Magnetic resonance image reconstruction system and method

Patent metadata

Field	Value
Publication number	US-10133964-B2
Application number	US-201715471079-A
Country	US
Kind code	B2
Filing date	Mar 28, 2017
Priority date	Mar 28, 2017
Publication date	Nov 20, 2018
Grant date	Nov 20, 2018

Abstract

A method for training a system for reconstructing a magnetic resonance image includes: under-sampling image data from each of a plurality of fully-sampled images; and inputting the under-sampled image data to a multi-scale neural network comprising sequentially connected layers. Each layer has an input for receiving input image data and an output for outputting reconstructed image data. Each layer performs a process comprising: decomposing the array of input image data; applying a thresholding function to the decomposed image data, to form a shrunk data, the thresholding function outputting a value asymptotically approaching one when the thresholding function receives an input having a magnitude greater than a first value, reconstructing the shrunk data for combining with a reconstructed image data output by another one of the layers to form updated reconstructed image data, and machine-learning at least one parameter of the decomposing, the thresholding function, or the reconstructing.

First claim

What is claimed is:

1. A method for training a system for reconstructing a magnetic resonance (MR) image from signals collected by an MR scanner, the method comprising:
under-sampling respective image data from each of a plurality of fully-sampled images;
inputting the under-sampled image data to a multi-scale neural network comprising a plurality of sequentially connected layers, each layer having a respective input for receiving a respective array of input image data and an output for outputting reconstructed image data, each layer performing a recursive process comprising:
decomposing the array of input image data received at the input of the layer;
applying a thresholding function to the decomposed image data, to form shrunk data, the thresholding function outputting a non-zero value asymptotically approaching one when the thresholding function receives an input having a magnitude greater than a first value,
reconstructing the shrunk data for combining with a reconstructed image data output by another one of the plurality of layers to form updated reconstructed image data, and
machine-learning at least one parameter of the decomposing, the thresholding function, or the reconstructing based on the fully sampled images and the updated reconstructed image data.

2. The method of claim 1 , wherein the thresholding function outputs a zero value in the case where the thresholding function receives an input having a magnitude less than the first value.

3. The method of claim 2 , wherein the thresholding function is non-linear.

4. The method of claim 3 , wherein the thresholding function is a garrote function.

5. The method of claim 2 , wherein each layer of the multi-scale neural network is configured to update the thresholding function.

6. The method of claim 1 , wherein the machine-learning includes machine-learning parameters of the decomposing, the thresholding function, and the reconstructing based on the fully sampled images and the reconstructed image data.

7. The method of claim 1 , wherein:
the decomposing includes performing a convolution of the image data and a first filter, and
the reconstructing includes performing deconvolution of the shrunk data and the filter.

8. The method of claim 1 , wherein the thresholding function of each of the plurality of layers is a shrinkage function.

9. The method of claim 1 , wherein each layer is configured for adding a gradient update to the reconstructed image data, and the gradient update is based on a gradient step size recursively computed based on the image data.

10. The method of claim 1 , wherein the under-sampling is performed using an under-sampling operator, and the method for training further comprises machine-learning at least one parameter of the under-sampling operator based on the fully sampled images and the updated reconstructed image data.

11. The method of claim 1 , wherein the computing includes generating a sampling mask based on the fully sampled images and the updated reconstructed image data.

12. A system for acquiring and reconstructing a magnetic resonance (MR) image from signals collected by an MR scanner, the system comprising:
a non-transitory machine readable storage medium for storing under-sampled image data from a plurality of fully-sampled images;
a multi-scale neural network for receiving the under-sampled image data, the multi-scale neural network comprising a plurality of sequentially connected layers, each layer having a respective input for receiving a respective array of input image data and outputting reconstructed image data, each layer configured for performing a process comprising:
decomposing the array of input image data received at the input of the layer,
applying a thresholding function to the decomposed image data, to form shrunk data, the thresholding function outputting a non-zero value asymptotically approaching one when the thresholding function receives an input having a magnitude greater than a first value,
reconstructing the shrunk data for combining with a reconstructed image data output by another one of the plurality of layers to form updated reconstructed image data, and
machine-learning at least one parameter of the decomposing, the thresholding function, or the reconstructing based on the fully sampled images and the updated reconstructed image data.

13. The system of claim 12 , further comprising:
a pseudo-random generator to generate a sampling mask according to a set of probabilities learned in the multi-scale neural network,
wherein the MR scanner acquires the under-sampled image data from the subject using the sampling mask.

14. The system of claim 12 , wherein the thresholding function outputs a zero value in the case where the thresholding function receives an input having a magnitude less than the first value.

15. The system of claim 14 , wherein the thresholding function is non-linear.

16. The system of claim 15 , wherein the thresholding function is a garrote function.

17. The system of claim 14 , wherein each layer of the multi-scale neural network is configured to update the thresholding function.

18. The system of claim 12 , wherein the machine-learning includes machine-learning parameters of the decomposing, the thresholding function, and the reconstructing based on the fully sampled images and the reconstructed image data.

19. The system of claim 12 , wherein:
the decomposing includes performing a convolution of the image data and a first filter, and
the reconstructing includes performing deconvolution of the shrunk data and the filter.

20. A non-transitory, machine readable storage medium encoded with computer program software, such that when a processor executes the computer program software, the processor performs a method for training a system for reconstructing a magnetic resonance (MR) image from signals collected by an MR scanner, the method comprising:
under-sampling respective image data from each of a plurality of fully-sampled images;
inputting the under-sampled image data to a multi-scale neural network comprising a plurality of sequentially connected layers, each layer having a respective input for receiving a respective array of input image data and an output for outputting reconstructed image data, each layer performing a recursive process comprising:
decomposing the array of input image data received at the input of the layer;
applying a thresholding function to the decomposed image data, to form shrunk data, the thresholding function outputting a non-zero value asymptotically approaching one when the thresholding function receives an input having a magnitude greater than a first value,
reconstructing the shrunk data for combining with a reconstructed image data output by another one of the plurality of layers to form updated reconstructed image data, and
machine-learning at least one parameter of the decomposing, the thresholding function, or the reconstructing based on the fully sampled images and the updated reconstructed image data.

Assignees

• Siemens Healthcare Gmbh

Inventors

• Guo Yi
• Mailhe Boris
• Chen Xiao
• Nadar Mariappan S

Classifications

• G01R33/5608 Primary

  Data processing and visualization specially adapted for MR, e.g. for feature analysis and pattern recognition on the basis of measured MR data, segmentation of measured MR data, edge contour detection on the basis of measured MR data, for enhancing measured MR data in terms of signal-to-noise ratio by means of noise filtering or apodization, for enhancing measured MR data in terms of resolution by means for deblurring, windowing, zero filling, or generation of gray-scaled images, colour-coded images or images displaying vectors instead of pixels (image data processing or generation, in general G06T) · CPC title

• G06T12/30

  Image post-processing, e.g. metal artefact correction · CPC title

• A61B5/7267

  involving training the classification device · CPC title

• G06T7/0012

  Biomedical image inspection · CPC title

• G06T2207/10088

  Magnetic resonance imaging [MRI] · CPC title