Motion-Corrected MRI Reconstruction Using Weighted Rigid-Region Signals
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing retrospective motion correction techniques in MRI suffer from long computation times and inaccurate motion estimation due to the rigid-body assumption, which is often not valid for the entire field-of-view, leading to motion artefacts in magnetic resonance images.
Innovation Solution
A method that weights magnetic resonance data by reducing signal from non-rigid and independently moving body parts, estimating motion parameters from the weighted data using a forward model with motion parameters, and performing separate steps for motion estimation and image reconstruction, utilizing techniques like SAMER and TAMER.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If rigid-body motion model is used for motion estimation, then computational complexity is reduced, but motion estimation accuracy deteriorates due to non-rigid body parts moving independently
Solution Approach 1:
The patent segments the body region into multiple sub-regions based on expected motion characteristics (rigid vs. non-rigid). Motion parameters are estimated separately for each sub-region, allowing rigid-body assumption to be applied locally where valid while capturing independent motion in non-rigid areas without requiring full complex modeling of the entire body.
Solution Approach 2:
The patent applies different motion modeling approaches to different spatial locations. Rigid-body motion parameters are used for regions where this assumption holds, while separate motion parameters are estimated for non-rigid regions. This local differentiation resolves the contradiction by matching computational complexity to the actual motion characteristics of each body part.
2Quantity of substance
If full body region data is used for motion estimation, then motion parameters are estimated using all available signal, but non-rigid moving parts introduce errors and reduce estimation accuracy
Solution Approach 1:
The patent extracts and separates the contribution of non-rigid body parts from the total signal used for motion estimation. By identifying and removing the signal component from regions expected to move non-rigidly or independently, the motion parameters are estimated using only the rigid-body portion of the signal, thereby eliminating the source of estimation errors while still utilizing the maximum useful data.
3Reliability
If alternating optimization between image and motion parameters is performed, then joint estimation is achieved, but computation time increases significantly
Solution Approach 1:
The patent performs preliminary segmentation of the body region into rigid and non-rigid sub-regions before the main optimization process. Motion parameters for non-rigid regions are estimated separately in advance, allowing the alternating optimization to focus only on the joint estimation of image and motion parameters for the rigid-body regions. This preliminary separation reduces the computational burden while maintaining joint estimation accuracy where applicable.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The invention relates to a method for generating a motion-corrected magnetic resonance image dataset of a body region of a subject, the method comprising (a) receiving magnetic resonance data acquired of the body region (40, 42); (b) optionally receiving information on the body region (40, 42) covered by the magnetic resonance image dataset; (c) weighting at least part (48) of the received magnetic resonance data by reducing the signal originating from parts (42) of the body region which are expected to have undergone non-rigid and/or independent motion during the acquisition, thereby producing weighted magnetic resonance data (46); and (d) estimating the motion-corrected image dataset (x) by minimizing (56) the data consistency error between the magnetic resonance data (14) acquired in the imaging protocol and a forward model described by an encoding matrix, wherein the encoding matrix includes motion parameters (θi), Fourier encoding (F), and optionally subsampling (M) and/or coil sensitivities (C) of a multichannel coil array (7.1, 7.2), wherein the estimation includes at least one step of estimating (54) motion parameters (θi) from the weighted magnetic resonance data (46).