Granted Patent B2

US 12,493,093 · App. 17/914,435 · Granted Dec 9, 2025

Inventors: Kay Nehrke (Hamburg, DE); Peter Boernert (Hamburg, DE)

Assignee: Koninklijke Philips N.V.

G01R33/56563G01R33/4818G01R33/5608G06N3/08

Abstract

Claims (42)

1 . A medical system comprising:

a memory configured to store machine executable instructions and a trained neural network, wherein the trained neural network is configured to output corrected magnetic resonance image data in response to receiving as input a set of magnetic resonance images, each magnetic resonance image of the set of magnetic resonance images having been reconstructed from k-space data assuming a different spatially constant frequency off-resonance factor, and

a computational system configured to control the medical system, wherein execution of the machine executable instructions causes the computational system to:

receive the k-space data acquired according to a magnetic resonance imaging protocol;

arrange for reconstruction of a set of magnetic resonance images according to the magnetic resonance imaging protocol, wherein each magnetic resonance image of the set of magnetic resonance images is reconstructed assuming a different spatially constant frequency off-resonance factor chosen from a list of frequency off-resonance factors; and

receive the corrected magnetic resonance image data in response to inputting the set of magnetic resonance images into the trained neural network,

wherein the trained neural network is trained by a method comprising:

configuring a topology of the neural network to receive as input a set of magnetic resonance images each having a different spatially constant frequency off-resonance factor;

configuring the topology of the neural network to output corrected magnetic resonance image data;

receiving training data, wherein the training data comprises data sets comprising a deblurred magnetic resonance image reconstructed from a k-space data set using a spatially varying frequency off-resonance map and multiple training images each reconstructed from the same k-space data set using a different spatially constant frequency off-resonance map; and

producing the trained neural network by repeatedly training the neural network with each of the data sets.

2 . The medical system of claim 1 , wherein the corrected magnetic resonance image data comprises an inhomogeneity corrected magnetic resonance image.

3 . The medical system of claim 1 , wherein the corrected magnetic resonance image data is complex valued.

4 . The medical system of claim 1 , wherein execution of the machine executable instructions further causes the computational system to assemble a composite magnetic resonance image by selecting voxels from the set of magnetic resonance images according to a pixel mapping.

5 . The medical system of claim 1 , wherein the corrected magnetic resonance image data output by the trained neural network comprises a voxel mapping, wherein each magnetic resonance image of the set of magnetic resonance images has an identically sized voxel matrix, wherein the voxel mapping comprises a selection of one magnetic resonance image of the set of magnetic resonance images for each voxel of the identically sized voxel matrix.

6 . The medical system of claim 5 , wherein execution of the machine executable instructions further causes the computational system to assemble a B0 inhomogeneity mapping and/or a frequency off-resonance mapping by assigning the constant frequency off-resonance factor from the set of magnetic resonance images according to the voxel mapping.

7 . The medical system of claim 6 , wherein execution of the machine executable instructions further causes the computational system to reconstruct a B0 inhomogeneity corrected magnetic resonance image from the k-space data and the B0 inhomogeneity mapping or the frequency off-resonance mapping.

8 . The medical system of claim 1 , wherein the machine executable instructions further cause the computational system to assemble the corrected magnetic resonance image data by applying the trained neural network to portions of the set of magnetic resonance images using a spatial sliding window algorithm.

9 . The medical system of claim 1 , wherein the machine executable instructions are configured to cause the computational system to assemble the corrected magnetic resonance image data by applying the trained neural network to sub groups of the set of magnetic resonance images.

10 . The medical system of claim 1 , wherein each magnetic resonance image of the set of magnetic resonance images is complex valued.

11 . The medical system of claim 1 , wherein the medical system further comprises a magnetic resonance imaging system configured to acquire the k-space data from an imaging zone, wherein the memory further contains pulse sequence commands configured to control the magnetic resonance imaging system to acquire the k-space data, wherein execution of the machine executable instructions further causes the computational system to control the magnetic resonance imaging system to acquire the k-space data.

12 . The medical system of claim 1 , wherein the k-space data has a non-Cartesian sampling pattern.

13 . The medical system of claim 1 , wherein the k-space data has a spiral sampling pattern.

14 . A method of training a neural network, wherein the method comprises:

configuring a topology of the neural network to receive as input a set of magnetic resonance images each having a different spatially constant frequency off-resonance factor;

configuring the topology of the neural network to output corrected magnetic resonance image data;

receiving training data, wherein the training data comprises data sets comprising a deblurred magnetic resonance image reconstructed from a k-space data set using a spatially varying frequency off-resonance map and multiple training images each reconstructed from the same k-space data set using a different spatially constant frequency off-resonance map; and

producing a trained neural network by repeatedly training the neural network with each of the data sets.

15 . A non-transitory computer readable medium storing machine executable instructions for execution by a computational system controlling a medical system, wherein execution of the machine executable instructions by the computational system causes the computational system to:

receive k-space data acquired according to a magnetic resonance imaging protocol;

arrange to reconstruct a set of magnetic resonance images according to the magnetic resonance imaging protocol, wherein each magnetic resonance image of the set of magnetic resonance images is reconstructed assuming a different spatially constant frequency off-resonance factor chosen from a list of frequency off-resonance factors; and

receive corrected magnetic resonance image data in response to inputting the set of magnetic resonance images into a trained neural network, wherein the trained neural network is configured to output the corrected magnetic resonance image data in response to receiving as input a set of magnetic resonance images each having been reconstructed from k-space data assuming a different spatially constant frequency off-resonance factor,

wherein the corrected magnetic resonance image data output by the trained neural network comprises a voxel mapping, wherein each magnetic resonance image of the set of magnetic resonance images has an identically sized voxel matrix, wherein the voxel mapping comprises a selection of one magnetic resonance image of the set of magnetic resonance images for each voxel of the identically sized voxel matrix,

wherein execution of the machine executable instructions further causes the computational system to train the trained neural network by:

configuring a topology of the neural network to receive as input a set of magnetic resonance images each having a different spatially constant frequency off-resonance factor;

configuring the topology of the neural network to output corrected magnetic resonance image data;

receiving training data, wherein the training data comprises data sets comprising a deblurred magnetic resonance image reconstructed from a k-space data set using a spatially varying frequency off-resonance map and multiple training images each reconstructed from the same k-space data set using a different spatially constant frequency off-resonance map; and

producing the trained neural network by repeatedly training the neural network with each of the data sets.

16 . The non-transitory computer readable medium of claim 15 , wherein the corrected magnetic resonance imaging data comprises an inhomogeneity corrected magnetic resonance image.

17 . The non-transitory computer readable medium of claim 15 , wherein the corrected magnetic resonance image data is complex valued.

18 . The non-transitory computer readable medium of claim 15 , wherein execution of the machine executable instructions further causes the computational system to assemble a composite magnetic resonance image by selecting voxels from the set of magnetic resonance images according to a pixel mapping.

19 . The non-transitory computer readable medium of claim 15 , wherein execution of the machine executable instructions further causes the computational system to assemble a B0 inhomogeneity mapping and/or a frequency off-resonance mapping by assigning the constant frequency off-resonance factor from the set of magnetic resonance images according to the voxel mapping.

Assignments (1)

Priority Claims (1)

Continuity (1)

References Cited (17)