Simultaneous multi-slice protocol and deep-learning magnetic resonance imaging reconstruction

Patent metadata

Field	Value
Publication number	US-12475614-B2
Application number	US-202217698199-A
Country	US
Kind code	B2
Filing date	Mar 18, 2022
Priority date	Mar 18, 2022
Publication date	Nov 18, 2025
Grant date	Nov 18, 2025

Abstract

Various techniques of reconstructing multiple Magnetic Resonance Imaging, MRI, images for multiple slices based on an MRI measurement dataset that is acquired using a simultaneous multi-slice protocol and undersampling and K-space are disclosed. A convolutional neural network can be used to implement a regularization operation of an iterative optimization for the reconstruction, i.e., an unrolled neural network or variational neural network. A combination with Dixon imaging, i.e., separation of multiple chemical species, is disclosed.

First claim

The invention claimed is:

1 . A method of reconstructing Magnetic Resonance Imaging, MRI, images, comprising:
obtaining at least one MRI measurement dataset, each one of the at least one MRI measurement dataset being acquired as a combined signal from multiple simultaneously excited slices using a simultaneous multi-slice protocol, an undersampling trajectory in k-space, and a receiver coil array, wherein signals from the multiple slices interfere in the combined signal, and wherein the simultaneous multi-slice protocol applies a phase modulation to multiple slices of the MRI measurement dataset; and based on the at least one MRI measurement dataset, performing an iterative optimization to separate the interfering signals and obtain, for each one of the multiple slices, at least one MRI image, wherein the iterative optimization comprises, for at least some iterations of multiple iterations of the iterative optimization, a regularization operation and a data-consistency operation to obtain respective current MRI images, wherein the data-consistency operation incorporates slice separation using sensitivity encoding or k-space correlations to disentangle contributions from individual slices and is based on differences between the MRI measurement dataset and a respective at least one synthesized MRI measurement dataset, the synthesized MRI measurement dataset being based on a k-space representation of a prior image of the multiple iterations and the undersampling trajectory, and wherein the regularization operation is implemented by multiple layers of a convolutional neural network for the multiple iterations, and the convolutional neural network comprises, for each convolutional layer, an associated circular padding layer.

2 . The method of claim 1 ,
wherein the layers of convolutional neural network jointly process the respective current MRI images for the at least some iterations.

3 . The method of claim 1 ,
wherein an input to the regularization operation comprises, for each iteration of the at least some iterations, a concatenation of the respective current MRI images concatenated along a channel dimension, and wherein convolutional kernels of the layers extend along the channel dimension.

4 . The method of claim 1 ,
wherein the at least one synthesized MRI measurement dataset is based on an antialiasing operation in image domain that is based on a sensitivity map associated with the receiver coil array.

5 . The method of claim 1 ,
wherein the at least one MRI measurement dataset comprises a sequence of MRI measurement datasets, the MRI measurement datasets of the sequence of MRI measurement datasets being acquired at multiple time offsets with respect to at least one excitation pulse and/or with respect to at least one refocusing pulse, and wherein the current MRI images of each iteration of the iterative optimization are associated with different ones of the multiple slices as well as different ones of the multiple time offsets.

6 . The method of claim 5 ,
wherein the current MRI images of each iteration are associated with different chemical species by considering a respective phase evolution map.

7 . A method of reconstructing Magnetic Resonance Imaging, MRI, images, the method comprising:
obtaining an MRI measurement dataset acquired for multiple slices using a simultaneous multi-slice protocol, an undersampling trajectory in k-space, and a receiver coil array, wherein the simultaneous multi-slice protocol applies a phase modulation to multiple slices of the MRI measurement dataset; and based on the MRI measurement dataset, determining multiple further MRI measurement datasets, each one of the multiple further MRI datasets being associated with a respective one of the multiple slices, for each one of the multiple further MRI measurement datasets, performing a respective iterative optimization to obtain a respective MRI image, wherein each iterative optimization comprises, for at least some iterations of multiple iterations of the iterative optimization, a regularization operation and a data-consistency operation to obtain respective current MRI images, wherein the data-consistency operation is based on differences between the respective further MRI measurement dataset and a respective at least one synthesized MRI measurement dataset, the synthesized MRI measurement dataset being based on a k-space representation of a prior image of the multiple iterations and the undersampling trajectory, and wherein the regularization operation is implemented by multiple layers of a convolutional neural network for the multiple iterations, and the convolutional neural network comprises, for each convolutional layer, an associated circular padding layer.

8 . A non-transitory computer program medium comprising program code that is executable by a processor, the processor, upon executing the program code being configured to perform the method of claim 1 .

9 . A device comprising a processor and a memory storing program code, the processor being configured to load and execute the program code and, upon executing the program code, to perform the method of claim 1 .

Assignees

• Siemens Healthcare Gmbh
• Univ New York
• Siemens Healthineers Ag

Inventors

• Raithel Esther
• Mailhe Boris
• Mostapha Mahmoud
• Fritz Jan
• Knoll Florian
• Nickel Marcel Dominik
• Körzdörfer Gregor
• Brinkmann Inge
• Nadar Mariappan S

Classifications

• G06T12/10 Primary

  Image preprocessing, e.g. calibration, positioning of sources or scatter correction · CPC title

• G06T2210/41

  Medical · CPC title

• G06T2211/424

  Iterative · CPC title

• G06T11/005 Primary

  Physics · mapped topic