lines obtained by a first undersampling scheme along a first direction;
acquiring, with the MRI device, second phase encoding lines obtained by a second undersampling scheme in a second direction different from the first direction;
generating, with the computing device, a first MRI image based on the first phase encoding lines and the second phase encoding lines; and
generating, with the computing device, a second MRI image different from the first MRI image based on the first phase encoding lines and the second phase encoding lines,
wherein,
the first undersampling scheme is a first data acquisition scheme that does not acquire a portion of the total K-space data that must be acquired to achieve a predetermined FOV and resolution of the first MRI image, and
the second undersampling scheme is a second data acquisition scheme that does not acquire a portion of the total K-space data that must be acquired to achieve the predetermined FOV and resolution of the first MRI image.

2. The method of claim 1 , wherein,
the generating of the first MRI image comprises: determining a first k-space based on the first phase encoding lines and the second phase encoding lines, and generating the first MRI image from the first k-space, and
the generating of the second MRI image comprises: determining a second k-space different from the first k-space based on the first phase encoding lines and the second phase encoding lines, and generating the second MRI image from the second k-space.

3. The method of claim of 2 , wherein,
the determining of the first k-space comprises filling the first phase encoding lines into a k-space required to obtain a predetermined FOV and resolution, and filling a portion or all of a remaining portion of the k-space using the second phase encoding lines, and
the determining of the second k-space comprises filling the first phase encoding lines into the k-space and filling a portion or all of the remaining portion of the k-space using the second phase encoding lines.

4. The method of claim 1 , wherein the first phase encoding lines are acquired using any one sampling pattern among a central sampling pattern and a random sampling pattern,
wherein the second phase encoding lines are acquired using any one sampling pattern among a central sampling pattern and a random sampling pattern.

5. The method of claim 1 , wherein the generating of the first MRI image comprises providing first data on the first phase encoding lines and second data on the second phase encoding lines to an input layer of a trained deep learning network to acquire the first MRI image from an output layer of the deep learning network.

6. The method of claim 5 , wherein a method for training the deep learning network comprises:
acquiring, with the MRI device, third data on third phase encoding lines obtained by undersampling along the first direction, with respect to a first image acquisition layer using the MRI device;
acquiring, with the MRI device, fourth data on fourth phase encoding lines obtained by undersampling along the second direction, with respect to the first image acquisition layer using the MRI device;
acquiring, with the MRI device, fifth phase encoding lines consisting of phase encoding lines obtained by full-sampling, with respect to the first image acquisition layer using the MRI device;
generating, with the computing device, an output layer MRI image for training by applying a Fourier transform to the fifth phase encoding lines; and
providing, with the computing device, the third data and the fourth data to the input layer of the deep learning network and providing the output layer MRI image for training to the output layer of the deep learning network to train the deep learning network.

7. The method of claim 1 , wherein the generating of the first MRI image comprises:
generating a third k-space based on the first phase encoding lines and the second phase encoding lines;
generating an input layer-MRI image by performing an FFT on the third k-space; and
acquiring the first MRI image from the output layer of the deep learning network by providing the input layer-MRI image to the input layer of the trained deep learning network.

8. A method for training a deep learning network comprising:
acquiring, with an MRI device comprising a first coil and a computing device, third phase encoding lines obtained by undersampling along a first direction, with respect to a first image acquisition layer;
acquiring, with the MRI device, fourth phase encoding lines obtained by undersampling along a second direction different from the first direction, with respect to the first image acquisition layer;
generating with the computing device, a fourth k-space based on the third phase encoding lines and the fourth phase encoding lines;
generating, with the computing device, an input layer-MRI image for training by performing an FFT on the fourth k-space;
acquiring, with the MRI device, fifth phase encoding lines obtained by full-sampling, with respect to the first image acquisition layer;
generating, with the computing device, an output layer-MRI image for training using the fifth phase encoding lines; and
providing, with the computing device, the input layer-MRI image for training to an input layer of the deep learning network and providing, with the computing device, the output layer-MRI image for training to an output layer of the deep learning network to train the deep learning network.

9. An MRI device comprising:
an MRI scanner comprising a first coil; and
an MRI computing device comprising a processing unit and
a storage unit,
wherein the processing unit is configured to perform:
acquiring, with a MRI device, first phase encoding lines obtained by a first undersampling scheme along a first direction;
acquiring, with the MRI device, second phase encoding lines obtained by a second undersampling scheme in a second direction different from the first direction;
generating a first MRI image based on the first phase encoding lines and the second phase encoding lines; and
generating a second MRI image different from the first MRI image based on the first phase encoding lines and the second phase encoding lines,
wherein,
the first undersampling scheme is a first data acquisition scheme that does not acquire a portion of the total K-space data that must be acquired to achieve a predetermined FOV and resolution of the first MRI image, and
the second undersampling scheme is a second data acquisition scheme that does not acquire a portion of the total K-space data that must be acquired to achieve the predetermined FOV and resolution of the first MRI image.

Assignees

• Korea Advanced Inst Sci & Tech

Inventors

• Park Sung-Hong
• Kim Ki-Hwan

Classifications

• G06N3/045

  Combinations of networks · CPC title

• G06N3/09

  Supervised learning · CPC title

• G06N3/0464

  Convolutional networks [CNN, ConvNet] · CPC title

• G06N3/08 Primary

  Learning methods · CPC title

• G01R33/5619

  by temporal sharing of data, e.g. keyhole, block regional interpolation scheme for k-Space [BRISK] · CPC title