Systems and methods for motion-artifact reduction in magnetic resonance imaging

Patent metadata

Field	Value
Publication number	US-2024410965-A1
Application number	US-202318331885-A
Country	US
Kind code	A1
Filing date	Jun 8, 2023
Priority date	Jun 8, 2023
Publication date	Dec 12, 2024
Grant date	—

Abstract

Systems and methods are provided for reconstructing images from motion-affected k-space data. In one example, a method comprises obtaining k-space data of a spin echo magnetic resonance imaging (MRI) exam of a subject, the k-space data comprising a plurality of echo train lengths (ETLs), with each ETL comprising a subset of lines of the k-space data. The method further comprises identifying a subset of ETLs of the plurality of ETLs of the k-space data corresponding to a dominant pose of the subject, generating an undersampled version of the k-space data, the undersampled version including only the subset of ETLs, entering the undersampled version of the k-space data as input to a reconstruction model trained to output a reconstructed image based on the undersampled version of the k-space data, and displaying the reconstructed image on a display device and/or saving the reconstructed image in memory.

First claim

1 . A method, comprising:
obtaining k-space data of a spin echo magnetic resonance imaging (MRI) exam of subject, the k-space data comprising a plurality of echo train lengths (ETLs), each ETL comprising a subset of lines of the k-space data; identifying a subset of ETLs of the plurality of ETLs of the k-space data corresponding to a dominant pose of the subject; generating an undersampled version of the k-space data, the undersampled version including only the subset of ETLs; entering the undersampled version of the k-space data as input to a reconstruction model trained to output a reconstructed image based on the undersampled version of the k-space data; and displaying the reconstructed image on a display device and/or saving the reconstructed image in memory.

2 . The method of claim 1 , wherein identifying the subset of ETLs of the k-space data corresponding to the dominant pose of the subject comprises identifying the dominant pose by:
identifying one or more motion-affected ETLs of the plurality of ETLs of the k-space data, the one or more motion-affected ETLs affected by motion of the subject; classifying each ETL of the plurality of ETLs as belonging to one of a plurality of poses based on the one or more motion-affected ETLs; and identifying the dominant pose as a pose of the plurality of poses with a highest number of ETLs belonging to that pose relative to other poses of the plurality of poses, wherein each ETL classified as belonging to the dominant pose is included in the subset of ETLs.

3 . The method of claim 2 , wherein identifying the one or more motion-affected ETLs comprises entering the k-space data into a motion detection model configured to output an identification of each motion-affected ETL.

4 . The method of claim 1 , wherein generating the undersampled version of the k-space data comprises:
generating a mask configured to mask all ETLs of the k-space data other than the subset of ETLs; and applying the mask to the k-space data to form the undersampled version of the k-space data.

5 . The method of claim 1 , further comprising transforming the undersampled version of the k-space data to an undersampled image and entering the undersampled image along with the undersampled version of the k-space data as input to the reconstruction model.

6 . The method of claim 5 , wherein the reconstruction model is an unrolled reconstruction model trained with a plurality of training data triads, each training data triad including an undersampled k-space dataset missing a set of ETLs of k-space data, a training image generated from the undersampled k-space dataset, and a ground truth image, and wherein each undersampled k-space dataset for each training iteration is missing a different set of ETLs.

7 . The method of claim 6 , wherein each undersampled k-space dataset is generated from a respective full k-space dataset by masking one or more ETLs of the respective full k-space dataset, and wherein each ground truth image is generated from the respective full k-space dataset.

8 . A system, comprising:
one or more processors; and memory storing instructions executable by the one or more processors to:
obtain k-space data of a subject, the k-space data comprising a plurality of echo train lengths (ETLs) each comprising a plurality of lines of the k-space data;
detect one or more ETLs of the plurality of ETLs affected by motion of the subject;
identify a dominant pose of the subject based on the one or more ETLs of the k-space data affected by motion of the subject;
identify a subset of ETLs of the k-space data corresponding to the dominant pose;
generate an undersampled version of the k-space data, the undersampled version including only the subset of ETLs;
enter the undersampled version of the k-space data as input to an unrolled reconstruction model trained to output a reconstructed image based on the undersampled version of k-space data; and
display the reconstructed image on a display device and/or save the reconstructed image in memory.

9 . The system of claim 8 , wherein detecting one or more ETLs of the plurality of ETLs affected by motion of the subject comprises detecting one or more ETLs of the plurality of ETLs affected by motion of the subject via one of entering the k-space data into a trained motion detection model configured to output an identification of each motion-affected ETL or based on output received from one or more sensors configured to detect motion of the subject.

10 . The system of claim 8 , wherein identifying the dominant pose of the subject based on the one or more ETLs of the k-space data affected by motion of the subject comprises classifying each ETL of k-space data as belonging to one of a plurality of poses by identifying one or more ETLs of the k-space data affected by motion separated by one or more ETLs of k-space data not affected by motion.

11 . The system of claim 10 , wherein identifying the subset of ETLs of the k-space data corresponding to the dominant pose comprises identifying the dominant pose as a pose of the plurality of poses with a highest number of ETLs belonging to that pose relative to other poses of the plurality of poses, wherein each ETL classified as belonging to the dominant pose is included in the subset of ETLs.

12 . The system of claim 10 , wherein generating the undersampled version of the k-space data comprises generating a mask based on the one or more ETLs of k-space data affected by motion of the subject and/or one or more non-dominant poses of the plurality of poses and masking the one or more ETLs affected by motion of the subject and/or each ETL classified as belonging to the one or more non-dominant poses of the plurality of poses.

13 . The system of claim 8 , wherein the instructions are further executable to enter a corresponding undersampled image as input to the unrolled reconstruction model, wherein the corresponding undersampled image is generated by transforming the undersampled version of the k-space data.

14 . A method, comprising:
obtaining k-space data of a subject, the k-space data acquired with a magnetic resonance imaging (MRI) system according to an acquisition protocol and including a plurality of echo train lengths (ETLs), each ETL comprising a respective plurality of lines of the k-space data; detecting one or more ETLs of the k-space data affected by motion of the subject; in response to the detecting, identifying a dominant pose of the subject during execution of the acquisition protocol; identifying a subset of ETLs of the k-space data corresponding to the dominant pose; generating a mask configured to mask all ETLs of the k-space data other than the subset of ETLs corresponding to the dominant pose; applying the mask to the k-space data to form zero-filled k-space data; transforming the zero-filled k-space data to generate a corresponding image; entering the zero-filled k-space data and corresponding image as input to an unrolled reconstruction deep learning (DL) model trained to output a reconstructed image based on the zero-filled k-space data; and displaying the reconstructed image on a display device and/or saving the reconstructed image in memory.

15 . The method of claim 14 , wherein the unrolled reconstruction DL model is trained on a plurality of training data triads, each training triad comprising training zero-filled k-space data, a corresponding training image of the training zero-filled k-space data, and a ground truth image.

16 . The method of claim 14 , wherein training of the unrolled reconstruction DL model comprises:
obtaining a pluralit

Assignees

• Ge Prec Healthcare Llc

Inventors

• Chatterjee Sudhanya
• Goel Megha
• Joel Suresh Emmanuel Devadoss
• Patil Rohan Keshav
• C Florintina
• Shankapal Preetham

Classifications

• A61B5/055

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

• G01R33/56509 Primary

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

• 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

• G06T7/0012

  Biomedical image inspection · CPC title

• G06T2207/10088

  Magnetic resonance imaging [MRI] · CPC title