Deep learning based magnetic resonance imaging (mri) examination acceleration

Patent metadata

Field	Value
Publication number	US-2022128640-A1
Application number	US-202017083074-A
Country	US
Kind code	A1
Filing date	Oct 28, 2020
Priority date	Oct 28, 2020
Publication date	Apr 28, 2022
Grant date	—

Abstract

Systems and methods for deep learning based magnetic resonance imaging (MRI) examination acceleration are provided. The method of deep learning (DL) based magnetic resonance imaging (MRI) examination acceleration comprises acquiring at least one fully sampled reference k-space data of a subject and acquiring a plurality of partial k-space of the subject. The method further comprises grafting the plurality of partial k-space with the at least one fully sampled reference k-space data to generate a grafted k-space for accelerated examination. The method further comprises training a deep learning (DL) module using the fully sampled reference k-space data and the grafted k-space to remove the grafting artifacts.

First claim

1 . A method of magnetic resonance imaging (MRI) examination acceleration, the method comprising:
acquiring at least one fully sampled reference k-space data of a subject; acquiring a plurality of partial k-space of the subject; and grafting the plurality of partial k-space with the at least one fully sampled reference k-space data to generate a grafted k-space data for accelerated examination.

2 . The method of claim 1 wherein subsequent scanning by the MRI system comprises acquiring only the partial k-space of the subject and grafting the partial k-space with the fully sampled k-space data to generate the grafted k-space data for accelerated examination.

3 . The method of claim 1 wherein grafting the partial k-space with the fully sampled reference k-space data comprises grafting a missing structural information in the partial k-space from the at least one fully sampled reference k-space data.

4 . The method of claim 1 wherein grafting the partial k-space with the fully sampled k-space data is carried out before a deep learning (DL) module-based reconstruction of the grafted data.

5 . The method of claim 4 wherein grafting the partial k-space with the fully sampled k-space data before the deep learning (DL) module-based reconstruction of the grafted data provides structural information to the deep learning (DL) module to correct the grafting artifacts.

6 . A method of deep learning (DL) based magnetic resonance imaging (MRI) examination acceleration, the method comprising:
acquiring at least one fully sampled reference k-space data of a subject; acquiring a plurality of partial k-space of the subject; grafting the partial k-space with the at least one fully sampled reference k-space data to generate a grafted data for accelerated examination; and training a deep learning (DL) module using the grafted data and the fully sampled reference k-space data to remove the grafting artifacts.

7 . The method of claim 6 wherein the deep learning (DL) module comprises a smart loss function.

8 . The method of claim 7 wherein the smart loss function comprises a plurality of regularization terms and the smart loss function is configured to change a weight assigned to the plurality of regularization terms based on a relevance of the regularization term during training of the deep learning (DL) module.

9 . The method of claim 8 wherein the weights assigned to the each of the regularization terms is dynamically modulated based on a training loss after each imaging epoch.

10 . The method of claim 9 wherein dynamically modulating the regularization terms include modulating a structural similarity index measure (SSIM) loss, perceptual loss and a mean absolute error (MAE) loss.

11 . The method of claim 7 wherein the smart loss function is defined as:
=α× MAE +(1−α)×(1− SSIM )

wherein represent smart loss;
MAE is mean absolute error;
SSIM is structural similarity index measure; and
α is weight of the regularizer;

wherein α is updated after each epoch depending on the MAE and SSIM values.

12 . The method of claim 6 further comprising acquiring a plurality of MRI images comprising a simultaneous multi-slice reading for accelerated examination.

13 . A magnetic resonance imaging (MRI) system comprising:
at least one radiofrequency (RF) body coil adapted to transmit and receive radiofrequency (RF) signals to and from a subject; a transceiver module configured to digitize the signals received by the radiofrequency (RF) body coil; a control system configured to process the digitized signals and generate a k-space data corresponding to an imaged volume of the subject, wherein the MRI system is configured to acquire at least one fully sampled reference k-space data of the subject and a plurality of partial k-space of the subject; and a computer processor configured to graft the partial k-space of the subject with the fully sampled reference k-space data of the subject to generate a grafted k-space data for accelerated examination.

14 . The magnetic resonance imaging (MRI) system of claim 13 , further comprising a deep learning (DL) module employed on a computer memory, wherein the deep learning (DL) module is trained using the at least one fully sampled reference k-space data and the grafted k-space data.

15 . The magnetic resonance imaging (MRI) system of claim 14 , wherein the deep learning module is an artifact prediction network comprising a smart loss function.

16 . The magnetic resonance imaging (MRI) system of claim 15 , wherein the smart loss function comprises a plurality of regularization terms and the smart loss function is configured to change a weight assigned to the each of the plurality of the regularization terms based on a relevance of the regularization term during training of the deep learning (DL) module.

17 . The magnetic resonance imaging (MRI) system of claim 16 wherein the weights assigned to the each of the regularization terms is dynamically modulated based on a training loss after each imaging epoch.

18 . The magnetic resonance imaging (MRI) system of claim 17 wherein dynamically modulating the regularization terms include modulating a structural similarity index measure (SSIM) loss, perceptual loss, and a mean absolute error (MAE) loss.

19 . The magnetic resonance imaging (MRI) system of claim 14 wherein the computer processor is configured to graft the partial k-space of the subject with the fully sampled reference k-space data of the subject before the deep learning (DL) module-based reconstruction of the grafted data.

20 . The magnetic resonance imaging (MRI) system of claim 13 wherein the MRI system is configured to acquire a plurality of MRI images comprising a simultaneous multi-slice reading for accelerated examination.

Assignees

• Ge Prec Healthcare Llc

Inventors

• Chatterjee Sudhanya
• Shanbhag Dattesh
• Joel Suresh

Classifications

• G06F18/22

  Matching criteria, e.g. proximity measures · CPC title

• A61B5/055

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

• G01R33/5619

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

• G01R33/56509 Primary

  due to motion, displacement or flow, e.g. gradient moment nulling (G01R33/567 takes precedence) · CPC title

• G01R33/4818

  MR characterised by data acquisition along a specific k-space trajectory or by the temporal order of k-space coverage, e.g. centric or segmented coverage of k-space · CPC title