Method and device for generating fat suppression magnetic resonance image using generative adversarial neural network based on the bloch equation

Patent metadata

Field	Value
Publication number	US-2022326328-A1
Application number	US-202217719931-A
Country	US
Kind code	A1
Filing date	Apr 13, 2022
Priority date	Apr 13, 2021
Publication date	Oct 13, 2022
Grant date	—

Abstract

The disclosed technology relates to a method and device for generating a fat suppression magnetic resonance image. The method includes: inputting, by an imaging device, a magnetic resonance image to an encoder of a neural network to extract features of the magnetic resonance image; and generating, by a generator of the neural network, a T2-weighted fat suppression image based on the features, in which the neural network is trained according to a result of discriminating, by a discriminator of the neural network, a loss due to a generation of a T2-weighted fat suppression image and as a result of reconstructing, by a decoder, the magnetic resonance image input to the encoder using a Bloch equation before the magnetic resonance image is input.

First claim

1 . A method of generating a fat suppression magnetic resonance image, comprising:
inputting, by an imaging device, two magnetic resonance images having different contrasts to an encoder of a neural network to extract features of the magnetic resonance images; and generating, by a generator of the neural network, a magnetic resonance image having a contrast different from the two magnetic resonance images based on the features, wherein the neural network is trained according to a result of discriminating, by a discriminator of the neural network, a loss due to a generation of a T2-weighted fat suppression image and as a result of reconstructing, by a decoder, the magnetic resonance image input to the encoder using a Bloch equation before the magnetic resonance image is input.

2 . The method of claim 1 , wherein the two magnetic resonance images input to the encoder are a T1-weighted image and a T2-weighted image obtained by capturing a specific tissue in a body at the same time, and
the magnetic resonance image generated by the generator is a T2-weighted fat suppression magnetic resonance image.

3 . The method of claim 1 , wherein the neural network is a generative adversarial neural network based on the Bloch equation (Bloch-GAN).

4 . The method of claim 1 , wherein the generative adversarial neural network includes an adversarial loss function, a normalized loss function, a pixelwise loss function, and a perceptual loss function as an overall loss function, and is trained so that the overall loss function is minimized.

5 . The method of claim 1 , wherein the decoder generates a hypothetical magnetic resonance parameter map and reconstructs the magnetic resonance image according to the magnetic resonance parameter map.

6 . The method of claim 5 , wherein the magnetic resonance parameter map includes a T1 magnetic relaxation rate map, a T2 magnetic relaxation rate map, and a proton density map.

7 . A device for generating a fat suppression magnetic resonance image, comprising:
an input device configured to receive two magnetic resonance images having different contrasts; a storage device configured to store a generative adversarial neural network based on a Bloch equation including an encoder, a decoder, a generator, and a discriminator as sub-networks; and a computing device configured to extract features of the two magnetic resonance images using the encoder and generate a magnetic resonance image having a contrast different from the two magnetic resonance images based on the extracted features using the generator, wherein the generative adversarial neural network based on the Bloch equation is trained according to a result of discriminating, by the discriminator, a loss due to a generation of the magnetic resonance images and as a result of reconstructing, by the decoder, the magnetic resonance image input to the encoder using the Bloch equation before the two magnetic resonance images are input.

8 . The device of claim 7 , wherein the two magnetic resonance images are a T1-weighted image and a T2-weighted image obtained by capturing a specific tissue in a body at the same time, and
the magnetic resonance image generated by the generator is a T2-weighted fat suppression magnetic resonance image.

9 . The device of claim 7 , wherein the generative adversarial neural network includes an adversarial loss function, a normalized loss function, a pixelwise loss function, and a perceptual loss function as an overall loss function, and is trained to minimize the overall loss function.

10 . The device of claim 7 , wherein the generative adversarial neural network generates a hypothetical magnetic resonance parameter map and reconstructs the magnetic resonance image according to the magnetic resonance parameter map.

11 . The device of claim 10 , wherein the magnetic resonance parameter map includes a T1 magnetic relaxation rate map, a T2 magnetic relaxation rate map, and a proton density map.

Assignees

• Uif Univ Industry Foundation Yonsei Univ

Inventors

• Hwang Do Sik
• Kim Sungjun
• Lee Young Han
• Kim Se Won

Classifications

• G06T12/00 Primary

  Tomographic reconstruction from projections · CPC title

• A61B5/055 Primary

  involving electronic [EMR] or nuclear [NMR] magnetic resonance, e.g. magnetic resonance imaging · CPC title

• G16H30/40

  for processing medical images, e.g. editing · CPC title

• A61B5/0033

  Features or image-related aspects of imaging apparatus, e.g. for MRI, optical tomography or impedance tomography apparatus; Arrangements of imaging apparatus in a room · CPC title

• G06T2207/10088

  Magnetic resonance imaging [MRI] · CPC title